In this special edition, we look back at the best issues of the past year to celebrate State of the Network’s one-year anniversary.
Over the last 52 issues we have taken a data-driven approach to elucidating the best (and worst) of crypto, powered by Coin Metrics’ network and market data. In this issue we look back at three specific themes that we have covered over multiple issues over the past year:
Transparency and auditability
Valuation and market analysis
Network security and health
We would also like to use this opportunity to wish a big thank you to everyone who has viewed and subscribed to State of the Network over the past year. Please let us know if you have any feedback about State of the Network, especially if you have ideas about how we can keep making it even better. You can submit feedback here.
Transparency and Auditability
At Coin Metrics, transparency and auditability are the foundational building blocks that our data and analysis are built on top of. This starts with running our own nodes for the cryptoassets we cover.
In Issue 30, we ranked full nodes in several tiers (A, B, C, and F) depending on their ease of synchronization, updatability, and maintenance. Below 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).
Running our own nodes also allows us to audit supply, and trace which addresses supply flows through. In Issue 32 we investigated the Stellar inflation experiment and found that about 98% of Stellar inflation payments accrued to the Stellar Development Foundation (SDF).
In Issue 26 we did a deep investigation into the nuances of Bitcoin’s supply to determine how many Bitcoins are permanently lost. From those estimates of lost coins, we constructed adjusted views of Bitcoin’s supply.
This supply investigation eventually led us to the introduction of CM Free Float Supply in Issue 48, a new metric that is being developed to more accurately represent the supply of an asset available to the market.
Accurate supply measurements have also played a foundational role in our research into cryptoasset usage patterns. In Issue 38 we analyzed the supply distribution of eight cryptoassets. We found that Bitcoin and Ethereum were gradually getting more distributed, while BCH and BSV were not.
Over the year, we also followed the rapid rise of both the supply and market capitalization of Tether and other stablecoins. Total stablecoin market cap grew from about $4.35B at the beginning of June 2019 to over $10B as of May 2020. We covered the rise of stablecoins in Issue 15, Issue 17, Issue 25, Issue 38, Issue 47, and more.
Over the past year we also devoted a lot of time to valuation research and market analysis.
In Issue 37, we released the first part of our “Cryptoasset Valuation Research Primer.” In Part 1 we explored six main categories of crypto valuation research: equation of exchange, discounted future utilities model, Metcalfe’s law, price regression models, cost of production models, and asset bubble identification.
In Part 2 of our cryptoasset valuation primer in Issue 40, we surveyed five additional facets of the literature: fundamental ratios, UTXO age analysis, realized capitalization-based analysis, factor investing, and social media-based analysis. We followed this up with a deep-dive into a specific fundamental ratio, Market Value to Realized Value, in Issue 41.
We also analyzed extreme market movements, both positive and negative. In Issue 19, after Bitcoin experienced its largest single-day decline in price in 2019 to date, we used on-chain data to show that selling pressure originated from traders that acquired Bitcoin at prices between $10,000 and $12,000.
In Issue 23, after Bitcoin bounced back and experienced one of the largest 24-hour returns in its history (at the time), we analyzed our trades data, which is part of our Market Data Feed offering, and found that spreads between major exchanges remained small, indicating that market participants have the ability to quickly transport liquidity across markets.
In Issue 42, after the March 12th 2020 crash, we again looked at the spread between exchanges and observed a large spread of around 16% between BitMEX and Coinbase during the sell-off. This was likely due to the BitMEX liquidation spiral after BitMEX went down for unscheduled maintenance in the early hours of March 13th, which we covered in detail in Issue 43.
We also used our realtime reference rates to track how Bitcoin reacted to geopolitical and macroeconomic events in real-time. In Issue 33, we analyzed Bitcoin’s response to growing military tensions in Iran.
And in Issue 12, published on August 13th, 2019, we began to analyze the theory that Bitcoin acts as a unique safe haven asset. We revisited the safe haven theory in Issue 33, Issue 39, Issue 45.
In Issue 46, we conducted a correlation analysis using our reference rates and intraday equity data, and found that while there are pieces of evidence that the correlation between Bitcoin and gold may be growing, Bitcoin’s overall correlation with gold is still relatively weak.
Security and mining analysis has been another theme for State of the Network over the last year. To assess the long-term health and security of a crypto network, it’s critical to track different security metrics like hash rate and miner rewards. It’s also crucial to keep track of hashpower distribution, both among different entities and among different types of mining hardware.
In Issue 23, we started exploring a novel technique of analyzing nonce distributions to find evidence of specialized types of mining hardware called ASICs.
We continued this line of research in Issue 45, where we extended the nonce analysis to identify different mining pools and estimate the usage of specific types of hardware.
We revisited the nonce distribution research once again in Issue 51, after Bitcoin’s third halving on May 11th, 2020. Using our techniques for estimating the types of mining hardware being used, we estimated that a significant number of formerly-offline S9s were turned back on prior to the halving. While the amount of hashpower that could be created by offline S9s is nowhere near enough to 51% attack the network, the change to the network’s security dynamics caused by their presence is significant.
We also analyzed prior halvings by looking back at historical data. In Issue 44, using a set of axioms, we provided a framework reasoned from first principles that illustrates how miners are a continuous and significant source of selling pressure that has a pro-cyclical impact on prices.
Our research on security also led us to designing improved ways to measure it. In Issue 49, we introduced the CMBI Bitcoin Index and ‘Observed Work’ as a more reactive, responsive and manipulation resistant way to measure the realities of mining activity when compared to traditional hash rate estimations.
Additionally, Observed Work and Coin Metrics’ CMBI Bitcoin Hash Rate Index can potentially serve as the foundational pieces of financial products that can provide markets with the required tools to effectively and efficiently trade and/or hedge Bitcoin’s hash rate.
It has been a great year of data-driven analysis focused on transparency and auditability, valuation and market analysis, security analysis, and much more. Thanks again for your readership, and please leave us feedback if you have any ideas or comments. We look forward to continuing to bring you the best data-driven crypto stories for years to come.
