Token Distribution Ratio: New Performance Indicator for Crypto? [Santiment Analysis]

Apr 19, 2019

One of the most recent metrics we at Santiment built for ERC-20 projects is the Top Holders Distribution. It gives an overview of the top 5-100 wallets of any ERC-20 token in our database since the ICO, both in net value and as a percentage of the token’s total supply.

For example, here are our Top Holder charts for BNB:

The addresses are also further divided by exchange and non-exchange wallets, so you can see exactly how much of a token’s supply is/was stored on exchanges at any point since the project’s launch.

1. Defining token ‘distribution’

Why are these metrics useful? For one, they give us a pretty good idea of how ‘distributed’ (or utilized) a project’s native cryptocurrency actually is. Simply put, if 90% of your project’s tokens are lying dormant in a handful of big-boy wallets, your network adoption has a ways to go.

Similarly, if half a token’s total supply sits in exchanges, that’s a strong indicator the coin is still mostly used for speculation, instead of whatever its purported utility.

Whether that’s good or bad will have to be determined on a project-to-project basis. That said, if a token’ is designed primarily as a robust in-game currency, it doesn’t make sense for most of it to sit on Huobi, does it?

In short, a ‘healthy’ token should be broadly distributed across the project’s network/addresses, and relatively sparse in exchange wallets.

With the help of our new metrics, we decided to test whether this token ‘distribution’ can help us predict its performance over time. Put another way - do distributed tokens perform better than concentrated ones?

Let’s find out.

2. Our methodology

To determine if (and how much) a network is distributed or concentrated, we compared the amount of tokens in the 100 biggest wallets to the total current supply of said token. Put simply, we’ve calculated the % of a coin’s total supply that’s sitting in the 100 largest wallets.

So the formula for a token’s “distribution ratio” is:

Don’t worry - that’s the only formula in this article, I promise!

To find the most distributed tokens each month and test their performance, we needed to calculate this ratio for all 700+ ERC-20 projects in our database - and not just once, but on a monthly basis.

After excluding the projects with a market cap of <1 million and the trading volume of <10k to

remove the biggest flops, we were able to build a portfolio of 15 most distributed projects, which we recalibrated each month.

For reference, we also created a portfolio of 15 most concentrated projects month-by-month.

Finally, it was time to pit the ‘most distributed’ vs the ‘most concentrated’ portfolio, and see which one comes out on top.

First, we evaluated the portfolios’ performance over the last 2 years. Here are the results:

The scale on the above chart represents the portfolios’ cumulative returns. Both start with an initial value of 1, so a final value of 2 would mean the portfolio’s value has doubled (100% profit) in the observed time period. Think of it as the amount of money you’d get back if you invested $1.

As we can see, the distributed portfolio (black) clearly outperforms its concentrated counterpart:

The distributed portfolio recorded a final value of 0.76 , which is a loss of 24%.
The concentrated portfolio recorded a final value of 0.22, which is a loss of 78%

Naturally, the end results are highly affected by the prolonged bear market of late 2018/early 2019. Still, all other things being equal, it would seem to make sense to opt for the more distributed projects when building your ERC-20 portfolio.

3. Benchmarking the Outcomes

While it does beat ‘most concentrated’ projects, we still haven’t proven if a distributed portfolio makes a sound trading strategy in general.

As we’re only observing ERC-20 projects in this analysis, let’s add the performance of Ethereum(red line) as a benchmark:

While more volatile, our distributed portfolio beats the ETH benchmark at almost all points in time. Holding Ethereum over this period would give us a value of 0.66, or a 33% loss.

That’s just one arbitrary time frame however. How about some more recent development? Here’s the performance of all three portfolios (distributed, concentrated and ETH) from mid 2018 onwards:

Again, the portfolio which consists of the most concentrated projects is vastly underperforming. With a value of 0.072, it lost ~93% by the end date.

And while the observed period hasn’t been kind to most crypto assets, our distributed portfolio and Ethereum have both performed better.

In a close race until the very end, the distributed portfolio scored a final value of 0.145 (a loss of 85%), while HODLing Ethereum clocked out at 0.235 (a loss of 76%). Still, we underperformed the benchmark in this case.

Of course, we also tested different variations of portfolio sizes and market cap limits. The distributed portfolio outperformed the concentrated one the majority of times.

