Inverting Stupid
Honda made a clever marketing campaign based on the old motorcycling insight that “Stupid hurts.” Like many keen observations, it’s applicable in many domains including trading and algorithm development. As a seasoned practitioner recently quipped to me: “The opposite of artificial intelligence is genuine stupidity.” If you’re trying to write smart systems, it may be worthwhile to spend some time reflecting on stupid.
I’ve recently had the opportunity to work with a brilliant mathematician. One of the first techniques he wanted to apply to our strategy development process is the mathematician’s old trick of inversion – instead of being clever and looking for smart algorithms, let’s see if we can happen across a dumb algorithm and invert it! This is a sound idea, but experience with algorithmic trading yields the unhappy result:
The opposite of stupid is not (necessarily) smart.
It turns out that this is true for at least two different reasons.
The first reason is that the primary form of “stupid” when one is considering trading strategies is a strategy that has no predictive ability. The simplest form of this is a strategy which bases trading decisions on a random condition. Such a strategy, run many times, will tend to create a return distribution which looks like a nice bell curve, centered around 0. Once one figures for fees, commissions and slippage, such a strategy will, in the long run, simply be a means for you to pay for the privilege to be involved in the markets. Contrary to an initial naive intuition, inverting this strategy will, sadly, not yield a fount of free money (unless you are a broker-dealer with a client who insists upon implementing it).
While a trading strategy based on some random condition sounds far-fetched, the great majority of trading strategies you can read about are actually equivalent to the random, non-predictive strategy. This is easily revealed by the returns of such strategies. They are non-predictive and their inverse is non-predictive. GIGO (Garbage in, garbage out) as applied to algorithmic trading.
But even strategies that are not wholly random are not necessarily invertible.
The second reason for our unhappy result is that the market itself is not necessarily symmetric. The market spread, price slippage and fees and commissions all conspire to potentially make the process of inverting a strategy a prohibitively expensive proposition. Until the uptick rule was recently abolished, US equities markets were explicitly asymmetric by law! Even without the rule, the atomic elements of strategies – orders – aren’t uniformly invertible; a limit on one side of the market isn’t going to behave exactly like a stop on the other. Likewise, the bid-ask spread need not be symmetric; one will frequently see the market trading on the bid while the ask is several ticks away. And once we trade in size or on illiquid markets, all of these effects are magnified.
The sad end result of this is that frequently a strategy which consistently loses money – when translated to its “inverse” – will consistently lose money.
The asymmetries displayed by markets and the strategies which operate on them shouldn’t prohibit creative thinking like that displayed by my mathematician friend. Instead, we should employ our creativity and our analytical tools, ensure that we understand and quantify our results … and always keep in mind that in markets as in motorcycles – stupid hurts!
