Last Updated on 11 September, 2023 by Samuelsson
This article is an extension of our article on curve fitting. If you have not read it, we recommend you do so here.
Less is more is a well-known adage that is very relevant to and helpful in trading. In short, it means that you always should try to keep your strategies as simple as possible, to prevent curve fitting.
Beginning traders often hold the view that a trading strategy should be a complex and advanced logic with many filters and conditions. The more advanced, the better, they believe. However, they could not be more wrong! Those understanding the concept of less is more in trading, will most times produce worse results in their backtest. But do not worry. Instead, your results going forward will be much better!
One Golden Rule
In trading there is one rule that every trader should remember:
The more advanced an edge, the higher the chance it being curve fit.
When we design edges and strategies, we want them to be as simple as possible. Many times, one line of code for the entry and one for the exit could be enough to have a strategy with fantastic performance not only in the back test, but also going forward. Edges and strategies with many conditions and filters are more prone to being curve fit.
But why is this?
Well, every time you add a condition or rule to your strategy, one thing happens:
You focus your strategy to utilize a more specific behavior than before you added a filter.
In other words, each time a new filter or condition is added, trades are removed from the backtest. What this means is that we define what part of our strategy we want to keep, and which we want to throw away. This is done under the assumption that we can derive what works and what does not work from backtesting. In our article on curve fitting, we established that this is not always the case and that what we observe could be nothing else than random market noise. If every condition in a strategy can be curve fit, that should mean that each time we add one more condition the probability of the edge being curve fit increases.
To Clarify Further
Let us try to clarify this with an example:
Assume that every condition added to a strategy has a 30% chance of being curve fit and a 70% chance of being true market behavior. That means that for every added condition, the chance of the edge being curve fit increases by 30%.
Now see what happens for every condition we add to our logic:
|Number of Conditions||alculation||Likelihood of the edge working in live trading|
And it goes on…
The best chances of our edge working going forward we have with one condition. For every added condition, that chance decreases by 30 %. Once we are up to 5 conditions, there is only a 16% chance of it not being curve fit. That means that in over 8 of ten cases, what you have built will fall apart in live trading. Not very good, right?
While this example is just an experiment under hypothetical circumstances, it nicely illustrates the dangers of having too many conditions in a strategy. As always, reality is more nuanced than any model that is used to describe it. Some edges with many conditions may work, and some with only one condition may fail.
Since we initially cannot know if what we backtest holds any true merit or not, the more we rely on market data to generate trading signals, the higher the chance that curve fitting appears in some of the conditions that comprise our edge. Due to that, we want to limit our reliance on market data as much as possible, which is most easily achieved by having as simple edges as is possible. That is why the adage “less is more” is especially helpful to traders.
Here you can read more about algotrading in our archives.