Determining the success of a UTXO fork is a fraught exercise. There is no agreed-upon standard for measuring adoption. Establishing the existence of genuine activity is therefore challenging to the say the least. The list of potential proxies is long: full blocks, a robust fee market, an elevated transaction volume, a high transaction count, to name a few. And, unhelpfully, given that forks are often competitive with the parent chain, and compete for scarce attention and resources, quantitative metrics which offer insight into the usage of a given chain are often targeted for spoofing or manipulation.
Usefully, however, forks are often similar in nature to their parents and so meaningful data-driven comparisons can be drawn. There are many reasons someone might want to dispassionately evaluate the uptake of a fork relative to its parent: custodians who must develop fork policies, exchanges deciding which asset to support, researchers interested in the drivers of success of competing systems, businesses selecting which protocol to build on, and of course users and investors determining whether or not to liquidate fork coins or to embrace the new chain.
In this post we discuss a few conventional approaches to comparing forks and introduce some novel metrics which can establish difficult-to-forge ground truths. By layering on alternative on-chain datasets, richer conclusions can be drawn.

Before we start, let us define our objectives. The task here is to find a measure that captures organic, non-contrived commercial or transactional activity and evidence of adoption on a given fork. In short, a metric which manifests the unique demand to use the blockspace of a given chain is sought. We’re limiting this analysis to UTXO forks of Bitcoin since metrics generalize easily – that is, they are eminently comparable – and they have the nice quality of maintaining ceteris paribus. That is to say, forks tend to hold all other things equal (the existing ledger of balances in particular) while altering a parameter or two (in the case of Bitcoin Cash, the block size, with Bitcoin Gold, the proof of work algorithm).

Popular methods

Market capitalization/price

While market cap, or network value, is a troublingly inconsistent measure when comparing projects with large treasuries or disparate issuance schedules, in the case of Bitcoin forks issuance schedules are roughly preserved and there are no treasuries to speak of. Some things confound the analysis slightly: Bitcoin Gold, for example, was subject to an additional premine of 100,000 units. Bitcoin Cash, due to an uncontrollable oscillating difficulty adjustment, raced about 84,000 units ahead of Bitcoin and is further along on the supply schedule.

Market cap straightforwardly evidences aggregate demand for any of these chains. However, it does not contain information relating to actual usage. If you were a business looking to build on a highly active chain, you might seek out a minority chain if it was clear that its users were more enthusiastic about transacting. Market cap also faces issues with supply: given that many forks have comparatively little uptake (activated coins) relative to a parent chain, it seems odd for them to inherit the parent supply. Imagine a Bitcoin fork where only 1000 units were activated – would it be appropriate to assign that fork a supply of 17.4m? We cover this issue in detail towards the end of this piece.

Transaction value

Transaction value, sometimes referred to as transaction volume, refers to the USD-denominated value of transactions occurring on a network in a given period. This same figure indexed to the value of the native unit is often referred to as unit turnover. Dividing yearly turnover by supply gives you annual velocity. It is also popular to divide market cap by transaction value to create a measure of usage relative to the size of the chain – this is called the network value to transactions (NVT) ratio.
Transaction value, while crucial in assessing a network’s usage and vibrancy, is exposed to significant complications. Firstly, in UTXO chains, it is heavily laden with artifacts and noise. Since most Bitcoin transactions include change, the raw output on chain in a given day will count these change outputs as the real thing, effectively double-counting outputs. This change must be netted out to derive a realistic view of transaction value; this is why we target known and inferred change in our adjusted transaction value metric.

Normal blockchain usage also has the side effect of inflating transaction value. Exchanges performing wallet management are a big culprit: the typical process for deposits involve users depositing funds to a unique address, which is then swept into a hot wallet. Liquidity is then managed between hot and cold wallets. This often results in triple counting for simple deposits. Mixers and Coinjoin services like the Wasabi wallet require multiple mixing rounds; these services generate a significant on-chain footprint as well. Our adjusted metric includes an early spend heuristic, which inserts a speedbump into wallets generating significant churn, targeting this class of activity. Eliminating these artifacts reduces transaction value by a full order of magnitude, and generates a more plausible estimate of actual usage.

We believe that the adjusted transaction value metric is a very useful approach to comparing UTXO chains which have similar characteristics.

Relative economic vibrancy can also be ascertained with the NVT metric. A lower NVT is a loose signal that a given chain has more usage relative to its price than its peers.

However, despite our efforts to refine transaction value and derive a metric which is less forgeable, significant opportunities for inflating the figure remain. Since transaction value and NVT are both considered glamor metrics – that is, they make a blockchain look attractive – they are obvious targets for manipulation. And even with our speedbump, large holders can generate an elevated transaction value by repeatedly cycling their coins around wallets that they own. A covert cat-and-mouse game has now developed between teams attempting to inflate transaction value, and blockchain analysts employing novel methods to subtract out the artificial churn. We consider this measure informative, yet fallible.

