Does Github Activity Affect Crypto Prices?

Ever wonder if developer activity drives cryptocurrency price? We explored if there are any trade-able insights to be gained from GitHub commits.

The research below is an excerpt from our Q4 Quarterly Report in partnership with eToro! Read the full report here:

The TIE & eToro Q4 Quarterly Report

Does Github Activity Affect Crypto Prices?

Most crypto-traders identify buy opportunities by analyzing market charts, reading recent news updates, or following social media trends. Few, however, utilize the number of Github commits as a measure for projected value. Here, we take a look at whether Github activity can (or should) be used as a measure of crypto value.

Figure 1 shows the price of six cryptoassets over 2020 (black lines) with the number of daily Github commits overlaid (blue). At first glance, there is no obvious, consistent relationship between github activity and price – the two variables seem to fluctuate independently of each other.

Crypto Github

To test whether there is any meaningful signal, we backtested a relatively simple trading strategy; if the number of github commits in one day is high (top 75% of all days), buy and hold that asset for 14 days. Otherwise, hold all six cryptoassets equally. Figure 2 shows the performance of this strategy (red line) compared to simply holding all six assets equally (black line), as well as the performance of each of the six assets individually (gray lines).

Github Altcoin Trading Strategy

We see that this strategy did indeed outperform the simple buy-and-hold strategy. However, this could have been due to luck – in fact, a strategy that randomly traded these six assets outperformed this commit-based strategy about 20% of the time.\ \ Perhaps the number of Github commits is not enough information on its own to measure increased value of these assets; commits can measure many different types of activity including simple bug fixes or documentation updates. Figure 3 shows the last four years of github activity associated with Ethereum broken up by repository.

There are some interesting patterns in this figure; for example, the change from Parity development dominance to Solidity, the brief bursts of development on the Sharding and Casper repositories, and the end of aleth development. These unique events can be quantified by measuring how the ratios of colors change over time in figure 3. Perhaps this measure of commit changes is a better indicator for notable changes in project development than the number of commits alone.

We compared these ratios of commit types at each point in time to the ratios in the preceding two weeks to get a measure of how different the types of commits are on each day. Following the simple trade strategy described above, figure 4 shows how we used this level of Commit Change as a potential buy signal; when the change is greater than 75% of changes seen (horizontal line) buy and hold for 14 days. Otherwise, hold everything equally.

This strategy seems to get some fantastic buy opportunities – the two late Fall signals on XMR, the first on EOS, and the second-to-last on XLM are a few worth noting. Indeed, this strategy does outperform the strategy of simply looking at the number of commits as shown in figure 5. This implies that if commits are indeed having an effect on cryptoasset values, it is likely more complex than simply the number of commits, but rather the content of those commits.

To measure the significance of this strategy, Figure 6 shows the final performance of 10,000 backtests that randomly traded these six cryptoassets.

The purple vertical line shows the final performance of the commit changes strategy while the red line indicates the performance of the first strategy. With less than 1% of random strategies outperforming the commit-change strategy, it may seem like convincing evidence that Github activity should be a main consideration for traders. However, there is still a chance that this result can be attributed to luck.

To verify if these strategies are significant, we back tested them on more historical data. It does seem likely that the results we found for 2020 may have indeed been attributed to luck; no single strategy was consistently the best across all the years as shown in the table below. However, it is more likely that the results of these strategies were attributed to luck in the historical data, especially the 2016-2017 backtests. This is because one lucky position in Monero (XMR) in 2017, for example, can return a nearly 4,000% increase alone. If a strategy was able to randomly capture only a few of these highly lucrative events in the 2016-2017 markets, it would define the entire year’s performance.

The honest (potentially unfulfilling) answer is it is unclear whether Github activity has an effect on price. The few strategies we tested in 2020 seemed to work well, but when they were back-tested against previous years their performances were inconsistent. However, due to the market’s highly volatile behavior (especially in 2016-2017) it is hard to signify the meaning of such backtests.

It is probably true that Github activity has some influence on market movements. However, there are many other factors that likely have much stronger effects. As the crypto markets mature, and we focus our attention on cryptoassets that don’t have erratic 50x gains, it becomes easier to identify more subtle patterns such as the potential influences of Github activity. It could very well be that 2020 marks the first year that crypto markets have matured enough to identify such signals. Only time will tell.

Contact us for a demo of our SigDev platform to view all market-moving events in real-time.