Pathfinder Trading System
- This topic has 1,834 replies, 139 voices, and was last updated 9 months ago by
.
-
Oh dear…. it seems we have another case of backtest not matching reality. Is reality more profitable though – if so that could be a PRT first!
This backtest of v8 can never match reality completely. The first reason is that the spread is set to 2 points, which is incorrect for most of the day, and the second reason is that there is no correction for slippage.
Look at the exit of today’s 8 contracts : They were closed at 9:00 due to line 297 in the code, because posprofit (defined in line 287) was positive at this time and the maximum number of candles with profit exceeded (lines 51, 290 and 296).
When the spread is set to 2, posprofit at 9:00 according to the backtest was only 10,60 € for all 8 contracts. Now imagine, that due to slippage (entry prices a little worse than in the backtest), posprofit in reality may have been slightly negative, say for example, -1,30 €. Could have well occured and would have prevented the closure of the 8 contracts at 9:00, because maxCandlesLongWithoutProfit = 75.
When you want backtest results to match reality almost completely, you need to build in a correction for this and other types of slippage in the code. The backtest calculates with the candles’ close or open +- spread as entry and exit prices, and there must be a correction so that the live trading system does this too. Normally, a live trading system calculates with the actual entry or exit prices (not close and open) and they may be slightly different from the candles because of slippage.
Also, the spread should be adjusted to the actual values at different times of the day in the code, when you want a better match between backtest and reality.
And DAX 1h v8 is running 2 long positions.
1 user thanked author for this post.
Do you have an example of code that you add to your strategies to adjust the spread throughout the day and on different days of the weeks?
Here is a little code for correction of spread and slippage issues :
123456789101112// each buy or sellshort order occuring LATER in the code is labelled with "entered = 1"// the code remembers a theoretical "backtest" entry price 1 bar after a position has been openedIf entered = 1 and longonmarket thenposopen = open + 0.5endifIf entered = 1 and shortonmarket thenposopen = open - 0.5endifentered = 0I add this code at the top of a trading system, before all buy or sellshort orders occur. Then, I calculate the return of a position not with positionprice or positionperf, but with posopen, which is the theoretical positionprice from the backtest.
0.5 is half of the normal DAX spread from 9:00-17:30 h, and it should be set to different values at different times of the day. 0.5 points from 9:00-17:30, 2.5 points overnight, 1 point for the rest.
When you calculate the theoretical backtest return of a position, you need to add the real spread to the “close” values, too.
Return of a long position from 9:00-17:30 is then :
(close – 0.5) – posopen,
not “pprice * positionperf”
So when you say : “close a position when its return exceeds or falls below a certain value at the end of a bar”, backtest and reality will match more frequently when you apply this correction.
Keep in mind that anything that can cause differences between reality and backtest due to spread and slippage (for example the total return “strategyprofit”) may make the real system behave differently than the backtest. So you may want to include this correction for all commands that refer to actual entry and exit prices and to the return of a position or strategy. This may be quite some work.
1 user thanked author for this post.
Find exclusive trading pro-tools on 