Coin Metrics Updates
This week’s updates from the Coin Metrics team:
This has been a special edition of State of the Network. The regular format including Network Data and Market Data section will return next week.
Coin Metrics is hiring! 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.
Cryptoasset data has been increasingly amassing the interest of
researchers and investors looking to build an edge in the market and inform
their trading strategies. At Coin Metrics, we believe that the greatest
insights can be generated when data is considered holistically. To that end, we
have developed a comprehensive suite of data that includes network data, market
data and indexes.
The following is a case study that highlights how any single
data strategy might overlook some valuable insights and opportunities that can
present themselves in a holistic data strategy.
Near midnight UTC on May 10, 2020, Bitcoin saw a rapid selloff
that brought the price down to $8,000 from $9,500, an almost 16% decrease in
less than 15 minutes. This move was preceded by a network data alert in
the Coin Metrics monitoring system, which allowed for the following market
action to be closely observed in real-time.
Block rewards are currently the primary source of revenue for miners. A reduction in this reward due to the halving causes some miners to exit the network. In the short term the sudden drop-off of miners can potentially leave the network more exposed to security threats like 51% attacks.
The distribution of hashpower among different types of mining hardware also has an impact on the network’s security properties. The availability of old hardware on secondary markets poses a potential threat to the network.
Novel techniques involving nonce distributions allow us to numerically estimate the amount of hashpower provided by certain types of hardware, including older hardware like S7s and S9s.
It now appears that a significant number of formerly-offline S9s have been turned back on. Currently, the Antminer S9 family of miners is responsible for about 32% of Bitcoin’s hashpower.
While the amount of hashpower that could be created by offline S9s is nowhere near enough to 51% attack the network, the change to the network’s security dynamics caused by their presence is significant.
Halving a Good Time
On May 11th, for the third time in Bitcoin’s history, the amount of new coins issued per block was cut in half. This event, known as the halving, occurs every 210,000 blocks, or approximately four years, until issuance is eventually rounded down to zero.
In the most recent halving, the block reward was reduced from 12.5 to 6.25 BTC. The period leading up to the halving was marked by pronounced market volatility, which has somewhat subsided since the reduction occurred. The impacts of the event on the network’s security are nuanced.
Block rewards are currently the primary source of revenue for miners, so a reduction in this reward causes some miners to exit the network. With less revenue to go around, margins are tightened and less efficient miners may suddenly find themselves operating at a loss. In the long run, these miners are typically replaced by more efficient operations as the market rebalances. In the short term, however, the sudden drop-off of miners can leave the network more exposed to security threats like 51% attacks.
State of the Network Issue 44 reasoned about the impacts of the halving on miner economics from first principles. This piece will address a similar topic, focusing on the implications of the halving on security and the economics of running old mining hardware. In the process, we’ll use nonce data to estimate the prevalence of certain types of hardware on the network today, and discuss how the presence of old hardware impacts Bitcoin’s security model.
To Halve and to Secure
Bitcoin miners are compensated through both block rewards, which are directly affected by the halving, and transaction fees, which are not. Transaction fees are generally a function of demand for block space, and therefore tend to spike during periods of congestion and high traffic.
Currently, fees make up a small percentage of total miner revenue. Over the past five years, only about 4.4% of miner revenue has been generated from fees.
As Bitcoin’s block reward continues to halve approximately every four years, transaction fees will need to increase in order to sufficiently incentivize miners to secure the chain. Since slightly before the halving, fees have surged to make up about 17% of total miner revenue. While this effect has been intensified by the reduction in the block reward, transaction fees themselves have also increased to levels not seen in almost a year.
This increase in fees may have been amplified by the reduction in hash rate that has taken place since the halving. This reduction, in turn, is caused by less efficient miners leaving the network. The drop in hashpower has increased the time between blocks, therefore reducing the amount of available block space.
To reduce the variance of their payouts, miners often aggregate into mining pools, which are loose coalitions of miners that are organized by an operator and share revenue, typically according to hashpower contribution. Individual miners often switch between pools depending on several factors, most notably fees charged by operators.
As long as no single dishonest entity controls more than half of Bitcoin’s hashpower, the network is secure. In a process known as a 51% attack, an adversary who controls more than half the network’s hashpower can censor transactions and perform double-spends.
The minimum number of pools who would need to collude in order to 51% attack the network is known as the Nakamoto coefficient. At this time, the top 4 pools would need to collude in order to 51% attack the network. This number has generally gone up over the course of Bitcoin’s history, indicating a steady increase in decentralization.
The Nakamoto coefficient is not a perfect metric, and makes Bitcoin seem significantly more centralized than it is. Individual miners, who face large up-front expenses on capital like hardware, are disincentivized from attacking the network. These miners would likely defect from a maliciously-operated pool.
Still, pools select the blocks that their constituents mine, and barring defection exercise a certain degree of control over them. It also may be possible for an attacker to censor transactions with less than half of the network’s hashpower through techniques like feather-forking. It’s therefore useful to have a pessimistic metric like the Nakamoto coefficient to quantify the degree of centralization among miners.
Stratum V2, an implementation of Betterhash with modifications to the original protocol, suggests letting individual miners select the blocks that they will mine, rather than the pool operators doing so. This potential improvement to the way pools are operated would put more power in the hands of individual miners, further decentralizing the network.
In addition to the distribution of hashpower among different entities, the distribution of hashpower among different types of mining hardware has a significant impact on the network’s security properties.
To add a block to the blockchain, Bitcoin miners attempt to find a nonce, or arbitrary value, that causes the block header to hash to below a certain target. The rate at which these hashes are computed and verified is known as hashpower, and a nonce satisfying this condition is called a golden nonce. Golden nonces are theoretically uniformly distributed throughout the space of potential nonces and valid blocks. The threshold that the hash of the block header must satisfy is set by the network’s difficulty parameter, which is periodically adjusted according to the rate at which blocks have been accepted to the chain.
While mining was initially performed with CPUs, the process was parallelized and made more efficient early-on through the adoption of GPUs. Today, almost all mining is performed using mining rigs that contain specialized chips known as ASICS. These devices are significantly faster, better at parallelization, and more energy-efficient than other hardware.
