Coin Metrics’ State of the Network: Issue 31

Weekly Feature

Revisiting the Block Reward Halving Theory 

by Kevin Lu and the Coin Metrics Team

The upcoming block reward halving for Bitcoin, anticipated in May 2020, has already given rise to intense discussion about its potential impact. Several theories have been advanced to study supply-side dynamics surrounding halvings and its eventual impact (or lack thereof) on cryptoasset prices. But up to this point, the short history and infrequent nature of block reward halvings have prevented us from drawing strong conclusions. 

In this article, we evaluate commonly proposed theories surrounding block reward halvings through the lens of the most recent instance: Litecoin’s halving that occurred in August 2019. 

History of Block Reward Halvings 

As a byproduct of mining, proof-of-work networks distribute new coins into circulation through block rewards. Block reward halvings are a common characteristic built into the protocol of Bitcoin-derivative proof-of-work coins as a means to gradually reduce the supply issuance. Other proof-of-work coins adjust issuance every block and/or do not have halving-induced supply shocks. 

Bitcoin has experienced two block reward halvings and Ethereum has similarly experienced two block reward reductions. While Bitcoin halvings are mandated by the protocol, Ethereum has a less transparent issuance model. There has, however, been broad consensus for the Ethereum block reward to trend downward over time and for eventual issuance to be maintained at a low level as the network shifts to a proof-of-stake model. 

The few historical instances of block reward reductions have been associated with increases in price in Bitcoin and Ethereum within the first 1.5 years of the halving. For this reason, an upcoming block reward reduction is often cited as a reason to be bullish about an asset’s future price appreciation. 

Litecoin was initially launched in October 2011 as a fork of Bitcoin. The Litecoin issuance model is similar to Bitcoin’s except that blocks are produced every 2.5 minutes, and the total supply of Litecoin will eventually be capped at 84 million. Block rewards are halved every four years, assuming that blocks are produced on average every 2.5 minutes. 

Litecoin experienced its first block reward halving in August 2015 when the block reward halved from 50 LTC to 25 LTC. The second halving occurred in August 2019 when the block reward halved from 25 LTC to 12.5 LTC. 

Theory: Block Reward Halvings are Priced In 

There is intense discussion regarding whether block reward halvings are priced in. In one camp, proponents of the efficient market hypothesis (EMH) state that block reward halvings are mandated in the protocol and are well-known to all market participants far in advance. Since all information regarding halvings is already known, any impact that halvings have on supply-side dynamics or price should be fully reflected in the cryptoasset’s price. Furthermore, demand should not change since the issuance rate schedule and final supply are already known. Indeed, for Litecoin, the approximate dates of each of the halvings that have occurred (and all future halvings) were known when the protocol was launched in October 2011. 

Detractors of this theory state that not all market participants are aware of the existence or significance of halvings and will still act on this phenomenon when discussion increases as the halving date nears. While this may be true, this is not incompatible with the theory that halvings are priced in. The EMH does not require all market participants or even a majority of market participants to be aware of information for prices to fully reflect that information. In fact, all that is required is the existence of a small fraction of market participants who control enough capital and act upon this information to force prices to react. 

Detractors of this theory also point to empirical evidence which shows that historical halvings have been associated with positive returns, arguing that the historical halvings were not priced in, and the trend is likely to continue. This argument deserves further consideration. Here, we propose an alternative theory which attempts to reconcile the arguments made by both sides. 

On one hand, empirical evidence supports a narrow interpretation of the efficient market hypothesis — at the instant that a halving occurs, there is no immediate reaction in the price, either up or down, as no new information has been revealed. On the other hand, halvings can still impact prices over a period of several months to years, not because new information is revealed, but because market participants understand the impact of a reduction in miner-led selling (discussed more fully below) and anticipate actions by other traders in a reflexive, game theory-like manner. 

Litecoin’s price performance this year supports the theory that market participants anticipate halvings and act on halvings by bidding prices up in advance. Among major cryptoassets, Litecoin experienced one of the strongest price appreciations this year, increasing in price by 350% between January 2019 and July 2019. Over the same time period, this performance was second only to Binance Coin and far outpaced the gains seen in Bitcoin, Ethereum, XRP, and Bitcoin Cash. 

While all information regarding halvings is widely known, the evidence suggests that market participants still act upon halvings in part because of a strong narrative that halvings are positive for prices. Even if no logical cause-and-effect relationship exists between halvings and prices, the narrative (or belief that others will act on this narrative) can cause a self-fulfilling increase in prices as market participants attempt to enter positions in advance of other market participants doing the same thing. 

In Litecoin’s case, this game theory-like “anticipation trade” caused prices to run up and sell-off in advance of the halving as market participants attempted to time their entries and exits while anticipating the actions of others. Poor performance of Litecoin’s price after the halving could be explained by a continuation of traders unwinding their positions, although Litecoin’s year-to-date performance is still among the strongest of the major cryptoassets. 

Theory: Reduction in Miner-Led Selling Pressure 

The protocols of major proof-of-work coins adjust the difficulty of the proof-of-work calculation regularly to ensure that blocks are produced at regular intervals. Mining represents a near perfectly competitive industry that constantly seeks a steady state where miner revenue is only slightly above miner costs. When the state is far from this equilibrium, miners will either enter or exit the industry until the equilibrium is achieved. 

Miners earn revenue in the form of native units of the cryptoasset they are mining, while their variable costs (primarily in the form of electricity) must be paid in fiat currency. This, combined with the fact that miners must, in the long-run, operate at a profitability level that is only slightly above breakeven, means that they represent the single largest cohort of natural, consistent sellers. According to this theory, miner-led selling pressure is likely significant, small perturbations in the amount they sell can have an impact on prices, and a halving of this selling flow may eventually have a large impact on prices. 

To illustrate, Litecoin’s annualized supply issuance since the halving has been running at around 4%. At current prices, this means that miner revenue denominated in U.S. dollar terms is roughly $100 million per year. Under the theory that miners must sell nearly all of their holdings for fiat, this is equivalent to nearly $300,000 in natural selling pressure every day over the course of an entire year. 

Immediately prior to the halving, these figures were doubled — issuance was running at 8% and daily selling pressure was $600,000. Such a drastic reduction in compelled selling pressure should be supportive to prices going forward. Importantly, this reduction in selling flow occurs regardless of the degree to which the halving was priced in or anticipated by market participants. 

Detractors of this theory point to the relatively small amount of volume that can be attributed to miners compared to all trading volume that occurs within exchanges or on-chain. Indeed, compared to these much bigger figures, miners appear insignificant. Although data is not readily available on capital inflows and outflows into a cryptoasset, it is likely that the majority of trading that occurs within exchanges and on-chain do not represent a net inflow or outflow of fiat. Under this perspective, miner-led selling pressure may represent a large amount of net capital outflow of a cryptoasset. 

The Current State of Litecoin 

Although still one of the strongest performers this year, Litecoin’s performance after the halving has been quite negative, perhaps reflecting an unwinding of the “anticipation trade”. Any positive price pressure as a result of a halving of miner-led selling flow has yet to occur or has been overshadowed by the unwinding of the anticipation trade. 

The halving of the block reward combined with the decline in prices have caused a sharp decline in mining difficulty. Current difficulty is approaching two year lows and have already exceeded the lows during the depths of the market-wide sell-off in December 2018. 

Looking at the one-month change in mining difficulty reveals that after the halving, difficulty declined at the fastest pace ever. Litecoin mining is clearly in a state where a significant number of miners are operating with a loss and/or less cost-efficient miners are exiting the industry. 

Inventory management of miners is not a well-studied topic since access to this information is not available, but it stands to reason that each miner makes their own decision on how much of their block rewards to sell for fiat and when to sell it. Since miner variable costs are relatively constant in fiat terms, during periods of rising crypto prices, miners are required to sell less of their block rewards to cover their expenses. On the other hand, when crypto prices are falling, they are required to sell more. Under this theory, miners have a procyclical effect on the market, in that they further exacerbates price increases during periods of increase and vice versa. 

During periods of capitulation, miner-led selling flow is likely to be high. Miners may play games of chicken in which miners that are barely profitable are attempting to hold on, perhaps even willing to temporarily operate at a loss, until less cost-efficient miners exit the industry. Miners may even be willing to sell block rewards earned in prior periods that they kept on their balance sheet to be used in an attempt to outlast other miners. 

Although this has not yet occurred, Litecoin difficulty will likely eventually stabilize as all unprofitable miners exit the network. The culling of inefficient miners combined with the halving of the block reward should eventually result in a significant reduction in miner-led selling flow and could be supportive of prices going forward. 

