Does history tends to repeat itself?

history repeats itself

This is the a basic principle of technical analysis: history tends to repeat itself. On the other hand, forex market continuously change, its dynamic can’t be predicted because it is definitely not a deterministic system (a system in which no randomness is involved). How can both statements be true at the same time? Every dynamic system has a minimum and a maximum. When we optimize an EA for a specific period of time, we tend to curve fit it which means to choose its maximum (maximum profits for the smallest possible drawdown). But the market conditions we optimize (read curve fit) our EA for will ever return? This is an interesting question to answer. What I’m going to do is to curve fit a test EA for a small 2 years period. From 2007 to 2009 then see how the best performing parameters behave on unseen data. This is a good way to spot when the same market conditions occurred over time. The results of this test can give me an answer to another question: can we really make money by trading using a single forex robot only?

 

Here is one of my EAs curve fitted for 2007-2009 period.

2007-2009-curve-fit

2007-2009-curve-fit

Profit: 8462 pips
Drawdown: 851 pips
Yearly average pips: 2820
Average profit / maximum drawdown: 2820 / 851 = 3.31

Here is how backtests look like if we keep the same parameters for a much longer period 2000-2013:

2000-2013-backtest

2000-2013-backtest

We can clearly see that FOR THIS STRATEGY which is a trend follower strategy the history tends to repeat itself. It performed pretty well during 2000-2007 using parameters from 2007-2009 only! The market changed since 2010 until 2011 and kinda invalidated the strategy (it survived in drawdown for 2 years, it didn’t blow the account but it didn’t make any profits either). Then, the same conditions popped up since the beginning of 2013 until now.

Conclusions

1. The history tends to repeat itself for valid strategies only! The strategy employed by this robot is based on candlestick patterns, no indicators or any form of curve fitting is used in the strategy itself. Entry and exit rules are crystal clear and very simple, everything is based on candlestick patterns.

2. Even if 1. holds true, history tends to repeat itself here and there, during small  periods of time for every strategy. But there is a  big difference between genuine strategies and curve fitted ones. For strategies that involve curve fitting (many rules, indicators, hardcoded data, etc) the overall equity curve will always point down because history tends to repeat itself only for very small periods of time, the rest of the time passes against them.

3. The method above is one of the best methods I know to test the validity and longevity of a strategy: curve fit it for 2 or 3 years then test the best performing parameters against a bigger period of time: 10 years for example. If the market goes in our favor most of the time and the equity curve points up then it worth to optimize the strategy (forex robot) for the whole period of time (13 years). Of course past performance is not a guarantee for future performance but at least we know that there is a strong correlation between the strategy and market movement.

4. Ok, fine, but what about those 2 years of drawdown? How do we know it won’t happen tomorrow? We don’t. Trading a single EA is not enough to make profits every year. We need another profitable forex robot to cover the gaps in profit without ruining the profitability of the first one. And this means a good portfolio. I will write in another article about  how to build a good portfolio and what rules should be followed. In  the meantime, please read my previous article about this subject. It does not cover every aspect but at least it prepares you for what comes next 🙂

 

Thanks for reading!