Purchasing these devices requires a large up-front capital expenditure. This benefits the security of the network by requiring miners to lock up capital in an illiquid asset and therefore disincentivizing them from acting maliciously.
The presence of old mining hardware changes this security model, since it tends to require smaller up-front investment at the expense of higher operating costs. While there are practical and logistical barriers to starting a mining farm aside from the cost of hardware, the presence of old hardware allows entry into the market with significantly reduced capital expenditure.
Due to secrecy in the mining industry, it’s generally difficult to discern which types of mining hardware are being used to secure the network. However, novel techniques allow us to numerically estimate the amount of hashpower provided by certain types of hardware.
Bitcoin’s nonce distribution offers hints at the types of hardware being used to mine on the network. By combining this data with information on the prices of hardware on secondary markets, we can quantify the degree of risk posed by the existence of inexpensive, slightly dated hardware. For a detailed explanation of our analysis of nonce distributions, see our series “The Signal and the Nonce” Part 1 and Part 2.
Since golden nonces are uniformly distributed throughout the nonce spaces of all potential blocks, we’d expect a plot of the winning nonces over time to look like random static. Bitcoin’s nonce distribution doesn’t.
Near the left-hand side of the plot, nonces are concentrated in the lower ranges of the distribution. This is a result of a sampling technique used by miners in the CPU-mining era, which involved iteratively testing values starting from zero and incrementing upward.
Bitcoin’s nonce distribution also has a characteristic streaked pattern that first appeared in late 2015, and has recently begun to fade. The striations in question start out broad, and then suddenly narrow out, before gradually fading away. There are four distinct streaks, each of which can be specified in terms of its narrow and wide bands.
These streaks were noted in State of the Network Issue 23, and their source was identified in State of the Network Issue 45. The striations were found to come from the way in which nonces are sampled by the Bitmain Antminer S7 and S9 mining rig lines. Each of these rigs was at one point the dominant miner on the network, with the S9 having recently been supplanted in this role by the Antminer S17.
The wide and narrow bands are attributable to the sampling techniques used by the S7 and S9 families, respectively. We can use this knowledge to numerically estimate the proportion of the network’s hashpower provided by S7s and S9s.
According to these numerical estimates, the proportion of hash rate provided by S7s and related hardware peaked in May of 2016 at about 61%. Today, S7s are not responsible for a significant portion of hashpower. The proportion of blocks mined by S9s and related hardware peaked in May of 2018 at about 78%. Today, about 32% of blocks are produced by S9s.
These estimates are based on the assumption that S7s and S9s do not sample within their respective excluded bands and that all miners sample uniformly outside of any excluded regions. These conditions are violated in the CPU-mining era, but appear to hold from the GPU-mining period onward. The excluded regions are determined manually, and estimates are corrected for any extrapolated values outside of the 0-100% range and normalized.
The estimates are subject to a certain amount of noise, which is visible toward the left of the graph. The small bump in the estimated proportion of S9 hashpower in 2015 could be due to noise, or may be a sign of something else such as the testing of experimental hardware.
These figures are consistent with other estimates of the hashpower output of these types of hardware. A CoinShares report on Bitcoin mining released in December 2019 estimated that S9s made up about two thirds of the hardware in their equivalence class. In March, the founder of Beijing-based Spark Capital estimated that S9s provided 20 to 25 percent of Bitcoin’s hashpower.
The proportion of hashpower provided by each type of mining rig provides further perspective on the threat posed by old hardware.
The most salient insight from this plot is the exponential growth in the hashpower securing the network.
The estimated amount of hashpower provided by S9s reached its peak in August of 2019, when they generated about 52 exahashes per second. In February of 2020, the estimated hashpower generated by S9s reached the bottom of a valley at about 21 exahashes per second.
It now appears that a significant number of formerly-offline S9s have been turned back on, likely as a result of a recent appreciation in the price of Bitcoin. This hardware now computes about 37 exahashes per second.
Due to rapidly changing market conditions, this effect may not be sustained. However, it illustrates the degree to which mining with old hardware may be viable given favorable conditions, and the ease with which this less-expensive hardware can be deployed.
S9s are being sold on secondary markets for a fraction of their retail price. The miners can be purchased for between $20 and $80, compared to an original price of about $3000. Given today’s economic climate and the inexpensive electricity brought on by China’s rainy season, miners have found it possible to operate these devices profitably.
While the amount of hashpower that could be created by offline S9s is nowhere near enough to 51% attack Bitcoin, the change to the network’s security dynamics caused by their presence is significant. This effect may be felt more acutely by other platforms that use Bitcoin’s proof of work algorithm, including Bitcoin Cash and Bitcoin SV, which are currently secured by about 2.5 and 1.8 exahashes per second of computational power, respectively.
The Epitaph of the Third Epoch
In anticipation of the halving and on optimism related to increased institutional interest, the price of Bitcoin increased dramatically before giving up some of its gains. Since the halving, volatility has subsided somewhat, but price has continued to trend upward. The halving has also accelerated an increase in transaction fees and precipitated a slight drop in hashpower.
Halving-related sentiment will continue to impact the market, and the halving itself will continue the test of whether Bitcoin can successfully transition to a model where miners’ revenue is predominantly based on fees. The long-term effects of this event remain to be seen, but its impact on the economics of mining and the market as a whole are already pronounced.
Bitcoin (BTC) and Ethereum (ETH) transaction fees continue to climb, despite a relatively flat week for most other metrics.
Bitcoin’s median transaction fee reached $2.88 on May 14th, its highest level since June 2019. Similarly, Ethereum’s median transaction fee reached $0.25 on May 14th, its highest level since August 2018. Median transaction fees tend to surge when blocks are relatively full. The causes for this surge are explored in today’s Network Highlights section.
Bitcoin’s hash rate has dropped to 81.66 TH/s following the halving, about a 40% drop from pre-halving highs. As noted in this week’s Weekly Feature, this hash rate drop-off is to be expected as less efficient miners exit the network. It will likely recover after a period of churn where efficient miners replace less efficient operations. However, it is unclear exactly how long this turnover period will last.