Coin Metrics is actively researching miner flows by tracking the movement of block rewards from miners to other entities in the ecosystem. We hope to publish more information on this topic as our understanding grows. 

Current market value to realized value (MVRV), calculated as market capitalization divided by realized capitalization, is significantly below 1 and is approaching levels close to all-time lows. At current prices, most Litecoin holders are now underwater with values below their cost basis. Historically, low levels of MVRV for Bitcoin have marked good entry points and have accurately identified periods of undervaluation. 

Conclusion

Several theories have been advanced to explain the impact or lack of impact of block reward halvings and active debate continues over whether block reward halvings are priced in and whether reductions in miner-led selling are impactful to prices. The few historical instances of block reward halvings, including Litecoin’s recent halving this year, deserve continued study. Over the course of 2020, a number of major cryptoassets are scheduled to experience a block reward halving, including Bitcoin, Bitcoin Cash, Bitcoin Cash SV, and ZCash, all of which will provide additional opportunities to test the current set of halving-related theories. 

Network Data Insights

Summary Metrics

BTC showed some early signs of recovery this past week, after a mostly negative December. Although BTC market cap was still down week-over-week, BTC estimated hash rate grew by over 10%. ETH, LTC, and BCH hash rate, however, all dropped by at least 2.9%, with ETH experiencing the largest drop of 8.3%. This is a potential signal that BTC may be gaining even more ground on the major cryptoassets in the wake of the latest downturn. 

ETH, on the other hand, continues to slide, losing over 10% of market cap week-over-week. Despite the drop in market cap, ETH active addresses, transaction, and fees all increased week-over-week, which could be a sign that ETH usage is somewhat independent of market cap. 

Network Highlights

Bitfinex’s BTC supply has increased sharply over the last month, despite the decrease in BTC price. Bitfinex held 204,150 BTC on December 22nd, up from 154,996 BTC on November 22nd. 

BitMEX’s BTC supply has also been increasing over the last month, despite an email leak that exposed client email addresses. On November 1st, it was reported that BitMEX suffered a security breach that caused over 23,000 email addresses to be leaked. Since then, BitMEX’s BTC supply has increased from 246,937 to 255,741 back to near all-time highs.

Market Data Insights

This week, the market has staged a recovery off the lows experienced on December 18, although many assets are still down over the past week. Bitcoin (+5%) has been one of the few assets that increased in price in addition to TRON (+7%). Tezos (-13%) has continued to decline despite the market staging a recovery in the past few days. 

Ethereum Classic (+10%) is the strongest performing asset among this set. Huobi Token (+5%), Maker (+2%), and UNUS SED LEO (+1%) also sustained small gains for the week. 

CM Bletchley Indexes (CMBI) Insights

Crypto assets had a very volatile week, ending with mixed results across the market. For the first time in months, Bitcoin not only outperformed most of the market, but was one of few assets that returned positive weekly gains. This is reflected in the below charts where only the Bletchley 10 (~70% BTC) and Bletchley Total (~66% BTC) were positive performers. The negative performance of the even indexes, which lower the weight of BTC, are testament to the fact that most of the Bletchley 10’s and Total’s positive performance was due to BTC.

This pattern was a sustained phenomena during early 2019 when BTC rose from $4k to $14k with very few other crypto assets managing to keep up. It will be interesting to monitor this over the coming weeks to see if markets will repeat their early 2019 trend.

Coin Metrics Updates

This week’s updates from the Coin Metrics team:

  • Coin Metrics is hiring! We recently opened up 6 new roles, including Blockchain Data Engineer and Data Quality and Operations Lead. Please check out our Careers page to view the openings.

As always, if you have any feedback or requests, don’t hesitate to reach out at [email protected]

Subscribe and Past Issues

Coin Metrics’ State of the Network, is an unbiased, weekly view of the crypto market informed by our own network (on-chain) and market data.

If you’d like to get State of the Network in your inbox, please subscribe here. You can see previous issues of State of the Network here.

Check out the Coin Metrics Blog for more in depth research and analysis.

Coin Metrics’ State of the Network: Issue 30

Weekly Feature

Ranking Crypto Assets by Auditability 

by Antoine Le Calvez and the Coin Metrics Team

In Medieval times, landed estates’ accounts were read out loud to a person charged by the local ruler to ensure that their steward had not swindled them. As the primary role of this person was to listen (audire in latin), they became known as auditors.

A parallel can be drawn between the genesis of auditing and crypto asset nodes that do not take part in that asset’s consensus; they primarily serve to listen to the peer-to-peer network and validate that everything is unfolding according to the protocol’s specification: they are auditing the network.

Coin Metrics takes this approach one step further: as well as running nodes for most major crypto assets, we also extract raw blockchain data from them to rebuild the asset’s ledger independently. This allows us to compute many of our metrics (for example, realized capitalization) but also to more deeply analyze each asset.

Our version of auditing a crypto asset is being able to rebuild its ledger (the mapping of who owns what) independently, for any point in time, using data provided by the asset’s protocol implementation and, using this reconstructed ledger, verify that its supply matches what it should be according to the protocol’s specification.

In this feature, we’ll dig into how we audit crypto assets, what difficulties can be encountered during this exercise and what can be learned from it. Finally, we’ll attempt a ranking of assets along two different dimensions of auditability: node operation and ledger reconstruction. Note that those rankings are arbitrary and that they reflect our subjective experience of working with each asset.

Crypto Assets Auditability Dimensions

Node Operation

The first step to audit a crypto asset is to synchronize a node that understands its protocol. The node is software that implements the asset’s protocol, connects to the peer-to-peer network, and downloads and verifies the blockchain. Depending on the asset, hardware and time requirements vary. For some assets, there are different node variants to choose from, either because parallel implementations of the same protocol exist, or there are various configurations possible depending on the user’s needs. 

In the case of an auditor, it’s better to use the configuration that will give access to the most data possible, sometimes referred to as an archive or full history node. Depending on what level of audit is needed (current or historical audit), a configuration that doesn’t store all historical data could suffice, for example, pruning mode in Bitcoin Core.

Synchronization

The first thing a crypto asset node does is synchronize itself with the current state of the network. Depending on the asset and configuration, this could involve downloading a current or recent snapshot (for example, downloading a recent Ethereum’s state trie using fast sync) or downloading the entire history of the block chain and replaying it in its entirety (like what Bitcoin does). While the former is much faster, the latter is better suited for auditing as it makes historical data available.

For some assets and node configuration, completing this process requires a lot of patience (and expensive hardware). As an example, synchronizing a full archive EOS node took more than one month and required a machine with terabytes of NVMe SSDs, some of the fastest storage available.

Finally, a handful of assets, most notably Ripple, require such huge amounts of storage (tens of TBs) that they are impractical to run and synchronize.

Normal Operation

Having a node synchronized with the rest of the network is only the first step of the audit process. The node also needs to stay synchronized in order to be able to audit the network on an ongoing basis. While this sounds easy to achieve, a lot of node software fails to stay synchronized and sometimes experiences catastrophic failures during normal operations.

An example would be that our EOS node needs to be shut down very carefully otherwise it risks corrupting its own database.

Code Audit

A major part of being able to audit an asset is being able to understand how it works. Since the implementation of each protocol lives in code, being able to read it is paramount to getting a deep understanding of the intricacies of each protocol. While some assets have deep, detailed specifications, they might differ ever so slightly from the actual implementation.

In that dimension, of the more than 35 unique nodes that Coin Metrics manages, one is unique: Binance Chain. It’s the only one whose source code is not available: only signed binaries are provided to would-be node maintainers.

Despite there being some documentation on Binance Chain’s protocol, it’s not detailed enough to be able to reconstruct its ledger using the data exported from the node.

Extracting Ledger Data

The last step before being able to reconstruct an asset’s ledger is to be able to extract the necessary data from the node. Few assets’ nodes offer the ability to directly obtain the latest (or historical) ledger; most of the time, it has to be rebuilt by replaying the asset’s history.

In order to do this, we need to be able to extract all the data required from the node. All nodes offer some form of API to access this data, with varying levels of user-friendliness, documentation, and completeness.

A key to easy auditability for an asset is to have as few ways as possible to credit or debit accounts. For example, Bitcoin only has 1 way to credit native units (the mining of an unspent output) and one way to debit native units (spending a previously unspent output). The more distinct things there are to track, the harder it gets. The more ways there are to credit/debit native units, the more likely that data for some is not easily accessible (an example would be some fees charged by the Binance Chain DEX).

Node Operation Ranking

We ranked full nodes in several tiers (A, B, C, and F) depending on their ease of synchronization, update, and maintenance. Here is our ranking for the top 10 assets by market cap (Coin Metrics doesn’t operate Ripple or Stellar nodes, but relies on the APIs provided by both Ripple and the Stellar Foundation).

