Price and Value in Initial Coin Offerings: Can we Tell the Difference?
Today I have a challenging article to write, but first I have an admission to make: Ether is presently below USD 300, I have no idea why, and I won’t really be working hard to find out. There are some claims that it is due to an Eth selloff due to ICOs cashing out. That may or may not be true (I suppose it sounds at least plausible), but fundamentally it’s not what most concerns me about the market right now.
One of the major use cases of Ethereum is ICO investment. At their core, ICO investors depend on two things: speculation, and liquidity. Speculation is an industry where marketing teams manufacture opinions on future prices. Liquidity is achieved by paying centralized exchanges to list your coin, and hiring market makers to buy and sell their own coin.
That’s not to say that quality teams and innovative products in the ICO space don’t exist — on the contrary, there are several ICOs whose products I fully expect them to complete and am very anxious to adopt once they are available. What I mean here is that the expectations of investors to make ten times (Moon! 10X!) their investment after a few months are not directly tied to products or teams beyond their function as a vehicle to further drive speculation and liquidity. Whether these expectations are tied in any way to reality is a story for another day, perhaps when I need to blow off some steam.
None of the above should be new or surprising to anyone involved in the ICO game for any substantial span of time. What I’d like to share with you is some data I gathered internally for discussion at Alpha Blockchain using a web scraper I wrote. It scrapes the web for ICO return on investment, measured in USD, at a given point of time. I had run it at the start of May, and then ran it again at the start of August in preparation for a short speaking engagement:
While the scales are different between the two graphs, and the ICOs being measured are at various stages in building their product, the message is clear — confidence in ICOs has plummeted. Looking a little deeper, when I created a normalized curve of ICO performance, the ‘average’ ICO was 60% likely to be profitable in May, and 49% likely in August.
For me, that represented a tipping point in confidence and my message to the audience at the resulting talk was one of concern: We seem to know the price of everything but the value of nothing (thank you, Oscar Wilde).
A few days afterwards, the Oscar Wilde quote slowly inspired a testable hypothesis — what if I scraped the ratings from a certain commonly-used ICO rating site and analyzed how well they explain today’s post-ICO token price from coinmarketcap? In short, does the value perceived by public ICO rating systems tie in to price at all? Do we really know the value of nothing?
I scraped the data and categorized the ratings into groups ranked 1–5, with 1 being the highest rating and 5 being the lowest (the exact sites used will remain anonymous). Then I ran a one-way ANOVA to determine whether there was a significant difference (α<0.05) between groups. For those that love statistics, here are the formal null and alternate hypotheses in this exploratory analysis:
H0: The token price between rating categories is not significantly different
HA: The token price between rating categories is significantly different
For those that don’t love statistics, an ANOVA is an ANalysis Of VAriance. It looks at the variance in token price between and within groups (our categories 1–5) and provides a measure of how well those groups explain the variance. The result of the ANOVA are provided in Figures 3 and 4 below.
I realize ANOVA tables are not commonly used in crypto articles, so I’ll go over the key elements. In Figure 3, the treatment columns are our 5 categories (1 is the best rating, 5 the worst). N is the number of samples (an ICO token price listed on an exchange) I found for each category. The Mean is the average ROI in USD expressed as a multiplier, e.g. 2.89 means if you invested a dollar, you would have $2.89 now. The Std.Dev. is the standard deviation, the average amount a given value differs from the mean.
In the Figure 4, what’s mainly interesting to us if the F-score. This is a measure of the probability of the observed difference between groups being due to random chance instead of due to an actual effect. Our F-score corresponds to a less than 0.001% chance of that being the case (my table doesn’t go any lower). It’s typical for anything below a 5% probability to be sufficient to reject our null hypothesis and consider our alternate hypothesis (that the rating categories explain some of the variance in token price) as correct.
One might argue though that the ratings website may adjust their ratings post-ICO to make it look like they are better at predicting token price than they actually are. While I can’t rule that out, I want to draw your attention to the means and standard deviations of each category in the first table. In every case, the standard deviation is very near to or greater than the actual mean. What this tells us is that while these public ratings do have some ability to explain the token price, in practice their ability to explain token price is not really sufficient. If the data were falsified in any way, I would expect it to paint a prettier picture!
What this all means to me as a consultant at Alpha Blockchain is that the services our clients need have changed, and I’m developing new tools to meet those changing needs. Since the rating systems I’ve tested above have real but insufficient predictive power, the first step is to develop and apply more effective quantitative tools to all projects we assess for investment. We’re pretty good at that already, but in a bear market it’s a matter of picking the best one or two projects rather than the best ten. As a result, we’re refining our tools to allow better and faster comparative analysis between projects with similar products.
The other angle is that we’re not going to be the only ones upping our game: other investors are going to be more rigorous as well. That means that when we partner with good ICO projects, we’ll be using these comparative tools to suggest specific improvements to areas such as token economics and allocations. The goal isn’t just to make projects look better on paper, but to bring strategic partnerships, help them build stronger teams and products, and become more sensible investments. It’s good enough to be the second-best project in your sector anymore, anything that’s not first place may as well be last place.
For me as a blockchain enthusiast, my position differs slightly. There are many ICOs out there that have proposed projects I’m quite excited to actually use! In the end, I want great ideas to be connected to funding, and I think a few things need to change for that to continue happening. We all know accountability for funds raised is lacking, and we all know that placing that much trust in a small, centralized team is not what blockchain is about.
We have smart contracts now. ICOs can happen in tranches, with staked and locked Ethereum only delivered if milestones are met, decided by investor vote. There are certainly better models out there developed by persons smarter than me. Trust is a weakness, so why not add some accountability to the ICO process?