As a result of the hash rate drop, the average interval between Bitcoin blocks has risen to its highest levels since late 2018 (excluding the period around March 12th 2020, where block interval shot up due to the sudden drop in Bitcoin price and subsequent hash rate drop-off).
Since there are less blocks being produced there is more competition for block space, which has led to the increase in transaction fees. Paying a higher transaction fee leads to a higher chance that miners will include the transaction in a block. So median fees tend to surge during periods where block space is at a premium.
The decrease in the overall number of blocks has also led to an increase in the size of each individual block. Bitcoin mean block size reached a new all-time high of 1.32 MB on May 17th.
Ethereum median transaction fees have also shown signs of growth since the Bitcoin halving. However, Bitcoin Cash (BCH), Bitcoin SV (BSV), Ripple (XRP), and Litecoin (LTC) median fees have not shown any significant increases.
This marks the third consecutive week that Bitcoin has outperformed other cryptoassets and forms a trend that cannot be ignored. Although it cannot be ruled out that this trend is simply a byproduct of the random walk of prices, one plausible explanation is that Bitcoin’s store of value properties are increasingly needed in today’s market environment. And as we observe the emergence of adoption by institutional investors, Bitcoin is the logical first choice as the gateway asset that may lead to the eventual adoption of cryptoassets as a distinct asset class.
There is an emerging narrative that Bitcoin is needed in a market environment of unparalleled monetary and fiscal policy by global central banks and governments. To examine this phenomenon, we show the year-over-year change in the Fed’s balance sheet highlighting the speed and magnitude of the Fed’s reaction to COVID-19. The Fed’s policy response has already exceeded the balance sheet expansions seen in the three previous quantitative easing programs following the financial crisis.
Bitcoin’s strong returns lately and the renewed interest in Bitcoin as a store of value in a rising inflation environment is remarkable because all indicators are still showing that inflation is not a problem, despite the strong growth in money and credit.
The shutdown of large swathes of the economy represents a demand shock which is deflationary by nature — energy prices reaching unprecedented lows are a prime example. Although monetary and fiscal policies are effective at getting money into the hands of U.S. businesses and households, the velocity of money has simultaneously declined. The most recent print for the U.S’s core inflation, which excludes food and energy items, fell 0.4% over the previous month, the largest monthly decline in the history of the series, according to the Bureau of Labor Statistics.
Similarly, inflation expectations either from survey-based indicators, such as the University of Michigan’s Survey of Consumers, or market-based indicators, such as 5-year, 5-year forward inflation expectations derived from TIPS, are well-anchored.
How can we reconcile the fear that Bitcoin will be needed in a rising inflation environment with the data that shows that realized inflation in the short-term is non-existent and inflation expectations over the medium to long-term are low? According to the Fed’s Survey of Consumer Expectations, median expectations for inflation have not meaningfully changed in response to the pandemic, but the level of uncertainty and disagreement across respondents have seen unprecedented increases.
One framework to bring clarity to this question is to view Bitcoin as a call option on inflation and to examine its greeks: the sensitivities of the price of the option based on the parameters of the underlying. What we are seeing now is an increase in the implied volatility of future inflation even if median expectations for the future level of inflation remain unchanged.
Standard option price theory indicates that increases in implied volatility of the underlying should lead to an increase in the price of a call option. Therefore, we can potentially attribute the recent increase in price of Bitcoin to the increase in implied volatility of inflation rather than the increase in the expected level of inflation.
CM Bletchley Indexes (CMBI) Insights
All CMBI and Bletchley Indexes had very good weeks, ending between 5% and 15% higher than the previous week’s close. Following the biggest news of the week, the Bitcoin Halving, the CMBI Bitcoin Index was the strongest performer, returning 14.4%. The CMBI Ethereum Index also had a strong week, closing 11.9% higher. Despite these two strong performances though, it was the Bletchley 40, small-caps, that was the best of the market cap weighted indexes, closing the week 12.1% up.
Stablecoin market capitalization has almost doubled since the Black Thursday crash. But the exact cause of the surge is still unknown.
To help elucidate how different stablecoins are being used, we analyzed the times of day that stablecoins are being transferred and created heatmaps to show daily usage patterns.
USDT-ETH has a clear pattern of heavy usage from about 2:00 to 16:00 UTC which corresponds with the hours that Asian and European stock markets are open.
Prior to March 12th PAX transfers appear to also mostly be clustered between 2:00 and 16:00 UTC (although not as dense as USDT-ETH). But as of April, PAX transfers have gotten noticeably more dispersed throughout the day.
Prior to March 12th USDC was slightly denser during U.S. hours. But after March 12th there has been a large uptick in usage between 1:00 and 8:00 UTC, which corresponds with Asian market hours.
DAI transfers have mostly been concentrated during U.S. working hours (14:00 – 22:00).
Stablecoin market capitalization has almost doubled since the Black Thursday crash, despite a drop-off for most non-stablecoin assets. As of May 7th, the aggregate stablecoin market cap has grown to over $10B.
There have been many theories about what has been causing this dramatic increase.
Some have speculated that the growth is caused by an increase in the amount of investors holding stablecoins as “dry powder” in anticipation of a new bull run. Others have proposed that it’s a reaction to a shortage of U.S. dollars, or a general rush to safety. Another theory is that Asian OTC traders are pouring money into stablecoins as an onramp to crypto markets.
While the exact cause of the market cap increase is still unknown, on-chain data can help point us in the right direction. In this week’s State of the Network we dive into on-chain transfer data to analyze the usage patterns of different stablecoins in order to shed light on some of the different theories about why market cap is growing.
Stablecoin Transfer Heatmaps
To help elucidate stablecoin usage patterns, we broke down stablecoin transfers by time of day.
The following heatmaps show the amount of stablecoin transfers by hour of day for different Ethereum-based stablecoins. The x-axis is the date, starting from the beginning of February. The y-axis is the hour of day ranging from 0 – 23 (UTC time zone). And the coloring represents the amount of transfers that occurred during that one hour block. So for example, the cross-section of March 1st on the x-axis and 0 on the y-axis represents the amount of transfers that occurred from 12:00 – 1:00 AM on March 1st.
