Why difference in two back testing

Dear @Streak
Kindly some info on this.

Both have I think same entry condition.
One with Switch to Advance.
Thank You with best regards…

Hi @Dhaval_Joshi,

Great that you asked this :slight_smile:
This one of the sutle but good differences b/w create and create plus.
Let me explain the reason why the results are slight y different.

  1. First of all, create plus has all the exactly same trades as that of basic create, plus few more, thats why you see 108 and 100 respectively.
  2. For the common trades everything is exactly the same, like entry time, price, trade type, etc, you can easily check that in the trade log.

So the final pnl has difference because of the extra trades.
So why the extra trades in create plus.
Answer : When you backtest say from dates D1 to D2 in create, then in backtesting it only uses the candles from D1 to D2. Now some of the early candles are of no use as they are used to calculate the indicators, and no trades can be calculated on these even though the condition might be true, hence some early trades near D1 are not accounted in the backtest, till all the indicator values have been calculated.

BUT in create plus, because we implemented a much faster and powerful version of our backtesting, we not only fetch candles between D1 to D2, but we also fetch, candles before D1(which not being done in basic create) such that all the indicator values are present from the first candle of D1 till D2. This make sure no trades are discarded during the period D1 to D2 and hence few extra trade depending on the conditions.

Hope this helps you see the working under the hood :slight_smile:

1 Like

Hi, @Streak

Thank you for your prompt and easy layman term reply… :ok_hand:

But as it seems “create plus/Basic” is more reliable, accurate and efficient, than why don’t you completely switch over to it…
Means its quite risky when pnl difference is more than 10-15% between them.

Thanks again for time…

Hi @Dhaval_Joshi,

Thats not correct prespective of look at it.
Both have same accuracy. The difference is , advance runs backtest on the full date range selected by the user, but basic create runs backtest on the same date range selected by the user minus few days/candles depending on the strategy.
Difference b/w % returns is redundant, when the final returns is very small. Eg: for a small number is even deviation of 0.1 point, the difference in terms of % can look big.
The correct way to analyse backtests should be with regards to win/loss ratios, streaks, and other backtest metrics. Which strategy imporves the win/loss ratios, risk rewards , reducing draw down is what is of importance, where as absolute pnl and returns are for quick filtering of good strategies from a top level.

Hope this helps.

1 Like