Since historical ledger auditing for Ethereum requires a node with tracing and they take a long time to synchronize, it gets a B.

Omni gets a B as, even though it’s a modified Bitcoin Core and is therefore easy to run and sync, it has many different ways to credit and debit native units, each accessible through its own API endpoint requiring some effort to understand and put together.

Despite having a very clean accounting model for an asset of its complexity, EOS gets an F due to the complexity of extracting all the necessary data to run a complete audit. An archive node running with an extra plugin is required, documentation for which is scarce. Finally, the amount of data to sift through in order to get credits and debits of EOS is very large which makes it impractical.

Binance Chain gets an F for two linked reasons: it has a very complex fee schedule for its DEX and it’s closed-source which makes reverse engineering this schedule very hard. 

Ledger Reconstruction

Once the node is synchronized, running correctly and its data exported, the task of rebuilding its ledger can begin.

The way this is accomplished depends on the asset’s accounting model: UTXO-based (like Bitcoin and its derivatives) or account-based (Ethereum and many others). The ledger for UTXO based chains’ consists of a set of unspent transaction outputs (UTXO). Therefore, rebuilding the asset’s ledger consists of tracking which transaction outputs are unspent. For account based chains, the ledger is closer to a mapping of accounts to their balance. In order to rebuild it, auditors have to keep track of credits and debits for each account.

UTXO Sets

Tracking unspent outputs is done by replaying the block chain: each block creates new outputs and spends old ones. After having replayed the whole chain, the outputs remaining unspent make up the asset’s ledger. Summing up their value gives the asset’s supply.

As straightforward as it seems, there are still some nuances to keep in mind. We detailed some of Bitcoin’s supply idiosyncracies in a recent installment of this newsletter. For example, some special outputs (OP_RETURN) are not counted towards the supply; Bitcoin’s genesis block’s output doesn’t count either.

Account Ledgers

For account-based chains like Ethereum, we have to track credits and debits for each account on the chain. Given the complexity of some chains driven by the existence of smart contracts, this can get complicated very quickly. 

Another difference between UTXO and account-based protocols is that transactions in the former are more explicit about how they affect the ledger:

  • A UTXO transaction consists of two parts: which previously unspent outputs it spends and what new outputs it creates. UTXO transactions describe how they are going to change the asset’s ledger. On all UTXO protocols, only valid transactions are included in valid blocks. There cannot be partially applied UTXO transactions.
  • Account-based transactions, especially smart contract invocations, only describe the intent of the user, not its effects on the ledger. To be able to recover their impact on the ledger, nodes often need to be run with what is often called “tracing”. Tracing consists of recording exactly what each transaction’s impact on the ledger was. Non-tracing nodes just apply transactions without making detailed information available on a per-transaction basis. Furthermore, transactions can be included in a block without being completely executed.

For some assets, there are additional factors listed below to take into account.

Implicit Block Rewards

Most protocols include a reward for whoever (miner or staker) added a new block to the chain. The reward’s amount is the implementation of the asset’s monetary policy as this is how most assets generate new coins. Therefore, tracking this is essential to determine an asset’s supply.

In most UTXO assets, this amount is visible in each block, as the first transaction of every block encodes this reward being given to the miner. However, this is not always the case for account-based ones, most notably Ethereum, where blocks just encode which account should get the reward. The exercise of determining the reward’s amount is left to nodes and auditors.

As Ethereum block reward changed twice so far (5 ETH→3 ETH→2 ETH), auditors and nodes have to keep track at which blocks those changes happened.

Implicit Ledger Edits

While the block reward being implicit is only a minor inconvenience, there are other times where an asset’s ledger can change without those changes being made explicit by the transaction or blocks.

For example, following the DAO attack, Ethereum experienced a hard fork to return funds withdrawn from the DAO to another address not controlled by the person behind the unexpected DAO withdrawals. Those changes to the ledger are implemented in the code run by the nodes, not in a transaction nor in a block. Unfortunately for auditors, neither the block raw data nor the tracing data indicate that those changes occurred. The only way to capture those credits and debits is to find the hardcoded list of affected addresses and emulate what edits the code ran over the ledger.

This type of software-only changes to the ledger also recently happened with Tezos. In the making of the Babylon hard fork, someone realized that thousands of accounts would have to be recreated by users in order to access some of their funds: an operation that costs 0.257 XTZ. In order to avoid having thousands of users pay this fee, the hard fork was changed to credit affected accounts with a tiny amount of XTZ to “recreate” them at a lesser cost. Once again, auditors were out of luck as this subtle change was not documented anywhere.

A final example of how implicit ledger edits make auditing assets more complicated lies with ERC-20 tokens. ERC-20 is a standard interface for Ethereum smart contracts. It lists a few methods and events they should implement to maximize compatibility with existing wallet software and explorers. In practice, ERC-20 is just a standard and lacks strong guidance on the semantics of its events. Developers are free to stray from it, leaving auditors like Coin Metrics the hard task of piecing together the asset’s full transactional history. For example, token generation and burn events are often recorded using token-specific methods (if recorded at all).

Ledger Reconstruction Ranking

We’ve ranked each of the top 10 assets in several tiers (A, B, C, and F) depending on the ease of rebuilding their ledger for any point in time:

Ripple and Stellar are not graded as we do not reconstruct their ledger completely independently as we rely on APIs provided by third parties.

Bitcoin and its derivatives (Bitcoin Cash, Litecoin, BSV) receive an A as tracking their UTXO set is a simple task.

The Omni protocol, used by Tether to operate on the Bitcoin blockchain, receives a B because it has many different ways to move native units which makes it harder to track.

EOS also receives a B because its sheer scale (tens of millions of transfers per day) makes it unwieldy to audit.

For Ethereum, the current state of the tools we use don’t allow a full reconstruction of the ledger only using the data exposed by tracing, the changes made in the DAO hard fork have to be manually implemented. It therefore receives a C.

Finally, Binance Chain received an F because we could not rebuild Binance Chain’s ledger using our node’s data, due to the complexity of its DEX and the absence of source code to look up the details of its implementation.

Validating the Supply

Once the historical supply of an asset can be computed, it still has to be validated against what it should be to ensure it’s correct. This category doesn’t have a ranking as it’s binary: either we can validate the supply or we can’t due to being unable to reconstruct its ledger. There’s been no case of an asset for which we couldn’t determine what the expected supply should be.

A few assets’ nodes let users query what the actual supply is (most notably Bitcoin and its derivatives) which makes this task easy.

Example of fetching Bitcoin’s supply

For the other assets, most of the time, they have a straightforward issuance (or none at all). Given the formula that gives the expected supply at a given height and the supply we found rebuilding the ledger, we can verify whether there’s been any unexpected inflation. It’s possible to find a lower supply than the formula’s due to users or entities burning funds.

Some assets, due to their use of privacy features, sacrifice supply visibility for transactional privacy. One interesting example is Zcash which has the particularity of having both a visible part of its supply (so-called transparent supply) and a private part (so-called shielded supply). Auditors have perfect visibility into the makeup of the transparent ledger but can only have an estimation of the size of the private part. Movements in and out of the private supply can be tallied to estimate its size, but if there’s an inflation-causing bug happening inside fully private transactions, it is undetectable by auditors.

Conclusion

It’s important to note that a lot of the issues we encounter in auditing crypto assets lie with the nodes and tooling available to users, not with the protocol themselves. We hope that over time, they will improve to make auditing assets easier, and we are already starting to see this today. Coin Metrics may therefore revisit this exercise in the future. 

While this exercise of validating an asset’s supply independently may seem futile, it nonetheless led to several discoveries of hidden inflation.

Coin Metrics detected that Bitcoin Private supply had anomalies using these auditing techniques:

Stellar suffered an inflation bug visible through supply auditing that was later remediated by the Stellar Development Foundation burning some of its own supply. As a further argument to why this process matters, the total supply as reported by the Stellar blocks headers diverged from the actual supply of XLM obtained by summing up all the account’s balances.

These two examples once again highlight the relevance of one of the industry’s adages: do not trust, verify.

Network Data Insights

Summary Metrics

The major crypto assets continued to downslide over the past week. BTC and ETH market cap both fell by over 2.6% and realized cap dropped 0.2% and 0.6%, respectively.

ETH active addresses count, however, grew 21.3% week over week. This large increase is potentially related to Ethereum’s recent Istanbul hard fork.  ETH transfer and transaction count were also up over the past week, despite a large decrease in transfer value.

XRP transactions, however, continued to decline after a large surge in recent weeks. XRP active addresses dropped by over 13%, far more than the other four assets in our sample. 