We start by investigating the transfer patterns of Tether issued on Ethereum (USDT-ETH).
USDT-ETH has a clear pattern of heavy usage from about 2:00 to 16:00 UTC which corresponds with the hours that Asian and European stock markets are open. Transfers go dark towards the end of the day – there are very few transfers after 20:00, which is when the New York Stock Exchange closes.
The amount of transfers has also grown significantly since the middle of March. There are clusters of red (i.e. high transfer counts) towards the end of April, which appears to correspond with the hours that Asian markets are open.
While the above heatmap shows the total number of transfers per hour, the following heatmap shows the percentage of the total daily transfers that happened within that hour.
For example, if there were 100,000 daily USDT_ETH transfers and 6,000 happened between 8:00 and 9:00 UTC that hour would account for 6% of the daily total and would be colored yellow/orange. This gives a slightly clearer picture of daily usage patterns, regardless of the total number of transfers.
The following heatmap also shows that USDT-ETH is mostly transferred during Asian and European hours, with a flurry of activity towards the close of Asian markets. This bolsters the theory that USDT-ETH is being used by Asian traders.
Paxos (PAX) usage has also increased dramatically since March 12th. In fact, PAX daily transfers have tripled since March 12th, and reached a new all-time high of 24.4K on May 5th.
As a result, PAX has passed both USDC and DAI in terms of daily transfer count.
Prior to March 12th PAX transfers appear to also mostly be clustered between 2:00 and 16:00 UTC (although not as dense as USDT-ETH).
But as of April, PAX transfers have gotten noticeably more dispersed throughout the day. The on-chain transfer data potentially shows that PAX is increasingly getting non-institutional, global usage.
USD Coin (USDC) had a huge amount of activity on March 12th and the following week, but has dropped off since then. Notably, MakerDAO added USDC as a collateral option on March 17th, which likely contributed to the flurry of transfers.
Prior to March 12th USDC activity was slightly denser during U.S. hours. But after March 12th there has been a huge uptick in usage between 1:00 and 8:00 UTC, which corresponds with Asian market hours.
Interestingly, USDC had a string of days in April where close to 20% of daily transfers occurred in a single hour. The other stablecoins in our study did not have over 12% of daily transfers in a single hour.
Similar to USDC, DAI had a large increase in transfers on March 12th and the days immediately following, but has dropped off since then.
DAI transfers have mostly been concentrated during U.S. working hours (14:00 – 22:00). However DAI transfers are relatively spread out, and not nearly as concentrated as USDT-ETH.
Stablecoin transfer patterns show that different stablecoins are potentially being used for different purposes, and are favored in different parts of the world.
USDT-ETH transfers are concentrated during Asian and European market hours. USDC transfers are also clustered during Asian market hours, but not as densely packed as USDT-ETH. PAX transfers are more dispersed, which could signal that it is being used for non-institutional purposes. And DAI transfers mostly occur during U.S. hours.
Stablecoins are a crucial part of the crypto ecosystem, and will only keep growing in prominence. We will continue to keep our eye on stablecoins, and track their usage and growth as they continues to develop.
Both Bitcoin (BTC) and Ethereum (ETH) had another mostly strong week in the lead up to BTC’s halving. BTC’s market cap briefly topped $180B as price approached $10,000. The week ended on a volatile note, however, with BTC market cap dropping back down to about $160B.
BTC and ETH transaction fees were both up over 30% again this week, after ETH fees grew by 31% last week and BTC fees grew by a massive 170%. Fee growth is a positive sign that demand for block space is growing, which generally signals that network usage is increasing.
BTC hash rate grew to all-time highs in the lead up to the halving. As the halving approached, miners rushed to get the last of the 12.5 BTC reward blocks, causing hash rate to increase and average block time to decrease.
The following chart shows BTC hash rate smoothed using a seven day rolling average.
The amount of BTC held by BitMEX and Bitfinex has reached new lows following the March 12th crash. Bitfinex now holds 93.8K BTC, down from 193.9k on March 13th. BitMEX’s BTC supply is now down to 216.0K BTC, down from a peak of 315.7K on March 13th.
The most noteworthy market news this week was the endorsement of Bitcoin by hedge fund manager Paul Tudor Jones. In a note shared to clients, he cites its attractive store of value characteristics in the face of unparalleled monetary stimulus that he has termed the “Great Monetary Inflation”.
Markets had a rational response to the news with Bitcoin outperforming most other major cryptoassets. This marks the second consecutive week of notable outperformance and deserves continued observation.
For nearly a year, correlation between Bitcoin and other assets has remained high and dispersions in returns has remained small, but history has shown us that these periods of calm are interspersed with periods of large shifts within crypto markets. In our previous State of the Networks, we have commented on these regime shifts and the difficulty it poses for fund managers.
The somewhat moderate reaction and the lack of large moves in traditionally high-beta altcoins suggests that we are still far from the irrational investor sentiment that characterizes the late stage of market bubbles.
Paul Tudor Jones’s prediction of unchecked monetary stimulus leading to an increase in inflation is still, surprisingly, a contrarian view that is not yet priced in according to five-year, five-year forward inflation expectations. This measure of inflation is widely cited as a proxy for long-term inflation expectations that is less sensitive to the demand shocks of today or the volatility of food and energy prices.
So far, despite everything that has happened, inflation expectations are stable and the Fed still has its credibility intact. In short, the market believes that the Fed will do what is necessary to defend its price stability mandate. The issue is that we could face a situation where the Fed’s dual mandate of maximum employment and price stability becomes untenable and it will have to favor one over the other. These fears manifested itself in higher inflation expectations in the years following the 2008 financial crisis leading to a multi-year period where assets that rose from rising inflation expectations benefited. Ultimately, the inflation alarmists were proven incorrect as the Fed was able to thread the needle and provide enough stimulus to heal the economy without stoking undue inflation.
The market is betting that it will do the same this time. But it is clear that if the market is incorrect, we could see much higher prices for Bitcoin in the future.
