Why streak gives different result for individual 3 month back test vs 3 month combine back test
What will the equation to compare both?
Thank you all for your time…
Yes that True and the reasoning is quite straight forward.
Lets take an example to understand easily!
Say strategy is simple MA crossover : Buy when SMA(9) crosses above SMA(21) and exit on TP 1% and SL 1%.
Now in the Standard Create, when you run backtest from date D1 to date D2, the backtesting service only fetches the candles from dates D1 to D2.
As you can see, to calculated the SMA(9) candles are required, and to calculate SMA(21) min 21 candles are required.
So basically no trades happens for first 21 candles the SMA value cannot be calculated yet(similar thing you can observer in chart when you plot 200 SMA and scroll back fast).
So when you run for 3months together, after the first 21 candles in this example, the trades are captured and backtested for and you get the results.
Now lets say you break the 3months into individual months and backtest, then for each of the months, after the first 21 candles only the trades are generated and captured to generate the backtest results.
So as you can see the 3month continous backtest only has misses the first 21 candles, but 3 individual backtest by the basic candles necessity, miss 21 candles each months, causing in few trades to differ, and hence causing the find backtest results to slighty differ.
In Create plus, which was recently released, this has been avoided, by alway fetching more candles than what users runs the backtest for, to make sure even the earliest trades are captured.
There is no as-such direct equation to compare the both, you can just add the pnl at the end of each month, to arrive at a net pnl/return.
What you actually should look for is win/loss counts and avg win/loss amounts in each month to check how good the strategy was doing in each specific month vs over the 3months.
Hope this helps.