Network Highlights

Bitcoin SV (BSV) transaction count hit new yearly highs this past week. However, as we reported back in SOTN Issue 8, a majority of BSV transactions are being used for data storage, and do not involve monetary transfers. At one point, over 90% of all BSV transactions were being initiated by WeatherSV, an app that records daily weather data onto the blockchain for a small fee. There is now another BSV app that is generating a large number of transactions: Preev, which allows users to write once-per-minute price updates for BTC-USD onto the BSV blockchain ledger.

The number of addresses with small balances of Tezos (XTZ) has been growing since early November. The following chart shows XTZ addresses with balance greater than X, where X ranges from 0.1 XTZ to 1K XTZ. 

On November 7th, Coinbase introduced Tezos staking, and added Tezos to Coinbase Earn, which allows users to earn up to $6 worth of XTZ by completing some lessons to learn how it works. There were a little over 82,000 addresses with at least 0.1 XTZ (worth about $0.16 at current prices) on November 6th. As of December 15th, there are over 107,000 addresses with at least 0.1 XTZ.

Market Data Insights

Crypto markets continued with a steady but moderate decline in prices over the past week, with a few noticeable exceptions. After consistently underperforming other major assets for the majority of this year, ZCash has achieved two consecutive weeks of positive price growth. 

ZCash has underperformed other major assets this year in part because of its high issuance rate. As a relatively young proof-of-work asset modeled off of Bitcoin’s issuance schedule, ZCash was launched in October 2016 with no premine and a block reward of 12.5 ZEC per block. ZCash has yet to experience a block reward halving and its annualized issuance rate is relatively high compared to other proof-of-work assets. 

Notably, ZCash annualized issuance rate has been declining rapidly as the constant block reward represents a smaller percentage of its total supply. Over the course of 2019, the annualized issuance rate has declined from 47% to 32%. By late 2020, the annualized issuance rate will further decline to 25% just prior to the block reward halving and will then decline to 12.5% immediately after the halving. These reductions in miner-led selling pressure should be broadly supportive to prices assuming that demand for ZCash remains constant. 

Other notable performers this week include ChainLink (+1%), Tezos (+5%), and Cosmos (+15%). All three assets are extending significant gains over the course of this year. 

CM Bletchley Indexes (CMBI) Insights

As the year comes to a close it is time to reflect back on where we have come from a year ago to put the current market performance into context. Despite the recent weakness in the market, crypto asset performance has been relatively strong over 2019 after an abysmal 2018. With renewed confidence in the market, it seems most of the attention has gone into larger assets, with Bitcoin close to finishing the year up over 80%, the Bletchley 10 (large cap) returning 52% over the year, the Bletchley 20 (mid cap) returning 28% and the Bletchley 40 (small cap) falling 40%. 

However, after a soft week for crypto asset markets that saw prices dwindle across the board, all Bletchley Indexes finished the week ~5% down. There was very little differentiation in performance between small, mid and large-cap assets as is evidenced by the below charts, demonstrating very little variance in weekly returns. 

Coin Metrics Updates

This week’s updates from the Coin Metrics team:

  • Coin Metrics is hiring! We have 6 roles posted including Blockchain Data Engineer and Data Quality and Operations Lead. Please check out our Careers page to view the openings.

As always, if you have any feedback or requests, don’t hesitate to reach out at [email protected]

Subscribe and Past Issues

Coin Metrics’ State of the Network, is an unbiased, weekly view of the crypto market informed by our own network (on-chain) and market data.

If you’d like to get State of the Network in your inbox, please subscribe here. You can see previous issues of State of the Network here.

Check out the Coin Metrics Blog for more in depth research and analysis.

Coin Metrics’ State of the Network: Issue 29

Weekly Feature

Analyzing the Supply Distributions of Projects with On-Chain Governance

by Nate Maddrey and the Coin Metrics Team

It takes a whole network of people to make a public blockchain work. You need miners to validate and secure the ledger. You need developers to maintain and update the protocol’s code. And you need users and investors who use the blockchain and value its native crypto asset.

Often times, these different groups’ interests are aligned. For example, investors and miners typically both want the price of a cryptocurrency to increase. But in other cases, like in the debate over whether Bitcoin should adopt SegWit, different constituents can have vastly different opinions. 

Most people would probably agree that for a blockchain to be successful in the long run it needs some sort of process for aligning on things like protocol upgrades and economic policies. But beyond that, there are countless different opinions about how blockchains should be governed, and even if they should be governed at all. 

In this piece, we analyze a specific subset of blockchain governance called on-chain governance, and look at the supply distributions of three different projects that make use of it.

The Governance Debate

Broadly defined, blockchain governance refers to the processes used to manage how blockchains change over time. This includes changes to the core protocol, but can also include changes to any other part of the blockchain’s ecosystem. The specific governance rules and procedures are unique to each blockchain.

There are two general forms of blockchain governance: off-chain and on-chain governance. In on-chain governance, voting is recorded on the blockchain’s ledger and is therefore publicly viewable and auditable. There are different ways that this on-chain voting can be structured, but it typically involves staking cryptocurrency to express support for an issue. 

Off-chain governance, on the other hand, does not involve recording votes on the blockchain ledger itself. Off-chain governance is more nebulous than on-chain governance and can involve many different forms of coordination and signaling, including forum discussions, informal polls, and formalized debate. But ultimately off-chain governance comes down to one critical decision: “voting” through deciding whether or not to change the protocol. 

Bitcoin and Ethereum both use forms of off-chain governance. Whenever there is a protocol upgrade, full node maintainers must decide whether they want to adopt the new changes. If they agree with the changes, they update their software. The version of the software endorsed by the majority of the community is considered the “main chain.” Full nodes maintainers can decide at any time to use a different version of the protocol than the main chain and thus create a fork as long as miners also run that version of the protocol. 

Off-chain governance typically works relatively smoothly for scheduled protocol updates, like those introduced in Bitcoin Improvement Proposals (BIPs) or Ethereum Improvement Proposals (EIPs). But it can sometimes lead to drawn-out, contentious hard forks, where two (or more) different factions of the community have drastically different ideas of how the chain should proceed. Notably, this occurred after the Ethereum DAO hack, which resulted in the split between Ethereum and Ethereum Classic, and the split between Bitcoin and Bitcoin Cash over what maximum size, if any, blocks should have.

These contentious hard forks led other blockchains to search for new, alternative ways to structure blockchain governance. Partially as a response to the perceived ineffectiveness of off-chain governance, on-chain governance began to gain favor. Starting around 2016, blockchains like Decred and Tezos began to experiment with forms of on-chain governance.

On-Chain Voting 

On-chain governance allows users to vote directly for the changes they want made to the core protocol code. Additionally, some blockchains allow the option to vote for economic changes such as setting fee price. These votes can be polls, that serve as a way for the community to coordinate and signal their opinions, or they can be binding, and immediately be put into effect as soon as the vote ends. 

Most on-chain votes involve some form of staking. In order to vote, users must “stake” a certain amount of crypto assets, which locks the staked assets into escrow until the voting period ends. If users are caught cheating or committing voting fraud they lose some or all of their stake, depending on the specific rules of the blockchain.

On-chain governance is pitched as solving several problems with off-chain governance. It is described as being more efficient, allowing for quicker decisions on key issues (although it’s still up for debate whether this is a good or bad thing). It also, in theory, takes the power out of the hands of miners and other powerful node operators and puts it into the hands of the token holders. 

However, on-chain governance is not without its problems. Most staking protocols give some form of advantage towards large balance holders. This means that the degree of concentration in supply distribution takes on increasing importance for governance systems that rely on on-chain staking. If supply is mostly held by a small number of addresses, those addresses gain huge influence over the governance decision-making process, which can lead to a plutocracy.

In the following sections, we analyze the supply distributions for three different projects that use on-chain governance: MakerDAO, Decred, and Tezos. 

Supply Distributions

MakerDAO

MakerDAO, the Ethereum-based decentralized finance platform, uses both polls and binding on-chain voting to govern the DAI stablecoin and to make other decisions for the MakerDAO ecosystem. MakerDAO polls are “symbolic votes used to poll community sentiment towards specific models or data sources.” The Maker Foundation uses polls to gauge community sentiment for different issues, like adjusting the DAI stability fee. MakerDAO also has “Executive Votes.” Executive votes are binding decisions, and the winning option is enacted once voting ends.  

Currently, MakerDAO votes are directly proportional to stake. When a poll or executive vote is opened, users vote by staking their MKR on a specific side of the issue. The side with the highest amount of MKR staked by the end of the voting period wins. For Executive Votes, the voting process continues until  “the number of votes surpasses the total in favor of the previous Executive Vote.”