CM Bletchley Indexes (CMBI) Insights
All CMBI and Bletchley Indexes gave up most of last week’s returns, except for the Bletchley 40 and Bletchley 40 Even which were the only indexes that finished the week positive, returning 4.6% and 5.1% respectively. Whilst the CMBI Bitcoin Index was down 3.9%, it was the CMBI Ethereum Index that retraced the most during the week, falling 11.1%.
Even weight indexes are a mechanism of gaining greater exposure to the smaller-cap assets and as such, can often provide different return profiles to market cap weighted indexes. This week’s return profile demonstrates just that, with even weighted indexes, excluding the Bletchley 10, outperforming their market cap weighted counterparts for the week. This type of performance demonstrates that the lower weighted assets within each index were some of the better performers.
To date, the critical role of Bitcoin miners has been unhedged and solely dependent on the price of Bitcoin. However, as the mining market continues to mature with the inclusion of traditional market participants, these companies will seek mechanisms to hedge their exposure and operations much like they do with other traditional assets.
The current way of estimating hash rate, which involves a calculation process that includes a set lookback period (e.g. 48 hours), makes it difficult to design financial products that could help miners hedge their risk.
Coin Metrics has designed the CMBI Bitcoin Index and ‘Observed Work’ as a more reactive, responsive and manipulation resistant way to measure the realities of mining activity when compared to traditional hash rate estimations.
Observed Work and Coin Metrics’ CMBI Bitcoin Hash Rate Index can potentially serve as the foundational pieces of financial products that can provide markets with the required tools to effectively and efficiently trade and/or hedge Bitcoin’s hash rate.
We welcome your feedback on how to refine these foundational pieces and pave the way for new crypto financial products. If you are a financial service provider that would like to discuss the CMBI Bitcoin Hash Rate and/or Observed Work, please reach out to [email protected]
With the upcoming Bitcoin halving, there has been much speculation about the impact it will have on hash rate. For the majority of the crypto community, this is a fun, speculative exercise with relatively low stakes. However, for the mining community, the outcome can not only dictate profitability but it can dictate the probability of survival.
This is largely due to the absence of a robust, market-wide accepted methodology for hedging mining operation uncertainties. In this week’s SOTN feature, we propose two new tools that will enable financial derivative markets to effectively provide a mechanism to hedge and speculate on Bitcoin’s hash rate:
The CMBI Bitcoin Hash Rate Index
But first, a quick recap on the importance of Bitcoin’s hash rate to the cryptoasset ecosystem.
The Importance of Bitcoin’s Hash Rate
The activity of cryptoasset mining is one of Bitcoin’s core functions and was one of Satoshi Nakamoto’s key innovative ideas. Simply put, without mining, neither Bitcoin nor cryptocurrencies in general would likely exist today. Mining helps to:
Secure the network, prevent corruption and disincentivize bad actors from tampering with the public ledger.
Mint new Bitcoin to go into circulation.
Order and broadcast all transactions that have occurred on the ledger.
Validate and append new transaction information to the ledger to allow users to transact in a trustless manner.
Miners can generally understand their costs, which are predominantly a function of Hardware / Facilities (Capex) and electricity to run the mining rigs (Opex), which, given a fixed amount of hash rate on the network, allows them to determine their ability to make a profit under particular Bitcoin price conditions. However, the total number of hashes (computing power) performed on the Bitcoin network is not constant or predictable. Rather hash power fluctuates significantly over time, and not always in line with the price of Bitcoin.
For this reason, the ability for miners to hedge their hash rate and improve their ability to maintain profitability through a broader array of price and hash rate scenarios is critical.
For example, consider a large institutional mining operation that is deciding whether or not to enter the market. They have the budget to acquire equipment that today will give them enough mining power to capture 1% of mining / hash rate. With this, they can expect to receive, on average, 18 BTC per day. At a price of $8,000 per Bitcoin, if the operational costs are more than $144,000 per day (at today’s reward level), a miner will not make any profits and therefore should not enter the market. However, if their operating costs are $100,000 per day, they will have a nice profit margin and should consider entering the market.
Three months later the mining rigs arrive at the facility and Bitcoin has gone up to $10,000 per coin, but Bitcoin’s hash rate has doubled. Now they will only have 0.5% of the total hash rate, representing a reward, on average, of 9 BTC per day. At $10,000 per BTC, they will clear $90,000 per day in revenue, for a net loss of $10,000 per day, assuming a maintained operational cost of $100,000 per day. This does not bode well for the longevity of their business.
However, if they were able to hedge their exposure to mining operations by trading hash rate derivative products, they could minimize their exposure to macro shifts in hash rate.
Note: The example utilizes illustrative figures and ignores the impact of the upcoming halving for simplicity’s sake.
Designing the Tools Required for a Hash Rate Financial Product
In a distributed process like mining, it is near impossible to obtain reliable hash rate figures from the universe of miners. Therefore, the current best practice of deriving hash rate is to generate an implied value from the rate at which blocks are produced at a given difficulty level. This approach might be like saying traders of oil futures used the price observable at gas stations to calculate the amount of oil being pumped globally. Essentially, this is deriving a price (or hash rate) right now, from historical data.
For hash rate, this introduces many undesirable issues when creating a financial product. Coin Metrics identified three key issues with a simple hash rate index that had to be overcome:
The predictability of the short term levels. Since the hash rate calculation depends on past data, the majority of the data used to generate short term future levels is already known. For example, if the hash rate is computed using a 48hr lookback window, you already have 47hrs out of the 48hrs of data points that will be used to calculate hash rate in an hour. Therefore, this short term future hash rate is relatively predictable.
Given the random block generation process, implied hash rate tends to follow an oscillating pattern (as can be observed below). This poses two types of settlement risk for contracts. Firstly, there is randomness as to whether or not a contract would settle at the top or bottom of an oscillation, which could significantly impact the outcome of a trade. Secondly, this rate is highly manipulatable by some of the large miners that control significant portions of hash rate.
A fixed contract length on hash rate does not account for what happens between the contract open and close. Imagine a 3-month contract opens and closes at the same level. If a miner wanted to hedge their position by longing this contract they would make $0 at settlement. But if the hash rate average over the period was 20% higher than the open/close rate, they also would have realized lower revenue than expected. This is a lesser consideration since, in theory, one could trade throughout the contract to overcome this, but one that could be overcome through some innovative design.
