By Jason Victor
August 22, 2019
It's hard to tell when you're doing execution well; it's easy to tell when you're doing it poorly.
This is the first part of a series about execution quality benchmarks and statistics. Today, we'll discuss perhaps the most important of execution quality benchmarks: we call it the liquidity-adjusted arrival price.
Before we discuss the details, let's recap the basics of execution quality. In general, we assume we have a trader looking to buy or sell some quantity at the best available price. By execution quality, we mean the trader's ability to obtain a good price.
So what's a good price? That's the critical question, and it's the one we'll address today. Traders interested in optimizing execution quality are usually doing fairly large volumes; for this reason, market impact — or the process of adversely affecting available market prices due to the force of one's own trade — is often the primary concern. But the things we'll discuss today don't only impact whales: indeed, this affect all traders. Even a 50,000 USD trade can create waves in crypto, depending on the circumstances.
Of course, it's a murky question what a "good price" is. While it's crystal clear what the trader's realized price was, it's always difficult to say what a "good" price might have been. It's impossible, for example, to predict how the market would have behaved in the trader's absence. (This is sometimes known as the Fundamental Problem of Experiments.) It's difficult to estimate how traders may have reacted to other strategies, particularly ones that may have substantively altered the public order book.
So let's come up with a "good price." Imagine we had a perfect trader — light-speed hands, access to all markets, the ability to do 10 things at once, and, most of all, faster Internet speeds than all other market participants. This trader would go through public order books and hit (lift) every bid (offer) they want until they reach the full amount requested by the user. (This process, known as "sweeping" the order book, almost never turns out the way it would for this theoretical trader, but let's go with it.)
Naturally, any sizable order will create some amount of market impact: by chewing through the available orders, and with no opportunity for liquidity to replenish itself, the price obtained would be potentially quite different from the top-of-book price before we started.
This price — the one we'd obtain by sweeping the order book at the time the trade is submitted — is the liquidity-adjusted arrival price. Quite simply, it tells you what you could've gotten with a really simple algorithm, a really attentive trader, or a really lucky market order.
In real life, sweeping the book won't actually get you this perfect price. You'll get something close but slightly different, simply due to changes in the order book before you have a chance to act. But, it shouldn't be hard to get this price (or a price close to it) with a fairly straightforward execution technique, like an inter-market sweep.
For large volume trades, if your execution price is worse than the LAAP, it might be worth considering tightening up the timeframes given to algorithms, or adopting a more passive trading posture. And, certainly, OTC quotes should never be accepted outside the bounds given by this number — this would be a literal arbitrage for the OTC desk.
We left out some critical details that are important when doing this analysis in practice.
First, we need to adjust for fees. The LAAP would incur taker-side fees, so these need to be factored into the price. Most execution techniques will attempt to stay passive in order to avoid fees, so it's important to create an apples-to-apples comparison by adjusting for fees appropriately.
Second, and far more complicated, is the question of risk-adjustment. There may be more risk in one execution technique over the other. If this is the case, a punitive factor can be applied to capture the value of risk mitigation. For example, the buy-immediately strategy is known to be the riskiest due to the possibility that a short-term liquidity shortfall create outsize price impact; for this reason, the LAAP is often adjusted by some value linear in the product of the asset's volatility and the square root of time.
If there's one execution quality metric you need to know, it's this one. If the simplest approach possible would perform better than what you're doing today, it's worth seriously reconsidering the current approach. You may be using a system (e.g. Routefire) that can provide LAAP statistics for you — if not, you can calculate it by keeping copies of the consolidated order book as it was at the start of each trade.
And it's well worth it. While this analysis can diagnose major breakdowns in execution quality, it can also be useful for ongoing refinement of algorithm parameters or internally developed strategies.