MakerDAO’s most recent Executive Vote, which enabled the community to adjust the DAI debt ceiling and SAI stability fee, was executed on December 6th. The vote passed with a total of 51,910 MKR staked in support. Two addresses accounted for over 66% of the total MKR staked, according to mkrgov.science. Furthermore, MakerDAO has had several votes where a single address has accounted for over 90% of the winning stake, including a vote in October where a single address contributed over 94% of the winning vote

Therefore, supply distribution is an important consideration when analyzing MakerDAO’s governance. Holders with large balances can have an outsized influence on votes.

To analyze supply distribution, we look at the number of addresses that hold above a certain fraction of the total supply, ranging from 1/1,000th of supply to 1/10,000,000,000th of total supply. 

However, MKR’s distribution is a bit skewed, because there are several MKR addresses that pool a large number of tokens for staking. Specifically, the Maker MultiSig Contract and Maker Governance Contract hold 219,296 and 137,084 MKR, respectively. Therefore we excluded those two addresses to create the following adjusted supply distributions, which brings the total MKR supply from 999,999 to 643,609. 

After this adjustment, 1/1,000th of the total adjusted supply of MKR, is 643.61 MKR (equivalent to about $324K at current MKR price), and 1/10,000,000,000th of the total supply of MKR is 0.000064 MKR (equivalent to about $.03). 

The below table shows stats for addresses holding greater than or equal to 1/X of the total supply of MKR, where X ranges from 1,000 to 10,000,000,000. For example, there are 102 addresses that hold at least 1/1,000th of total MKR supply (i.e. at least 643.61 MKR). These 102 addresses, which are only 0.58% of the total amount of addresses, collectively hold over 509,991 MKR, which is over 79% of the total supply. Given that the most recent vote required 51,910 MKR, these top 0.58% of addresses could be able to control MakerDAO votes if they cooperated.

Note: MKR’s supply is adjusted in the following chart to exclude two Maker Foundation contracts, as explained above.

The number of addresses holding smaller amounts of MKR has been growing steadily over time. But the number of addresses holding greater than 1/100,000th of the total MKR supply (i.e. holding at least 6.44 MKR) has remained relatively flat. Therefore the majority of voting power has been concentrated in a relatively small number of addresses over the course of most of MakerDAO’s history.

Decred 

Decred’s on-chain voting system also uses staking but does not have staking directly proportional to votes. Instead, Decred users stake DCR in exchange for voting tickets, which gives them an opportunity to both vote and validate the previous block. Every block randomly picks five tickets to vote. Staked DCR is locked until the ticket is selected to vote, and is then returned to its owner along with a PoS reward (source). 

Although votes are not directly proportional to the amount of DCR staked, there is still an advantage to having a relatively large amount of DCR. Decred voting tickets each currently cost 144.71 DCR each, which is equivalent to about $2,980. There are currently 28,158 addresses that hold at least 107.83 DCR (representing about 24% of the total DCR supply), which is a little less than the current price of a voting ticket.

Jumping up one level, there are only 792 addresses that hold at least 1/10,000th of DCR supply (i.e. at least 1,078 DCR) which is equivalent to about $22,260 at current price. These 792 addresses hold over 55% of total DCR supply. Given that this group of addresses holds over 50% of supply, they could conceivably control Decred voting if their owners cooperated. However, Decred’s voting ticket system makes this much more difficult to do than in MakerDAO. Since Decred voting tickets are randomly selected, a single user cannot come in and immediately dictate the vote.

The number of addresses holding greater than 1/100,000th of the total DCR supply (i.e. greater than 107.83 DCR) has almost doubled over the past year. Similarly, the number of addresses holding smaller amounts of DCR has also grown. The number of addresses holding more than 10,783 DCR (1/1,000th of the supply) decreased over the course of the year, dropping from 73 to 64. 

Tezos 

The Tezos governance process is more similar to Decred than to MakerDAO. Tezos bakers (which is Tezos’ version of miners) vote on issues by staking their tokens. Bakers validate blocks in addition to voting on governance issues, but we will be focusing solely on their voting duties for this analysis.

Additionally, Tezos users can delegate their tokens to bakers, which allows the baker to vote on that user’s behalf. This allows users with less than 8,000 XTZ to still participate (indirectly) in the baking process. 

In order to be eligible as a baker, Tezos users must stake a certain amount of XTZ, known as “rolls.” The more rolls a baker has, the higher the chance they have to be selected as a block validator. Tezos rolls currently cost 8,000 XTZ. 

There are currently 5,424 Tezos addresses that hold at least 7,399 XTZ (i.e. at least 1/100,000th). These 5,424 addresses collectively hold over 96% of the total Tezos supply. Furthermore, there are 883 addresses that hold at least 1/10,000th total supply (73,999 XTZ), equivalent to about $118K. These 883 addresses hold over 82% of the total supply.

The number of addresses holding greater than 1/100,000th of the total XTZ supply (i.e. greater than 7,399 XTZ) has grown at a slower pace than DCR, increasing from 4,625 at the start of the year to 5,424 today. The number of addresses holding more than 1/1,000,000,000th of the total supply of XTZ (i.e. at least .74 XTZ) has grown significantly faster, jumping from 24,556 on January 1st to over 44,000. 

Future of On-Chain Voting

Governance, in general, is an incredibly difficult problem to solve; societies have been trying to govern themselves for thousands of years, to varying degrees of success. 

On-chain governance is still at very early stages. MakerDAO, Decred, and Tezos are still in relatively experimental phases, and their governance systems will certainly evolve as the projects progress. 

Many people are currently working on ways to improve on-chain governance and blockchain governance in general. For example, quadratic voting has been put forth as one potential improvement.  Furthermore, several projects are working on solutions for online identities, which could also improve the on-chain voting process by aligning voting power to the individual rather than to their stake in the network.

Token distribution will likely continue to be an important metric to monitor, especially as long as staking is involved in on-chain voting. We will continue to track these projects moving forward.

Network Data Insights

Summary Metrics

Most major crypto networks stabilized over the past week, with BTC and ETH market cap dropping 0.3% and 1.7%, respectively. Transfer value and fees, however, dropped significantly over the past week. BTC and ETH fees both dropped to weekly lows at the end of the week. On December 8th, BTC had $108,452 total daily fees, while ETH had $39,987.

XRP transactions decreased by 34.1% over the past week, after staggering growth the two previous weeks: XRP transactions increased by over 50% last week after being up over 121% the week before. 

Network Highlights

Coin Metrics recently released updates and enhancements to our exchange flow metrics. The following two charts are generated using the updated metrics.

The PlusToken scam has been back in the news recently as the wallet continues to shed coins. Reports going back to August claim that Huobi has processed a large amount of PlusToken withdrawals, and on-chain data appears to back that up. 

The following chart shows our estimate of Huobi’s BTC supply over the past year. Huobi’s supply started increasing mid-year and shot up towards the end of the year, which coincides with reports of the PlusToken sell-off. 

Poloniex has also been in the news for spinning out of Circle, after being acquired in early 2018. Poloniex’s supply of BTC and ETH plummeted after Circle’s acquisition, and are now at their lowest levels since early 2016. It remains to be seen whether Poloniex can recover after the recent spin out. 

Ethereum recently completed its scheduled Istanbul hard fork. The protocol upgrade made several protocol changes, including changes to some gas costs, which some developers were apparently not fully prepared for. 

The below chart shows the percent of Ethereum contract calls that ran out of gas, from December 6th to December 9th. Contracts running out of gas spiked up to over 1% on December 8th, immediately after Istanbul went live.

Notably, this change appears to have affected Gemini. Gemini has not swept user deposits into its hot wallet since the launch of the Istanbul fork. Each of their attempts has resulted in an “out of gas” error, as noted by Coin Metrics’ resident data archaeologist Antoine Le Calvez. Gemini later fixed the issue, whose root cause was a bad estimate of post-fork gas cost to sweep user funds.

Market Data Insights

This week, ZCash (+5%) and Bitcoin Cash SV (-7%) saw significant moves in price while the other top assets remained unchanged, although brief periods of elevated intraday volatility were seen. 

Tezos, however, continues its track record of short-term outperformance with a +26% gain this week. 

The recent launch of Bakkt’s monthly Bitcoin options product marks an important step in the institutionalization of the asset class. Both CME and OKEx have announced they will offer Bitcoin options in the future, joining the ranks of Deribit who already offers option markets. A robust and liquid options market is significant because it allows portfolio managers additional tools to hedge away certain portfolio risks and opens the possibility of creating portfolios with a defined volatility target — an attractive proposition for pension funds and large institutional investors. 