Another approach to developing derivative financial products to speculate and hedge hash rate is through the use of difficulty. While difficulty provides some benefits above a hash rate derived index, it too has some issues that need to be overcome:
Difficulty only adjusts every 2,016 blocks (~2 weeks). This means that in the early parts of the contract, it is incredibly difficult to price, as estimating the future difficulty is essentially impossible and subject to significant fluctuations.
Long term difficulty contracts do not consider the difficulty levels throughout the duration of the contract. Difficulty can start and finish at the same level, but if it was higher through the middle of the contract, it does not act as an effective hedge (unless the contract has the liquidity to be managed in real time).
Difficulty can be susceptible to heavy manipulation. Perhaps somewhat surprisingly, in the days before a contract closes miners can significantly impact the outcome of the next difficulty adjustment by deliberately switching off equipment.
Coin Metrics has reduced the impact of many of these issues by designing the following tools, together which can form the basis for an effective derivative market hash rate product.
1: The CMBI Bitcoin Hash Rate Index
There is no definitive way to understand the amount of hash rate that is being contributed to the Bitcoin network. Rather, an implied hash rate can be calculated by looking at the recent historical time it takes miners to produce blocks.
Bitcoin was designed to have an average block time of 10 minutes, and every two weeks, Bitcoin’s difficulty adjusts to maintain a 10-minute average block time. Since solving for Bitcoin blocks is a random process that follows a Poisson distribution, the time between blocks can vary greatly.
This can lead to volatility in the determination of hash rate on the network. The industry standard to date has been to view hash rate using a 24hr lookback. However, for a structured financial product, Coin Metrics deems this to be too unpredictable and volatile. For that reason, we have introduced and will leverage a 48hr lookback for the CMBI Bitcoin Hash Rate Index.
This was especially evident when 3 blocks in a 24hr period took over 50 minutes to mine in September, resulting in the industry-standard implied 24hr hash rate to drop over 30%. However, this was just likely the result of a random, low probability but explainable outcome. Whilst the 48hr lookback period was also impacted, its <20% fall was less severe. We discuss this and more in issue 19 of SOTN.
More generally, in the image above you can observe that the 48hr lookback follows a lot less volatile movements than its 24hr counterpart.
2: Observed Work
As discussed above, it was not enough to release a hash rate index given the predictability, undesired randomness (from the oscillations) and tradability issues of traditional hash rate calculations. For this reason, we created Observed Work.
The traditional chainwork calculation assigns a fixed number of hashes per block based on the current difficulty (i.e. between difficulty adjustments, regardless of whether a block takes one second or one hour to find, chainwork will assign the same value).
To better reflect the work that is done by miners and the number of hashes conducted over a fixed period of time, Coin Metrics’ Observed Work is calculated as follows:
By introducing both the implied hash rate level and the time taken to find the most recent block, this representation of work conducted is more reactive and responsive to the realities of mining activity when compared with chainwork.
Coin Metrics’ Proposed Observed Work Futures Contract
Recall from above that a miner knows the number of hashes that their equipment can produce, but not what other miners can and will produce in the future. Based on the implied hash rate, a miner understands their current share of total hash rate and thus their expected revenue / share of block rewards.
Observed Work has been developed for financial service providers to build structured financial products:
For market participants to speculate on hash rate.
For miners to effectively manage their hash rate exposure by being able to hedge against the total observed number of hashes over a period of time.
Bitcoin’s difficulty level provides great insight into the network’s expectations of hash rate over each 2,016 block epoch. For this reason, a financial product that utilizes Observed Work would allow users to effectively trade expectations of the number of hashes with the unknown reality of hash rate movements over fixed timeframes.
Below is a theoretical example of an Observed Work 50-minute contract (3,000 seconds), assuming:
The market expects the hash rate to be 100 exahashes per second at time t=0
Bitcoin blocks are expected to take 600 seconds per block as defined in the Bitcoin whitepaper
Note: The implied hash rate values are directionally correct but for simplicity use rounded and easily digestible values.
At the opening time of the contract, a reasonable expectation of settlement price would be 300,000 exahashes (3,000 seconds ✖ 100 exahashes per second). However, as evidenced below, despite the implied hash rate closing at the same level that it opened, the contract would adjust over time and close higher than expected at 308,050 exahashes.
Walking through this result block by block:
Block 1 takes the expected 600 seconds to be found, thus there is no change to the implied hash rate and the observed work of 60,000 exahashes equals the expected work.
Block 2 is found in 10 seconds, which will increase the implied hash rate as block time was less than the protocol defined 600 seconds. At time 0, the expected work in 10 seconds was 1,000 exahashes (10 seconds * 100 exahashes per second). However, since the hash rate went up, the observed work increased to 1,050.
Similarly, block 3 was a fast block and therefore the implied hash rate increased and observed work was higher than what was expected at time 0.
Block 4 is the first slow block, taking over 600 seconds and thus reducing the implied hash rate to 105 exahashes per second. This is still higher than the 100 exahashes per second that we expected at time 0, which results in this block too having a higher observed work than expected.
Block 5 is another slow block, resulting in an implied hash rate of 100 exahashes per second, the same rate at which hash rate started. Given this, the observed work for this block equals the expected work, producing 119,000 exahashes.
This scenario is not uncommon in the Bitcoin protocol. One recent period that demonstrated this result very clearly was in mid-January this year (depicted below). As can be observed, the implied hash rate began and ended the 2,016 block period at approximately the same value. Despite this, the difficulty rose 5% since hash rate spent the majority of the period above expected levels. If a miner were to have taken a long hash rate position during this contract term, they would have not achieved the returns forecasted at the start of the period from their mining operation, as the average hash rate was higher than expected. Additionally, they would have made very little profit on the long hash rate futures contract because hash rate closed around the level that it opened.
However, exploring how a two week observed work futures contract over this period would have performed, it is evident from the below that:
The total amount of work conducted by miners throughout the period was higher than expected.
