Saturday, February 17, 2018

Some of my beliefs about backtesting

When it comes to trading, I like to think differently. I'm not afraid of being a bit of a heretic when it comes to thinking about the markets. While I adopt and use 'classical' trend following and breakout principles, I also like to think for myself, outside the box.

As an example, lots of traders swear by running backtests and tuning their trading models based on the results of their testing and analysis. Certainly, in some cases with getting to grips with a basic concept, backtesting can be of some use.

However, you need to be aware of the dangers of fine-tuning your backtested results too much. Complexity for complexity's sake is worthless, and prone to fragility from being overly 'curve fit' or dependent on past market action and conditions remaining the same.

As an example, someone who fine-tuned a short-term method based on the low volatility environment we had last year would have had a nasty surprise over the last couple of weeks...

Trend following on stocks is a marmite approach to trading, depending on who you listen to. Some traders, authors (and fund managers) swear it doesn't work. Other traders and CTA's use it just fine. Based on my own experiences and results, it should be obvious in which camp I fall into.

So, did I spend hours testing and refining my method?

In a word, no.

When I came up with my basic trend following method, I NEVER did any backtesting.

I can imagine now what some of you are saying: "WHAT - are you crazy???"

Let me explain my rationale.

The basic concept of trend following has been used successfully over decades. I simply took two simple, well known trend following methods and parameters, combined them, and tailored them by adding in some additional elements and rules as my own method was to be used on trading stocks.

I sure as hell wasn't going to waste time re-inventing the wheel. I simply got stuck in, to see if it worked, and more importantly to see if I could follow the rules.

It is a simple method. It is robust. And the typical performance characteristics achieved are similar to those used by well-known CTA's and other trend followers when trading commodities, forex, indices and the like.

As one example, Richard Dennis once said that 95% of his profits came from 5% of his trades. Last time I calculated the same based on my own results, 94% of the profits came from 6% of the trades. The win rate also falls into line with what has historically been achieved by trend followers.

So, given the metrics achieved by the trading legends who have adopted such an approach over time, I am in the right ball park.

If with all their computing and brain power I have been able to achieve roughly the same performance (albeit on a MUCH smaller level of equity), by adopting their basic principles and rules, then I am happy. Who am I to say I can outperform them? I just utilised their knowledge and experiences, and made it work for me.

I also know that, any time I have tried to test things and improve it, performance has dropped off. So I have dumped the refinements and the added layers of complexity. What worked a decade ago is still working now. So now I have the sense to let it flow.

There is a far more important point here though about backtesting in general.

If your basic beliefs about the markets are like mine, then you might even see backtesting as a complete waste of time.

All backtesting does is help you develop a wonderful system based on the past market action and conditions. As with all trading methods, past performance is not indicative of future results.

As I have talked about on here and on social media repeatedly, the future doesn't exist. Also, every moment in the markets is unique. Therefore, you do not know what market conditions and price action will be like going forward.

You don't know what the level of volatility will be. You don't know if price will be stuck in a range or in a trending phase in the markets you trade. Therefore, all that work backtesting, refining and honing a method based on the past may count for nothing.

It has also been proven by others that there is a positive expectancy based on random entry into markets combined with the basic rules of cutting losing trades, letting profitable positions run, and using good risk control. Refer to my post on The Golden Rule for more.

All this will no doubt generate resentment from those authors and traders who are adamant that trend following doesn't work on stocks, or those who backtest extensively and make continual refinements to their own trading approach. Well, my own results demonstrate that adopting simplicity and robustness works for me.

There are people out there who are stuck in a endless loop of refining a theoretical method based on the past, without ever getting their hands dirty and actually interacting with the markets. Then, when they finally do, they find out that the market has changed, and moved the goalposts. To put it another way, they are always chasing the market's tail.

I'd far rather people formulate some basic beliefs about how they want to trade, set out the basic rules which will govern their approach, and get stuck into the market using a small level of capital and very strong risk control. At least that way, they also get to experience the real-time emotions and psychology that comes with having money at risk (which is a crucial element of successful trading that backtesting cannot compensate for). That way, they will learn what works for them.

The whole issue of backtesting and trying to improve your existing method was brought home again during a recent conversation with a successful trader, when he threw in the following question. This revolved around whether he could (or should) make changes to his existing method in an attempt to improve performance. It was put to me along these lines:

"Say my method has been achieving average returns of 20%. Should I make changes in an attempt to improve to 40%, but which may result in a deterioration so that returns drop to 10%?"

The honest and only answer I could give was that it might work, or it might not. Because the big 'known unknown' in all of this is what future price action and the market conditions will be - which we do not know or cannot predict with any certainty.

By the same token, carrying on doing the same as he is now, he may make the 40% this year. Or maybe 100% or more. Or maybe even a loss. Who knows?

What we DO know as trend followers is the basic premise that, for us to be able to make money, prices need to trend. When they are not trending, then we tend to struggle. That is a given. We don't know when trading phases will start or end in a particular market or sector. That is what diversification is for.

That has always been the way in the markets. You can go back decades, to the days of Jesse Livermore or Bernard Baruch, or even further back to the likes of Dickson Watts, before computers and complicated technical analysis or backtesting were even dreamt about, and those basic truths have remained unchanged.

In my own case, fully accepting that you are unable to predict the future, and that every moment in the market is unique, is all thanks to the writings of the late Mark Douglas. Truly understanding these concepts, and sticking to a simple, robust strategy has, in my own case, proven to be extremely liberating.

I think what is irrefutable is that trend following, as a concept, worked in the past and still works now, on any market where there is buying and selling, in spite of what some may say.

And you don't need a backtest to prove that.

1 comment:

  1. Very good article, thank you. I like your style and attitude.

    ReplyDelete