Options also allow market participants to implement specific market views that previously were not possible, namely the ability to bet on future levels of volatility. In time, an index calculated off option-implied volatility levels (similar to the VIX for U.S. equities) are possible. Opportunities in volatility trading will likely be present as market participants gain experience in accurately pricing Bitcoin options.

Current levels of realized volatility, measured over a trailing one-month period, are reaching levels rarely seen over the past three years. Historically, annualized volatility has rarely dipped below 50% and exhibits mean-reverting behavior. For crypto assets, low levels of volatility breed complacency and increase risk-taking through increased leverage or futures positions. Under such conditions, heightened future volatility is likely, as forced liquidations can exaggerate a price move in either direction. Currently, Bitfinex leveraged long positions are approaching all-time highs although BitMEX’s open interest are still at moderate levels. 

CM Bletchley Indexes (CMBI) Insights

Most Bletchley Indexes were up slightly over the week, returning 1%-2% as the whole crypto asset market remained relatively flat. Mid-cap assets were the exception with the Bletchley 20 falling 0.8%.

A testament to the stability and general uniformity across the market over the last month is the results of the Bletchley December rebalance. The only change to the large-cap, mid-cap, and small cap indexes was the promotion of Tezos (XTZ) to the Bletchley 10 and the demotion of Binance (BNB) to the Bletchley 20.

Coin Metrics Updates

This week’s updates from the Coin Metrics team:

  • Coin Metrics is pleased to announce the release of CM Network Data Pro Daily Macro Version 4.4 and CM Network Data Pro Block-by-Block Version 2.1. These releases are centered around a large enhancement to our on-chain exchange-related metrics (including flows and transfers into/out of exchanges as well as the total supply of native units held on exchanges). Read more about the updates here.
  • Coin Metrics has also enhanced and improved the CM Real-Time Reference Rates methodology. Specifically, the methodology now incorporates inverse price variance weighting, in addition to volume weighting, which reduces the weighting of markets that contain extreme price outliers. Read more about the changes here.
  • Additionally, Coin Metrics has released version 2.0 of our CM Market Data Feed (MDF) product. CM MDF now includes Trades and Candles data for Futures instruments. Currently, BitMEX and Huobi futures for all available markets are supported. Read more about the release here
  • Coin Metrics is hiring! We recently opened up 6 new roles, including Blockchain Data Engineer and Data Quality and Operations Lead. Please check out our Careers page to view the openings.

As always, if you have any feedback or requests, don’t hesitate to reach out at [email protected]

Subscribe and Past Issues

Coin Metrics’ State of the Network, is an unbiased, weekly view of the crypto market informed by our own network (on-chain) and market data.

If you’d like to get State of the Network in your inbox, please subscribe here. You can see previous issues of State of the Network here.

Check out the Coin Metrics Blog for more in depth research and analysis.

Release of CM Market Data Feed Version 2.0

Today, Coin Metrics is pleased to announce the version 2.0 release of our CM Market Data Feed (MDF) product. 

CM Market Data Feed version 2.0 includes the below items.

Methodology Documentation

With the release of CM MDF version 2.0, we have posted a Methodology document to our website. Similar to the CM Reference Rate’s Methodology documents, the CM MDF Methodology document aims to provide detail on the  calculation process as well as product specific information. 

New Instruments

CM MDF now includes Trades and Candles data for Futures instruments. Currently, BitMEX and Huobi futures for all available markets are supported. Note that complete historical Trades and Candles for these exchanges is expected to be included in the new release. Subsequent releases of CM MDF will include Futures instruments from additional exchanges. In the future, Coin Metrics expects to add additional instrument types to CM MDF such as Options, Swaps, Fixed Income instruments, etc. 

New Exchanges

In addition to the Futures exchanges noted above, CM MDF now includes Trades and Candles data for all spot pairs traded on Binance.US, bringing the total list of integrated exchanges to 27. Coin Metrics expects to add more more exchange integrations in the near future.

New Order Books and Quotes Markets

The following markets have been enabled for Order Books and Quotes: cex.io-btc-usd-spot, bitflyer-btc-usd-spot.

Please reach out to Coin Metrics ([email protected]) for more information on CM Market Data Feed.

CM Network Data Pro Release: Updates and Enhancements to the Exchange Flows Metrics

Coin Metrics is pleased to announce the release of CM Network Data Pro Daily Macro Version 4.4 and CM Network Data Pro Block-by-Block Version 2.1. 

Flows Metrics Updates and Enhancements

These releases are centered around a large enhancement to our on-chain exchange-related metrics (including flows and transfers into/out of exchanges as well as the total supply of native units held on exchanges). Because these metrics requires sourcing information external to the chain (to tag addresses related to an exchange) and because this information needs to be monitored closely, in order to offer professional-grade on-chain exchange-related data, a rigorous daily human review process is necessary (in addition to automated processes). 

This update therefore included i) a deep historical review of this data, ii) the implementation of a new daily human quality control process (in addition to existing and revised automated processes), and iii) the implementation of a new API with status labels that identify when a particular observation has been manually reviewed and revised (as necessary). The end result is a far more complete picture of this information for exchanges both historically and moving forward. 

To our knowledge, this process will be the first of its kind in the industry and highlights Coin Metrics’ commitment to providing professional and innovative services. This provides clients greater confidence in the accuracy of these data and greater immediacy and transparency into the review of anomalous activity and any subsequent adjustments. 

Finally, please note that CM intends to release a more complete methodology and background document in the coming weeks.

Updates and Enhancement Benefit

Please see below for a demonstration of the benefit of the updated methodology versus the old methodology. You can see that the current, enhanced methodology captures more of the total BTC and ETH supply across exchanges than the original methodology. 

Other Updates and Notes

  • Binance Chain updated its transaction fees schedule, reducing fees by as much as one half. Binance Chain’s code is closed source and Coin Metrics cannot cannot audit the logic used to compute fees. Instead, Coin Metrics has hard coded the fee structure. The change in fees has not yet been reflected in Binance Chain’s fee-related metrics in our Network Data Community data and the current fee-related metrics are likely overestimates. This change will be reflected in a future release and we will update at that time.
  • Removal of Certain Pre-Mainnet Assets. The following assets will be removed from our Network Data Community API and toolings. These assets were supported during their initial launch as ERC-20 tokens but these projects have since migrated to their own mainnets. Since these mainnets are not currently supported, it makes little sense to continue to offer these historical (and now outdated) data. Please note: the historical data will not be deleted from our storage (it will be retained) but it will no longer be available in our Community products. Below is the list of removed assets:
    • Aeternity (ae)
    • Aion (aion)
    • Bytom (btm)
    • ICON (icx)
    • Loopring (lrc)
    • Nebulas (nas)
    • QTUM (qtum)
    • VeChain (vet)
    • Ziliqa (zil)

Please reach out to Coin Metrics ([email protected]) for more information on CM Network Data Pro.

Updates and Enhancements to the CM Reference Rates

Today, Coin Metrics is happy to announce a series of updates and enhancements to our CM Reference Rates products. 

CM Real-Time Reference Rates Version 0.1 (Beta 2)

The release of CM Real-Time Reference Rates Version 0.1 (Beta 2) includes the changes noted below. An updated methodology document is available here. Please note that the recalculation mentioned below has a cross impact on the CM Network Data Pro – Block-by-Block (Real-Time) Version 2.1 release which is detailed here.

Methodology Enhancements 

Coin Metrics has enhanced and improved the CM Real-Time Reference Rates methodology. Specifically, the methodology now incorporates inverse price variance weighting, in addition to volume weighting, which reduces the weighting of markets that contain extreme price outliers. 

Under normal circumstances, the variance of the price for each of the selected constituent markets will be very similar. So each market will get a similar inverse price variance weighting. However, when one constituent market contains a significant outlier, due to unusual circumstances such as an errant trade or error in the exchange matching engine, that market will have a very high variance in their price compared to other markets which are orderly. The inverse price variance of the constituent market with the outlier will be very small and that market will receive less weight.

Overall, this methodology update makes the CM Real-Time Reference Rates more stable and less susceptible to outliers, particularly for assets that are thinly traded. 

Coin Metrics has recalculated and published all historical real-time reference rates which are now available in the API. The CM Real-Time Reference Rates methodology will be left in “beta” while Coin Metrics continues to test-stress and monitor performance. Once satisfied that the methodology is sufficiently robust, we will promote the methodology to version 1.0.

Methodology Enhancements Examples

As a brief supplement to the CM Real-Time Reference Rates Research Summary published as part of the Beta release, please see below for an example of the updated methodology versus the old methodology. You can see that the current, enhanced methodology more intelligently manages the Poloniex outlier trade when compared to the original methodology. 

