Backtest uses 2 x Spread??

• 04/04/2020 at 9:41 AM #124590
  GraHal

  I hope this is me missing something simple , I really do??

  My System is onmarket 96.24% of the time (Buy / SellShort reversal).

  When I backtest my System …

   1.  With Spread = 0, I get 100 Trades at an Average Gain per Trade = £24.45

  2.   With Spread = 10, I get 100 Trades at an Average Gain per Trade = £4.45

  Seems too much of a coincidence other than spread is being accounted for twice by the backtest engine??

  If you try it (please would you) then it would need to be a Buy / Sell Short Reversal System.

  

  GraHal

  Participant

  Topics: 214

  Replies: 8846

  Been thanked: 2211 times
  04/04/2020 at 10:26 AM #124593
  Vonasi

  The way to test this would be to have two strategies. One that simply buys at a precise time and one that goes short at the exact same time then run them as back tests. As long as there is no slippage then the difference in price between what one strategy bought at and what the other strategy went short at is your spread.

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1811 times
  04/04/2020 at 11:08 AM #124597
  GraHal

  Yeah but there could be something weird going on (re spread) where the direction reverses?

  I guess I only noticed it as it was 100 trades and £10 spread and so the maths leapt off the screen at me?

  Previously I may have put it down to a bad performing System.

  I’ll knock something up later so I am testing my theory on at least 2 Systems.

  Thanks

  

  GraHal

  Participant

  Topics: 214

  Replies: 8846

  Been thanked: 2211 times
  04/04/2020 at 12:03 PM #124599
  AutoStrategist

  Hi Grahal, out of curiosity I tried this by knocking up a random system that does a long/short reversal and didn’t get the result you are seeing.  Results with spread of zero at 1 per contract were £6.71 per trade and results after with spread of 5 were £1.71.  If you look in the orders list you will be able to see the prices that the trades were taken at and see the price difference between 2 trades and compare that with the results for the other spread setting.  For my test it confirmed the spread was the expected 5.

  1 user thanked author for this post.

  GraHal

  

  AutoStrategist

  Participant

  Topics: 6

  Replies: 125

  Been thanked: 71 times
  04/04/2020 at 12:11 PM #124601
  Vonasi

  GraHal  perhaps with such a large spread of 10 you just happened to have a lot more losing trades that just happened to add up to £20?

  You could just create a dummy strategy that keeps reversing its position every minute and then a second dummy strategy that reverses its position in the opposite direction at the exact same time and then back test them and see what prices they trade at.

  1 user thanked author for this post.

  GraHal

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1811 times
  04/04/2020 at 1:02 PM #124608
  GraHal

  AutoStrategist wrote:

  see the price difference between 2 trades

  Good Call!

  With a spread of 10 … 5 is accounted for at entry and 5 at exit.

  Vonasi wrote:

  you just happened to have a lot more losing trades that just happened to add up to £20?

  Yep it was just that … pure coincidence and the simplicity of the maths lead me to a flawed conclusion!

  Thank You to Both

  Just because we Believe something … it does not mean it is True 

   

  

  GraHal

  Participant

  Topics: 214

  Replies: 8846

  Been thanked: 2211 times
  04/04/2020 at 1:09 PM #124611
  Vonasi

  Case solved and closed. Topic marked as resolved!

  1 user thanked author for this post.

  GraHal

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1811 times
• Author
  Posts