System Building Lessons continued
I keep getting distracted, but I’d like to move ahead quite rapidly now since Mole will be back on deck soon enough. Today’s post has enough information for you to go ahead and build working mean reversion trading systems right now, with not that much work involved.
I think we’ve covered in enough detail the idea that the best systems are based around a simple and clearly articulated idea, which can be tested for veracity.
The problem with complicated ideas is that our brains are built to believe all kinds of bullshit that isn’t true. Look at the recent election for a perfect example, we are a walking stew of cognitive biases that fuck with us relentlessly.
For an example of how I’ve personally been taken in by this kind of fraudulent bullshit let me examine the Top Down System. Ivan has a system with 13 different buy and sell setups, which happen on average every 1.6 bars. Overall they are nether a positive nor a negative edge, but there are significant times they outperform and they appear at almost every turning point, leading to significant confirmation bias. Ivan’s top down idea stacks a bunch of stupid ideas together to render them effectively untestable. When he gets his students to build systems he insists of ever more layers of complexity stacked up on top of the already complicated ball of wax. Which is why he’s broke, not to put too fine a point on it, and why I’m continually fielding emails from his bitter students asking me to put them on the right path.
The system works like this (there are a few variations but its the same concept) Look at the daily and 8 hour charts. When you are supposed to be in a long setup (according to his rules) you drop down to a lower timeframe like 15m, 60m, and 240m and search for setups in the long direction (if the system is long on a higher timeframe).
It sounds logical at first, and I admit I was taken in by it. But how to test the assumptions here? We have some critical assumptions
- That the system as a whole is an edge, so trading in the direction of “the system” is going to improve the edge
- That all timeframes are equal, ie that setups have equal edge and validity on any chart you want to pull up
- That if there is a discrepancy between timeframes the higher timeframe is the “dominant one”
- That all of the 13 setups are just as good as each other
- That all the setups are just as good long as short
- That one campaign management is appropriate for both rangebound and trending markets
- That you should exit a long position when a short appears
Taken together, it is quite literally impossible to test all these assumptions together. I proceeded to produce some impressive backtests, comprising a lot of work, that showed an expectancy of .2 or above in the short term.
Long story cut short, it didn’t work at all. It wasn’t an edge, and the fault was mine. I did not understand how fragility affected trading systems, and thought you could keep adding more and more rules ad infinitum.
The best systems are very simple. Simple ideas, and simple mechanisms.
Most of the systems you will build are variants on trend following, breakout, volatility breakout, momentum and mean reversion systems.
These are the low hanging fruit. When I see people want to get fancy with fancy ideas I see it as an affectation, like wearing a bowtie. Showing the world how smart you are, except you just look like a dick when you wear a bow tie. And you trade like a dick when you design systems with too many moving parts and too many untested assumptions.
Today I’d like to deep dive a little on mean reversion systems, and show you some of the classics, which have strong results, and provide you a cookbook for building your own systems. Stock and index futures mean reversion systems are dramatically easier than other kinds (which require real maths and not just the sort of “I can barely follow along” maths that I have)
Mean reversion systems are fundamentally different than trend following systems. Here’s where I have to eat a little humble pie.
I’d always been dismissive of those who traded without stops. Even when traders I deeply admire, like Victor Neiderhoffer, don’t believe in them, I thought that was why he blew up.
I was wrong. Ernest Chan explained it very well (and I’m running Ernest’s mean reversion systems in one of my own accounts to great effect) that if I am buying a breakout or a trend it makes logical sense to have a stop, because only a minority of trends ever pan out like we hope. However if we think mean reversion is probable with a 3 standard deviation move, logically it should be even more probably with a 4 standard deviation move, therefore a stop is illogical for a mean reversion system.
Trends and mean reversion are fundamentally different in every way, it stands to reason that stops are no different
In the real world we might choose to have stops in our systems, but we should do so knowing that virtually any mean reversion system we can build will be adversely affected by the presence of even a very wide stop.
So how do the professionals deal with this? They deal by trading for what you and I would probably regard as tiny, insignificant size.
For example this is Ernest Chan’s system on one my accounts for yesterday, on a roughly $100K account it took a few trades making or losing $20 a time. Hardly life changing stuff, is it? But the sharpe ratio is kept artificially high by taking enormous numbers of small trades.
And yet, the long term results are more than acceptable. This is the secret sauce of mean reversion systems. I’ve looked at a lot of them now, and they share the same basic ingredients.
- A metric fuckton of trades
- Very wide or no stops
- Tiny targets
- Win rates > 65%
- Very carefully chosen markets
If you are interested in learning more about these systems I hope you are a member. This isn’t a charity and professionals get paid. You can get plenty of free stuff from the dumbasses at that “bear blog”, the latest wave counts and pretty charts and so on.
It's not too late - learn how to consistently bank coin without news, drama, and all the misinformation. If you are interested in becoming a subscriber then don't waste time and sign up here. The Zero indicator service also offers access to all Gold posts, so you actually get double the bang for your buck.