Current Methodology Version 0.1 (Beta2):

Previous Methodology Version 0 (Beta):

New Assets

As part of the CM Real-Time Reference Rates Version 0.1 (Beta 2) release, Coin Metrics has expanded the Coverage Universe to include two new assets: Beam (beam) and Algorand (algo).

CM Hourly Reference Rates Version 2.1

The release of CM Hourly Reference Rates Version 2.1 includes the changes noted below. An updated methodology document is available here. Please note that the recalculation mentioned below has a cross impact on the CM Network Data Pro – Daily Macro (EOD) Version 4.4 release which is detailed here.

New Assets

As part of the CM Hourly Reference Rates Version 2.1 release, Coin Metrics has expanded the Coverage Universe to include two new assets: Beam (beam) and Algorand (algo).

Please reach out to Coin Metrics ([email protected]) for more information on CM Reference Rates.

Coin Metrics’ State of the Network: Issue 28

Weekly Feature

The Psychology of Bitcoin Bubbles as Measured by Investor Cost Basis

by Kevin Lu and the Coin Metrics Team

Bubbles and subsequent crashes in financial assets occur with regular frequency in economic history, largely due to deeply-rooted cognitive biases in human psychology. In Bitcoin’s ten year history, at least three complete bubble-and-crash cycles have been observed. 

We examine the state of investor psychology at six critical moments in Bitcoin’s history (including its recent state) using the distribution of investor’s estimated cost basis, an extension of the realized capitalization concept. 

Understanding Estimated Cost Basis Distribution

Unlike market capitalization, which values each coin at the current price, realized capitalization values each coin at the time of its last on-chain movement. We further extend this concept by using one important assumption: an on-chain transfer represents a trade between a willing buyer and a willing seller, such that the price at the time of the transfer represents the cost basis of the buyer. Thus, realized capitalization can be thought of as the aggregate cost basis of all holders.

This assumption does not always hold in practice — many transfers do not represent a change in ownership and instead are motivated by wallet maintenance, wallet shuffling, and other reasons. The majority of trading also occurs on centralized exchanges where coins are transferred between parties within an exchange’s internal ledger and do not require on-chain transfers. For these reasons, the idea of realized capitalization as a representation of aggregate cost basis should be taken as an imperfect estimate of the true value. 

Despite these important caveats, an analysis of an estimated cost basis for each native unit of a crypto asset allows for rich analysis of investor behavior in a way that is not possible for other financial assets. For example, data regarding the cost basis for each individual share of a company’s stock is not available. 

Cost basis data can be valuable because it provides insight into the degree of euphoria or pain that investors are experiencing (due to having either unrealized gains or unrealized losses), two fundamental emotions that affect investment decision-making during asset bubble formations and crashes. 

Six Critical Moments in Bitcoin’s History

We examine the state of investor psychology through snapshots of estimated cost basis distribution during the previous cycle and during the current market cycle. For each market cycle, three points are chosen: at bubble peak, at lows for the cycle after the bubble has been completely deflated, and at a mid-cycle correction that involved a significant decline in prices after a recovery. All six moments are annotated on the chart below. 

A brief description of each moment is described in this table and more fully explored in the sections below. 

  1. Peak of the Previous Cycle on November 29, 2013 

We first examine the distribution of estimated cost basis distribution on November 29, 2913, the peak of the previous cycle where prices increased from roughly $100 to $1,100 in the span of two months. Below we introduce a visualization that contains a snapshot of the estimated cost basis for each native unit of Bitcoin at this time. Units of Bitcoin are assigned to bins where each bin is segmented into a $50 price interval. 

For instance, almost 5.5 million Bitcoin had a cost basis between $0 and $50 (although some of these Bitcoin are presumed to be lost) and almost 400,000 Bitcoin were bought at bubble peak with a cost basis between $1,100 and $1,150. At the time this snapshot was taken, prices were at a new all-time high, such that 100% of the Bitcoin held had unrealized gains. 

Despite the explosive price growth observed at this stage in Bitcoin’s development, holdings of Bitcoin remained heavily concentrated in the hands of early adopters with approximately 6.4 million of the 12 million Bitcoin in existence at the time having a cost basis between $0 and $100. As prices exceeded the previous all-time high of $200, all holders sat on unrealized gains and investor sentiment reached its most positive phase. Although there was likely interest in buying more at this phase from new adopters, existing holders find few reasons to sell, leading to constricted available supply and few net transfers of Bitcoin into the $200 to $700 price intervals. 

As prices rapidly exceed $800, however, existing holders begin to be incentivized to sell some of their holdings as they experience a several hundred percent return in less than two months. Compared to the $200 to $700 price intervals, many more Bitcoin was transferred from early adopters into the $800 to $1,150 price intervals as prices peaked at approximately $1,129 on November 29, 2013. This indicates that for Bitcoin, bubbles exhaust themselves not from a lack of buying interest but because the large cohort of early adopters are incentivized to bring their previously dormant holdings to market. 

  1. Lows of the Previous Cycle on January 14, 2015 

Following the peak of the previous cycle on November 29, 2013, Bitcoin experienced an 84% decline over the course of slightly more than a year. Here we show the distribution of estimated cost basis on January 14, 2015 when Bitcoin prices reached a low of $176. On this date, 43% of Bitcoin held had unrealized gains and 57% had unrealized losses. 

The shape of the distribution of estimated cost basis changed substantially over the course of slightly more than one year. In contrast to the distribution seen at bubble peak, there are few holdings above $800 and the holdings of early adopters are significantly reduced. 

Here we introduce another visualization that shows the change in the distribution between two points in time instead of a snapshot of the distribution at one point in time. For instance, almost 1.5 million Bitcoin that was originally in the $0 to $50 price interval was sold between November 29, 2013 and January 14, 2015.

The change in distribution illustrates a picture of complete capitulation, a state of investor psychology that is necessary for prices to completely bottom. Early in the price declines, both investors that bought at bubble peak (above $800) and large numbers of early adopters (below $150) reach the point of maximum pain and sell massive existing holdings to investors who buy the dip. 

Although not easily seen in this visualization, subsequent waves of dip buyers are also seen to sell as prices continue to decline. Significant transfers are observed — a total of 5.7 million Bitcoin out of the 13.7 million Bitcoin in existence moved from one price interval to another. Using this market cycle as a study of deeply-rooted investor psychology, prices do not bottom until capitulation is seen from investors that bought at the peak, large portions of early adopters (although not all), and from early dip buyers who represent investors with the most conviction but also reach a point of maximum pain as prices continue to decline. 

  1. Mid-Cycle Correction of the Previous Cycle on August 24, 2015 

Although usually not seen as a critical event in Bitcoin’s bubble-and-crash cycles, we examine a mid-cycle correction where prices declined significantly from recent highs. This market environment most closely resembles the current state of Bitcoin as prices have declined from around $13,000 during the summer to a low of $6,500 in late November of this year. During the previous cycle (November 2013 to January 2015), after bottoming at $176, prices recovered to around $310 over the next few months before correcting 35% to $200 over the course of two months. Here we show the snapshot of the distribution of estimated cost basis on August 24, 2015, immediately following the 35% correction. 

We also show the change in distribution between the lows of the cycle and immediately following the mid-cycle correction. Over the course of eight months, a significant change in the shape of the distribution is observed. We see some limited selling from early adopters with cost basis below $200, although the magnitude of selling is much less than was observed in the previous section, suggesting that capitulation of early adopters is near complete and even a 35% drawdown is not enough to cause this cohort of investors to bring more supply to market. 

Importantly, we see nearly zero selling pressure from investors who bought at bubble peak (above $800) indicating that capitulation for this cohort of investors is complete. Instead, we see most selling pressure coming from investors who bought the dip as prices declined from bubble peak and from recent investors who bought as prices increased off the bottom. This cohort of investors had not yet reached the point of maximum pain prior to the sharp decline in prices. 

A study of investor psychology suggests that prices cannot truly bottom until all investors have reached the point of maximum pain and capitulation is complete. An examination of the mid-cycle correction that occurred during the previous cycle indicates that selling pressure from investors who bought at the peak and most early adopters is complete. Significant capitulation was also observed from dip buyers and recent investors who had not yet reached the point of maximum pain. After this mid-cycle correction was complete, prices never declined to these levels again as the state of investor psychology had reached the point where most investors who wanted to sell had already sold. 

  1. Peak of the Current Cycle on December 17, 2017

Over the course of one year, prices passed the previous all-time high of around $1,100 and peaked at near $20,000 on December 16, 2017. Here we show again a snapshot of the distribution of estimated cost basis on this date but assign each native unit of Bitcoin to $500 bins. At this point in time, about 7.4 million out of the 16.75 million Bitcoin are held by early adopters with an estimated cost basis of between $0 and $1,000 and 98% of Bitcoin held have unrealized gains. 