In another test, we limited our analysis to the top 50 projects by market cap, and compared the performance of the 10 most distributed and 10 most concentrated assets. Here are the results for both, once again plotted against the Ethereum benchmark:

As before, the distributed portfolio outperformed the concentrated assets significantly, proving the same effect also applies to top performers overall.

[For those with a more statistical background:

We of course got a bit more statistical to see if there really is a statistically significant connection between the distributednes of a crypto project and its returns. For this we took the monthly distribution ratios for all the projects and regressed it against their monthly return (output in the appendix).

The result showed a significant (p=0.045) negative (slope=-0.2307) relationship between ratio and returns. In other words, the smaller the ratio (more distributed) the higher the returns. Or more precise: If you would manage to decrease the ratio by 0.1 the model would predict an increase of monthly returns by 2.3 percent. The r-squared was expectedly low, given that we do not expect to actually predict the price movement exactly with this metric but rather to see an overall trend. So with this model we show that there is in fact a statistically proven relationship between distributednes and returns.]

4. Long-short Strategy

Finally, let’s try some ‘advanced’ investing using our new distribution ratio.

At this point, our distributed portfolio based on all 700+ ERC-20 projects beats the concentrated portfolio and performs OK against the ETH benchmark. But we’d still be losing money with it.

Luckily, we now have a custom valuation metric that generates one “good” and one “bad” portfolio. We can then do it like the hedge fund pros, and execute a long-short strategy on the two distribution-dependent portfolios.

In a nutshell, a long-short strategy (or “market neutral” strategy) picks good and bad assets based on a certain criteria, and then equally buys the good and shorts the bad apples.

The idea is to cancel out any major shifts in the market; if the market goes up overall, you make money on the assets bought, and lose money on the shorts. If the market plunges, you make money on the shorts and lose on the buys.

This way, the market movement is removed from the equation and you just profit purely based on how much your ‘good’ assets outperform the ‘bad’ assets.

So we have our custom criteria (distribution) for picking the long-short targets, which we would recalibrate each month. Here are the results of such a strategy, compared to HODLing ETH:

Impressive! Not only does a long-short strategy (blue) generate a value of 4.37 (more than 300% profit) it also exhibits a relatively-stable growth (at least by crypto standards) even during the bear market.

2018 until now was one of the worst periods for crypto (from a market perspective). In that time frame alone, our long-short strategy generated around 50% profit.

So... did we just find the holy grail? Well, the above results definitely needs to be viewed with some caution:

For starters, just the process of implementing a long-short strategy is relatively complex and fairly costly. The monthly rebalancing of two portfolios of this size comes with significant transaction costs not included in this calculation, as it varies strongly between platforms.

Also, not many platforms out there facilitate shorting altcoins at the moment. That said, with the rise of projects like dydx that work on all kinds of derivatives implemented via smart contracts, shorting an ERC-20 portfolio is definitely within the realm of possibilities.

5. To the first few reading this...

With all this in mind, it might pay to know the most concentrated and most distributed assets in the top 50 at this very moment.

Pen and paper time. At the moment, the 10 most distributed coins in the top 50 are:

At the moment, the 10 most distributed coins in the top 50 are:

This list changes as whales move their coins and the projects' distribution ratios adjust. Given the potential value of this information, we are creating a very small, first-come-first-serve email list, where traders will receive monthly updates to the top 10 most concentrated and top 10 most distributed projects to use in their portfolio building.

There is no sign up form. If you want to join, send me an email titled 'Distribituviness ratio' to [email protected]

6. Conclusion

Overall, our decentralized portfolio is by no means a perfect investment strategy. Even though it’s not terrible against the benchmark and massively outperforms the centralized portfolio, its performance is still somewhat time dependant.

That said, our results certainly indicate that a more distributed network performs better than the one that’s centralized, making our distribution ratio a valuable litmus test for crypto - one that could definitely see play in portfolio building.

Of course how much distribution is good/required is at least partially dependent on the purpose of each considered project.

However, given the limited amount of info still available in crypto, the distribution ratio could help us gain a better understanding of how much any token is (or has been) actually used, making it a novel indicator of any network’s quality and health.

7. Appendix:

Result regression analysis (monthly returns ~ distribution ratio):