A miner with a fixed rate of hashes per second would have received less yield through the period than they expected.
If the miner was long this contract, they would have profited from the increase in observed work throughout the period.
This can further be modeled over longer contract lengths to provide long-term exposure and hedging to hash rate as may be required (e.g. hedging between ordering and receiving equipment). Below is an example 3-Month Observed Work Contract from the first difficulty adjustment of 2020. It can be visually observed how such a contract’s expectations would change over time as more information came to light and the contract settlement date got closer.
From the above examples, we can observe that such a structured financial product would overcome many of the issues that have hindered successful hash rate products as discussed earlier:
Predictability – the work conducted during a contract is always increasing, making it less susceptible to the predictability issues that hash rate faces due to its oscillating pattern. Further, whilst expected work is well understood, the observed work that takes place is highly dependent on the randomness of block times coupled with fluctuations in hash rate. To this extent, the divergence between expectation and observed work is less predictable than short term hash rate movements.
Measure performance over the duration of the contract – hash rate contracts could close at the top or bottom of an oscillation, which introduced an unwanted random risk to traders. Further, hash rate levels do not reflect the behavior of the metric over the duration of the contract. Observed work reflects the whole history of ‘work’ over the contract, and isn’t as impacted by the oscillating pattern of hash rate.
Manipulability – Given the large amount of hash rate that some miners possess, they could significantly and rapidly impact both hash rate and difficulty. Observed work can improve the manipulation resistance of hash rate products by adding time-weighted dimensions that follow a random Poisson distribution.
Mining is one of Bitcoin’s core functions and innovations that has allowed us all to benefit from a decentralized, distributed and sovereignless currency. As such, hash rate is a very important on-chain metric that provides markets and network participants with an indication of network strength and security.
To date, the critical role of miners has been unhedged and solely dependent on the price of Bitcoin. However, as the mining market continues to mature with the inclusion of VC-backed operations and traditional market participants, these companies will seek mechanisms to hedge their exposure and operations much like they do with other traditional assets.
Together, the CMBI Bitcoin Hash Rate Index and Observed Work hope to be the foundations of financial products that can finally provide markets with the required tools to effectively and efficiently trade and/or hedge Bitcoin’s hash rate. We welcome your feedback on how to refine these foundational pieces and pave the way for new crypto financial products.
If you are a financial service provider that would like to discuss the CMBI Bitcoin Hash Rate and/or Observed Work, please reach out to [email protected]
The major cryptoassets had their strongest week since the Mach 12th crash, with Bitcoin (BTC) leading the way. BTC market cap grew 17% week-over-week, breaking the recent trend of Ethereum (ETH) outperforming BTC. Despite this, ETH also had a relatively strong week, with market cap growing 12.8%.
BTC and ETH usage also continued to trend upwards and recover after the crash. BTC fees are up 170% week-over-week, which signals a large surge in network demand. As a result, BTC’s fee-to-revenue ratio (a key indicator of network health) reached over 6% on April 30th, its highest level since June 2019.
There is now over $1 billion of Tether issued on Tron (USDT-TRX). Over $350 million USDT-TRX has been issued since April 1st.
Additionally, there is now over $5.6 billion Tether issued on Ethereum (USDT-ETH), and $1.34 billion on Omni (USDT). If considered separately, USDT-ETH, USDT, and USDT-TRX are the first, second, and third biggest stablecoins by market cap.
One of the interesting things about the crypto asset management industry is that there is no consensus on what to benchmark returns to. Three logical candidates have emerged: use Bitcoin only, use Ethereum as a proxy for altcoin returns, or use a market capitalization weighted index.
Asset management in crypto is hard because not only do fund managers need to be right on how much long or short exposure to have, they also need to be right on the mix and weighting of the assets in the portfolio. Cryptoassets tend to fall into certain market regimes where one of the three candidates vastly outperforms the others, and correctly predicting which regime we are in is one of the key alpha producing decisions a fund manager can make. Depending on which benchmark you decide to use to assess fund performance, the dispersion in returns can be so large between these three candidates that it can be tough to tell if a fund is outperforming or underperforming.
This previous week is one instance where Bitcoin (+16%) vastly outperformed most other large capitalization cryptoassets. For the past several months, Bitcoin, Ethereum, and the long tail of altcoins have more or less performed similarly, but we are starting to be in the phase of the cycle where large divergences could be possible.
During the last cycle that ended in January 2018, it was Bitcoin that led to first run up. After wealth was made in Bitcoin, capital was shifted to Ethereum and altcoins. Eventually, the bubble crashed in part because all remaining buyers were exhausted and because the launch of so many altcoins raised the global supply of cryptoassets to unsustainable levels.
Tether Market Share Grows
Many respected industry observers expected Tether to fail, but it continues to defy its critics. The most salient criticisms assert that Tether would eventually crumble under its own weight due to its lack of transparency, investigations by government regulatory organizations, and its troubled banking relationships.
What was underappreciated by the market is how these characteristics actually make Tether more useful to certain market participants. By operating in a legal gray zone and taking a stance that it will not operate in a regulatory-compliant and transparent manner, it has attracted all sorts of traders that need these protections. Tether issuance is through the roof and its lead in the stablecoin market is still unrivaled.
Here we show Tether’s volume market share for Binance. Back when Binance launched, most volume was still denominated in Bitcoin or Ethereum. We see some declines in Tether market share in late 2018 as competing stablecoins such as USD Coin and Paxos Standard launched. But since then, Tether has crowded out almost all other cryptoassets, and the trend shows no signs of slowing down.
All CMBI and Bletchley Indexes had another good week off the back of continued global market strength. After a month of underperformance, the more risky low-cap assets performed the best this week, returning almost 13%. The CMBI Bitcoin Index and CMBI Ethereum Index also had strong weeks, returning 6.9% and 7.6% respectively.
After a March to forget, cryptoassets bounced back strong in May with an almost uncanny uniform performance across the top 70 assets, with the Bletchley 10, 20, 40 and Total all returning between 34% and 36%. However, it was the CMBI Ethereum Index that outpaced all other indexes, returning 58.7% in what was its second-best month in the past two years.