The shape of the distribution at bubble peak of the current cycle looks similar to the distribution at bubble peak of the previous cycle with some important differences. A large amount of transfers were observed at prices between $1,000 and $7,000 whereas transfers at intermediate prices in the previous cycle were much lower. This caused the price appreciation to occur at a slower pace over a period of about 8 months, with many significant corrections along the way. 

On the other hand, we see very few transfers of Bitcoin into the $8,000 to $16,000 price intervals, not because of a lack of buying interest but because existing holders saw few reasons to sell. This coincided with an extremely rapid uninterrupted ascent in prices — prices increased by $8,000 in only two weeks. We see increased transfers as the prices exceeded $16,000 as the high prices again incentivized early adopters to bring more supply to market. Several hundred percent increases in price in a short amount of time seem to draw early adopters, who hold significant amounts of Bitcoin, to sell their long-held Bitcoins.

  1. Lows of the Current Cycle on December 15, 2018

The speed and magnitude of drawdown during the current market cycle are remarkably similar to the previous cycle — both experienced a drawdown of around 84% and required about one year for the bubble to completely deflate. At the lows of the current cycle, 39% of Bitcoin held had unrealized gains, also similar in magnitude to the 43% of Bitcoin that had unrealized gains during the previous cycle. This suggests that following a bubble, maximum pain and capitulation can only be reached when prices decline to a point where only roughly 40% of Bitcoin held have unrealized gains. 

Both a snapshot of the distribution of estimated cost basis and the change in the distribution also show strong similarities to the previous cycle, providing support for the assertion that bubbles-and-crashes in Bitcoin (and other financial assets) are driven by deeply-rooted cognitive biases which lead to repeating cycles. 

Here we show the change in the distribution between the peak of the bubble and the lows of the current cycle. Similar to the previous cycle, strong capitulation is observed by investors who bought near the bubble peak (with cost basis above $16,000) and early adopters (with cost basis below $1,000). Few holdings remain with cost basis above $12,000 so investors belonging to this cohort are not likely to be a source of selling pressure going forward. 

A total of 5.4 million Bitcoin (out of 17.4 million Bitcoin in existence) moved price intervals over the course of this year. Although not clearly seen in this visualization, investors who bought near the bubble peak and early adopters were the first group of investors to capitulate. During the final decline between $6,000 and $3,000, selling pressure was observed from investors who bought the drip as prices declined. They represent the final group to experience maximum pain and reach the point of capitulation. 

One important caveat when looking at the change in distribution at this time is that Coinbase in early December 2018 migrated its holdings to an alternative set of cold wallets. According to their statements, they migrated 5% of all Bitcoin in existence, or almost 900,000 Bitcoin, most of which likely had a low cost basis. 

  1. Mid-Cycle Correction of the Current Cycle on November 25, 2019

The current state of the distribution illustrates that holdings are becoming less concentrated over time and at an increasingly higher cost basis. Whereas at bubble peak, early adopters held 7.4 million Bitcoin with cost basis between $0 and $1,000, these holdings have been gradually reduced to 5.0 million Bitcoin today. Excluding early adopters, cost basis now appears to be roughly normally distributed around $8,000, with a noticeable spike at $3,500 which coincides with the lows of the current cycle. Consistent with the lows of this cycle, holdings with a cost basis above $12,000 are at a very low level and represent a cohort of investors that have completely capitulated. 

Below we show the change in the distribution between June 26, 2019 (where prices reached a local peak of $12,863) and November 25, 2019 (where prices recently bottomed at $7,139). The complete peak-to-trough drawdown during this time period has approached 50%, a very steep decline for a bull market, which has caused many market observers to question the current market regime. 

Analysis of the sources of selling pressure reveals that investor concern is warranted. Unlike the mid-cycle correction observed in the previous cycle, selling pressure is broad-based and originates from many cohorts of investors. Investors that bought at the local peak with cost basis of around $12,000 to $12,500 have been heavy sellers. Dip buyers have also seen signs of capitulation with heavy selling observed from Bitcoin with cost basis between $7,500 and $8,000. Recent investors that bought after prices bottomed with cost basis between $3,000 and $6,000 also have sold significant amounts. And early adopters have been a source of small but not negligible selling pressure. 

Compared to the previous changes in distribution, the current change in distribution is most similar to when prices were reaching a bottom of the cycle. This interpretation may be bullish if the types of sellers have fully reached capitulation and are likely not to be a source of selling pressure going forward. Analysis of the current distribution supports this theory. However, seeing this broad-based selling pressure may reveal that capitulation, originally thought to be complete when prices bottomed in December 2018, may actually require more time or further price declines. 

Network Data Insights

Summary Metrics

BTC and ETH mining revenue are down significantly for the second straight week, mostly due to the large price decreases. Similarly, ETH fees came back down to earth, dropping 14.4% after growing by over 20% the week before (likely due to the launch of the Gods Unchained marketplace). 

After dipping to a six month low of 1.23 last week, BTC market value to realized value (MVRV) ratio, calculated by dividing market cap by realized cap, started to increase again over the past week. As of Sunday, December 1st, BTC MVRV was 1.32.

XRP transactions jumped by over 50% this past week, after being up over 121% the week before. Ripple recently completed a $50M investment in MoneyGram. However, it is unclear whether this directly led to the increased activity. Despite the increased usage, XRP’s market cap dropped by 6.8% over the past week, which is a larger decrease than both BTC and ETH.

Network Highlights

After reaching all-time highs in May, the amount of Bitcoin that has not moved in over one year has since been declining. As of Nov. 31st, 3,174,760 Bitcoin had not been moved in at least one year. Comparatively, there was 4,500,526 Bitcoin that had not been moved for at least a year on May 18th, 2019. 

The below chart shows the amount of Bitcoin not moved in over X years, where X ranges from one month to 5 years.

Decred daily active addresses are approaching new all-time highs. On November 16th, Decred had 25,315 active addresses, which is its highest daily total since April 23rd, 2016. The below chart shows Decred active addresses smoothed using a seven day rolling average.

Note: November’s active addresses appear to be all-time highs in the below chart because the all-time highs in April 2016 were outliers, surrounded by days with low relatively low active addresses.

Zcash, on the other hand, is trending in the opposite direction. Zcash active addresses are approaching all-time lows. As of December 1st, Zcash had 11,218 daily active addresses, which is the lowest since October, 2016. The below chart shows Zcash active addresses smoothed using a seven day rolling average.

Market Data Insights

As November concludes, we examine indexed prices over the past month. Bitcoin is down 20% for the month, although slightly higher than the lows that occurred on November 25. Most other major assets are down a similar magnitude. Over the past week, Monero, EOS, and Cardano have staged decent recoveries although all assets are down for the month. 

Among smaller assets, Tezos leads with a +44% return for the month of November, driven in part by the announcement that Coinbase would be offering staking rewards. Cosmos also had a positive return of +18%, but all other assets are down. Curiously, although almost all assets displayed high correlation to Bitcoin, UNUS SED LEO has exhibited unusually low correlation for an unknown reason. 

CM Bletchley Indexes (CMBI) Insights

This week crypto assets experienced a slight market wide recovery, with all Bletchley Indexes returning 6-9%, after the ~20% market wide drop in the prior week. Coincidentally the even indexes outperformed the market cap weighted indexes this week, indicating that the best weekly performers within each index were the lower market cap constituents.

Despite strong weekly performance of all indexes, it was not enough to overcome a bad November against the USD, with all indexes experiencing significant losses. The Bletchley 20 and Bletchley 40 performed best over the month, indicating that mid and small-cap crypto assets outperformed large-cap crypto assets for the second month in a row.

Large-cap assets seem to have largely moved in tandem with Bitcoin over November, evidenced by the negligible returns of the Bletchley 10 in BTC terms, whereas the Bletchley 20 and Bletchley 40 both had positive returns, 10% and 7.5% respectively, against a BTC pairing.

Coin Metrics Updates

This week’s updates from the Coin Metrics team:

  • Coin Metrics is hiring! We recently opened up 6 new roles, including Blockchain Data Engineer and Data Quality and Operations Lead. Please check out our Careers page to view the openings.

As always, if you have any feedback or requests, don’t hesitate to reach out at inf[email protected]

Subscribe and Past Issues

Coin Metrics’ State of the Network, is an unbiased, weekly view of the crypto market informed by our own network (on-chain) and market data.

If you’d like to get State of the Network in your inbox, please subscribe here. You can see previous issues of State of the Network here.

Check out the Coin Metrics Blog for more in depth research and analysis.