Transaction count

Long considered the metric of choice by analysts, onlookers, and journalists, transaction count has come under pressure in recent years. Transaction count is trivially gameable on chains where fees are low and blockspace is cheap. It is very common for teams to employ “stress tests” to show the promise and capabilities of their chain. These can spike transactions well above normal levels, confounding analysis. Since transaction count is considered to be indicative of usage and vibrancy, many teams have strong incentives to game it on their networks. 

Both BCH and BSV were subject to stress tests in late 2018 in apparent attempts to demonstrate their extensibility. In the last month, however, BCH and BSV account for 3 and 2 percent of Bitcoin’s transaction count, respectively.
Unlike transaction value, where there are very obvious candidates for exclusion, it is harder to distinguish meaningful transactions from the contrived ones. Other traits impair comparability. Take batching in Bitcoin, it is common for heavy users of the network to aggregate many payments to different recipients in a single transaction, for efficiency’s sake. These can contain thousands of total payments. Batched transactions account for about 30 percent of total output volume today and 25 percent of all outputs in Bitcoin today. If the users of a fork don’t have similar habits (because fees are lower, for instance, and efficiencies are not sought), transaction count will appear higher on the non-batched chain.

So transaction count, the perennial comparative metric of choice, not only fails to provide a good comparative basis among Bitcoin and radically different chains (like Ethereum, EOS, or Ripple), it also doesn’t quite capture the differences between Bitcoin and its own forks.

Block fullness

Full blocks are another interesting measure of demand to use the chain. They can prove a negative – empty blocks are clear evidence of no demand to use a given fork – but not a positive: blocks may be full due to genuine demand or due to block stuffing.

At present, bitcoin’s many forks appear to be in relative states of abandonment according to this measure. However, when some of these forks were new, blocks were stuffed full, perhaps as a proof of concept for larger blocks.

Aggregate fees

One measure which is often overlooked is quite simply the aggregate value of fees paid to transact on a given chain. This is a good measure of demand, and is very costly to fake. One way to game this would be miners paying themselves high fees to simulate a vibrant fee market. We view fees as important because Bitcoin and all of its forks are ostensibly capped in supply, and will have to replace the miner subsidy with fees once issuance declines. Attracting meaningful demand to transact is a prerequisite for the development of serious fees which could generate sufficient security, post reward. A total lack of fees demonstrates poor preparedness for a rapidly-approaching fee market regime, and evidences weak demand to use the chain.

Aside from BTC, fees are almost totally absent in Bitcoin’s forks. Combined fees on BCH, BSV, and BTG have averaged under $100 daily over the last month. Fees are not, on their own, a particularly rich measure; they don’t tell you much about the chain aside from whether demand to use block space is elevated or not.

Alternative methods

Through chain analysis and creative metric design, alternatives to the above popular metrics can be used to ascertain a richer set of insights regarding the adoption of forks.

Uptake

Uptake measures the number of units which have been “activated” or awoken on a given fork. Take a fork like Bitcoin Cash; rather than starting as a new chain, every holder of Bitcoin was granted equivalent numbers of Bitcoin Cash on August 1st, 2017. Owners then had to decide whether to recognize the new asset, split their Bitcoin Cash balance from their Bitcoin balance, and sell or hold the new coins, or simply do nothing.
Since forks are opt-in in this manner, and recognition of new coins leaves a mark on-chain, this is a natural phenomenon to track.

The above charts show the total number of units awoken on each of the forks over time; Bitcoin’s supply – the total possible set of units which could have been activated – is added for context. Analysis is fairly simple: if a fork receives little uptake, users are saying: “even though I have been airdropped what is essentially free money, it’s not even worth my time to recognize the gain and liquidate it.” In fact, most forks receive comparatively little attention and minimal uptake.

One interesting consequence of this analysis is that it casts circulating supply figures of these forks into doubt. The common approach for data services is to have fork coins inherit the supply of the parent. But if uptake is only 10 or 30 percent of the parent, it doesn’t make sense to assign them a supply which is mostly inert. In BSV, for instance, only ~4.5m coins were claimed out of a 17.5m total possible, yet the BSV market cap is calculated as if 17.5m units were actually circulating. Applying the uptake discount would grant BSV a market cap of $343m rather than its claimed $1,308m, a 74 percent difference. This phenomenon is another piece of evidence which undermines measures like “total market capitalization” or dominance indexes: forks inheriting their parents’ supply generates significant illusory economic value.

