How much market data do you need for a Forex strategy back-testing? What time interval is long enough to run tests?
One obvious consideration is that it depends on the time frame you use for trading. You need to have many data points for your selected interval, to be able to make any judgement.
The longer the time frame is, the more data you will need. E.g. for M15, one hour of ticks will only give you 4 bars. For H4, you will need to have almost one day of ticks to get the same amount.
Another consideration is, how long are you going to be in the market. This depends on how you trade – do you use leverage or not. If you use leverage, it is very risky to keep a position for long in certain cases – e.g. over a weekend, as there can be a big gap at the opening. To say nothing of news release times.
These considerations should give us some idea on the minimum and maximum data interval we need to test on.
But let’s assume we have all needed tick data available and don’t have any restrictions coming from leverage. And we just want to understand, how much data do we need to test on, exactly? When should we stop testing?
The answer to this question is – it depends on the trading strategy. For the strategies that doesn’t adapt to the market (“static”), no testing can give any guarantee of success.
You can optimize the strategy parameters on 3 years data for several pairs – and it still can start losing the next day.
This is why we are working on the dynamic Forex strategy optimization. This approach requires much less test data, as well.
Basically, for the dynamically adapting strategy, you can move from “extensive” testing on long intervals, to “intensive” one – using price patterns as “test cases“. These test cases should simulate the market conditions where trading strategies were noticed to consistently lose.
This approach also saves a lot of time, as we can only concentrate on the essential behavior of the strategies.
Smart Forex Tester will support running of test batches in summer 2017.