Should a strategy backtested against three years of tick data continue to produce positive results?
Let's say we have a Binary Options 5-minute trading strategy that relies on multiple indicators and exploits price reversals in currency pairs. Now let's say there is a combination of inputs for the strategy's indicators that work really well together and produce a 65% average win rate when backtested against three years of minute-by-minute tick data.
In theory, can we expect this strategy continue to produce a 65% average win rate?
For one of the indicators, we use a [Polynomial Regression Channel](https://c.mql5.com/2/12/fig4_1.gif) of length 250. This tells us when the price spikes outside the recent average price range and thus offers a clue as to when the price will likely reverse.
In testing such a strategy, I have noticed that some days it performs extremely well (sometimes yielding a 75% win rate with 12/14 trades winning) whereas others it bombs (20% win rate with only 2/10 trades winning).
It's unlikely that backtested data will anticipate future results. The market tends to digest these lessons pretty well and as soon as patterns so simple to observe become observable the markets tend to arbitrage them quickly enough.
