Survival Approach to Developing Trading Systems

Leveraging Survival Techniques to Optimize Trading Strategies

Introduction

I’ve been very intrigued over the past year in the idea that markets are a game of survival. The longer your portfolio survives in the market, the longer you are exposed to regimes where your trading really shines. Let’s take the example of 2020-2021. That was a massive money making market if the trader was long. The only way to take advantage of such market was to naturally be trading it. If you blown up your portfolio prior to that year, you did not take advantage of that market regime.

So the first task of the trader is survival. But what does survival mean? Survival means that your edge can be exploited for large periods of time without blowing up the portfolio. In order to achieve that you need both a robust edge and also a strong risk management approach. So let’s think through these concepts with that survival aspect behind them.

What is a robust edge?

This is what a robust edge means to me:

• It’s not easily destroyed by minor market dynamic change

• It can survive across different regimes (bullish, bearish, trend, mean reversion, etc)

• Has a high degree of frequency

• It survived over many years out of sample

• Can survive basic stress-testing

These robust edges are the ones that survive the longest. You can have such system active for many months/years without it becoming obsolete. A robust edge is not a system that gets easily destroyed when under stress.

This doesn’t mean that an edge that will only be available for a short period of time is wrong. There are unique edges that become available due to inefficiencies with protocols, projects, pricing of derivatives, etc, that usually only last for a limited amount of time. In this article I am not talking about such edges. Rather I am speaking about the ones that can last for larger periods of time, that might not be as optimal as more “shiny” strategies.

These robust strategies are usually not the strategies that are optimal in a return perspective. In order to endure harsh market dynamics that are constantly changing, the strategy needs to be also flexible enough to endure these changes. In order to endure change, it needs flexibility. This flexibility doesn’t allow for the edge to be extremely specialized in a market regime, which in itself, makes the overall performance, sub-optimal.

On the picture above it’s actually one of these sub-optimal systems. These are the assumptions:

• Asset: BTC

• Timeframe: 6h

• Risk per trade (of equity): 1%

• Period: 2011-2023

• Number of trades: 534

I haven’t included more specific performance because that’s not the point. The point is that it isn’t a shiny equity curve. It does produce a profit, and it was built with all the robustness techniques I normally use, but it’s not optimal. You can also see that from 2020-2023 it barely produced a profit, which is concerning considering that that time period had one of the best performing years for the asset class.

Evolutionary and Survival Approach

When you are building a trading system, you start with an hypothesis. You have a basic theory of how to exploit some sort of edge from the market. This theory is usually based on a few criteria. This criteria can be based on basic indicators, price, order book data, fundamentals, advanced mathematic models, prediction models, etc.

The point is that your initial theory will not be perfect the first time you test it. And you will have to tinker with the criteria in order to achieve a higher degree of performance that is acceptable for deployment. Different traders have different goals when it comes to system performance:

• Returns

• Win Rate

• Drawdown

• Sharpe

• Average win to loss ratio

• Etc

In order to achieve the closest to best strategy according to your desired outcome, you must keep evolving the strategy with better criteria. This happens over large iterations of testing of different inputs to see which works best.

Example:

• Buy when Daily is above 200SMA and 6h has a deviation of 2 ATR from the 20SMA.

• Buy when Daily is above 500SMA and 6h has a deviation of 3.5 ATR from the 20SMA.

• Buy when Weekly is above 20EMA and 4h is above the high band of the bollinger bands with a length of 30 and a standard deviation of 3.

You get the point.

The trader starts working around different criteria and different lengths in order to achieve the best outcome for the data set. But how to get there faster?

One is way is to use evolutionary thinking. Charles Darwin said:

“It is not the strongest of the species that survives, nor the most intelligent that survives. It is the one that is most adaptable to change.”

You can do that by start experimenting with the following idea. If evolution is the process of finding the individual that is most adaptable to the environment, how does that translate to trading? How can you use that theory for finding trading systems?

Here’s an example:

• Test 100 strategies.

• Rank them from highest to lowest based on your desired outcome (fitness)

• Combine the criteria of top 25% strategies

• Rank them again

• Delete the bottom 75% of strategies

• Add 75 new strategies to the mix

• Repeat the process until you are happy with the outcome

Make sure that you are not running into over-optimization problems and use basic measures of robustness. But that is not the topic of this article. We can get into robustness testing in the future.

The point of this article is to show you how you can build a trading system using a commonly accepted scientific method of survival and adaptation. You need the fittest strategies to survive and keep evolving with the other fittest strategies. Keep doing this until you find a strategy that you can work with and then expand on that.

Hope you have enjoyed the article!