By Jason Victor
April 2, 2019
Last night, a trader sold $100 million into an equivalent amount of Bitcoin over the course of several hours. The trade was executed — I'll be generous — "naively" through a simplistic method of buying a constant value of Bitcoin every hour. The unanticipated consequences were huge, and cost the trader plenty — our estimate is $19 million in losses!
Whether it was an algorithmic misfire, a novice trader, or simply a bad algorithm, we'll never know. Here's what we do know so far:
We'll go through the story in more detail and discuss these three points in detail, so the next $100 million buyer doesn't make the same mistakes — both for their benefit, and for the benefit of the industry at large.
(Note: this analysis comes before we've had the chance to fully analyze the quantitative data available ourselves, which will be the subject of a future post.)
The trade itself wasn't particularly special: $100 million is big, but in fairness, at Routefire we routinely see these sizes get done with no market impact and in total stealth. That the trader lost $15 million over the course of the trade is truly shocking to us.
The trader in question used a naive algorithm to enter the order. Specifically, this algorithm does not seem to have taken market conditions or liquidity into account. Its behavior in the presence of the market impact it was itself creating leads us to believe it was shockingly naive: "every X seconds, buy Y BTC at each of three exchanges, over and over until complete."
This cost the trader dearly. Not only is there a substantial change in price that is projected by most to retrace at least halfway, if not most of the way, to the prior steady-state level — which would add more to the $4 million he has already lost since the trade completed — the trader would likely have incurred extremely high fees for executing this way. All while collecting $19 million in losses along the way.
Long story short: either there's something in the data we'll find that provides an explanation, the trader and/or algorithm was extremely naive, or the trader was intentionally looking to change the price of Bitcoin, even if at his own cost.
After a big trade happens, financial theory would dictate that arbitrageurs would plug the holes created in various order books and force price discovery across exchanges, even if the big trade happened on only a couple exchanges (as in this case).
And that's precisely what happened. As we discussed in our talk on responsible liquidation, there is a process of arbitrage and price discovery that follows a heavy-impact trade. (Video will be posted to this page when available.) Indeed, this happened to such an extent that top-of-book dispersion — a measure of the degree to which exchanges agree with one another on price — is at 1.5%, its lowest level in months, evidence of the fact that the market machinery for price discovery worked beautifully.
And thus, what began as blowing through the order book on a couple exchanges changed prices throughout the market, swiftly turning a single person's trade into an international news story.
This might seem all well and good to the reader — and they'd be right if the only goal in the community were to see the BTC/USD rate increase.
Unfortunately, this kind of event has unwanted knock-on effects. For one, it demonstrates just how easy it is to dramatically influence the price of Bitcoin; the fact that no safeguards exist to prevent against market manipulation was the principal reason for the SEC's decision not to allow a Bitcoin ETF, and this story proves them (unfortunately) correct.
Beyond giving the regulators more to worry about, it also increases the cost of doing business in Bitcoin. Events like this increase traders' view of volatility. This means they quote higher spreads on trades, derivatives cost more, and payment services need to charge more. The ripple effect up the value chain is material and will be felt for some time.
And finally, for the trader who made the buy, the worst part of all is the $19 million he lost for no reason. A $100 million Bitcoin buy on Routefire could be configured in many ways, none of which would have cost the trader so dearly. Indeed, our estimate is that a well-executed algorithmic order (like those offered via Routefire or some OTC desks) would have saved the trader $19.1 million net of algorithm usage fees (assuming Routefire was used — OTC fees are higher.)
Moving a market by unexpectedly huge amounts is destructive regardless of price direction — both for the trader and the industry.
The unfortunate part to me is that it was such a preventable mistake. A passive TWAP would have filled this in around 24 hours effectively for free; an active VWAP would have gotten near-spot pricing and finished quickly; and a PVOL would almost certainly take a couple days, but it would be doing so with an average cost better than spot.
We've always reminded our users to "liquidate responsibly," and as we move into what may be the next phase of the market cycle, we would like to also remind you of one other thing:
Buy responsibly, and don't throw away $19 million,