We like uptake as a measure of vibrancy of a fork, as it demonstrates user enthusiasm to cross the hurdle of recognizing those fork coins, which is often a nontrivial exercise. Uptake is nevertheless imperfect for estimating enthusiasm for forks; as exchanges and custodians get better at splitting and liquidating forked coins, the uptake rate should increase, without necessarily reflecting more enthusiasm for the assets. And if a user’s sole interaction with their fork coins is to sell them, this doesn’t necessarily reflect demand to use the chain – but it does get imprinted in the uptake.

Active supply

Another approach to determining the vibrancy of a forked chain is to look at the active supply. This means the fraction of number of units which have moved in a given period of time. This is distinct from transaction value, as active supply only counts a given unit at most once, while transaction value can count the same unit many times. What active supply is trying to determine is what fraction of supply is actually moving (if at all) during this period, while transaction value aims to ascertain the totality of economic volume. Active supply was designed to be more resistant to spoofing than transaction value, since the most an adversary can do is to make their coins look permanently active. A sustained active supply is a potential indication that a given chain has genuine usage, outside of just some small population of power users.

We currently track active supply for the trailing month, six months, and year, although it can be done for any period.

Interestingly, Bitcoin Gold looks rather inert by this analysis. The fraction of supply which is active is bottoming on all the timeframes.

One important caveat here is that active supply does not make any judgment about the dispersion of users. A chain where a single entity controls all liquidity and periodically refreshes their wallet would look just the same on an active supply basis as a chain with many participants transacting frequently. And indeed, the more concentrated supply is, the easier it is for a single entity to game active supply. So active supply must be heavily caveated – it is probably not very informative for chains with highly concentrated ownership.

Addresses with balance

Usefully, there exist measures which can give analysts insight as to the dispersion of the ownership of a given chain. One promising measure is the number of addresses with a nonzero balance. While these don’t perfectly correlate to individuals, it is a useful proxy. Some custodians own massive wallet clusters, either deliberately or through poor address management. The periodic declines in the figure are consolidations: the dramatic collapse in Jan/Feb 2018 came about because fees finally declined, enabling custodians to consolidate cheaply.

Addresses present in the UTXO set with a nonzero balance is an interesting measure, but it is possible to semi-cheaply forge it on a low-fee chain, by sending dust outputs to thousands or millions of new addresses. Thus we devised an alternative version: addresses with a balance totaling at least one billionth of present supply.

In Bitcoin, a qualifying address requires a balance of about $62. We indexed it to supply, rather than a dollar threshold (say, $10 across all chains) in order to capture the actual supply dispersion, and to insulate the metric from changes in price.
This measure is illustrative: you can clearly see the rapid expansion in late 2017 as Bitcoin gained adoption and a new user base; through 2018, Bitcoin consolidated and continued to improve its dispersion (the early 2018 reduction in addresses with a balance was a consolidation of UTXOs after fee pressure declined). Bitcoin Cash and Bitcoin Gold, on the other hand, were most dispersed at their moments of inception: what followed was a concentration, in which the base of unique holders actually contracted.

This is consistent with the theory that Bitcoin Cash and Bitcoin Gold were redistributive – indifferent sellers recognized their coins and sold them off for more BTC, with a few large committed buyers scooping up this liquidity. The continued contraction is potential evidence of stagnation and an inability to onboard new users.

Paired with the active supply analysis, this suggests that while BCH has a good fraction of active supply, it is held in a more concentrated user base.

UTXO set size

The below chart shows the size of the UTXO set by fork. It is a nice visual demonstration of how forks inherit the UTXO sets of their parent chains and build on them or concentrate them. In BCH, the UTXO set shrank significantly after the fork, potentially demonstrating a re-concentration as BTC holders liquidated their balances and that loose supply was aggregated by buyers. The rapid increase in UTXOs on BCH appears to be a consequence of the stress tests run at that time. BTG also been subject to a concentrative slide for its entire existence.

While useful and interesting, active supply, addresses with a balance, and the UTXO set size alone are not sufficient to characterize the concentration of new chains. A fuller analysis might involve determining user balances on chain and on exchanges to derive a measure of inequality like the Gini coefficient, although this presents significant difficulties.


In this post, we introduced several new measures to determine how active, vibrant, and dispersed new forks are: in particular the notions of uptake, active supply, and addresses with a balance. Together, these metrics give observers a comprehensive and nuanced understanding of the vibrancy or morbidity of a given fork of Bitcoin. While no metric is individually sufficient to analyze the uptake and success of a fork, taken together, a pattern of evidence can be built. The notion of uptake in particular challenges market cap, suggesting that large fractions of the troubled metric are entirely illusory. We believe that these metrics can be powerful tools for entities which require hard-to-forge quantitative measures of enthusiasm and usage for major forks.