Strategy Back Testing Issue
-
Hi, I've been Testing a strategy recently which shows fantastic results however the results don't reflect the reality very well. This is because my strategy absolute time is between 7am - 10am however on the back tests at the time the selected horses sometime don't actually fit the criteria but then later on will do and the strategy will count these. Basically the software doesn't match up with the odds in the selected time frame.
Hopefully i'm missing something in the criteria to make my results true, would appreciate the help.Tom
-
@Tom-Williams We're releasing a "favourite price" setting for the betting rules this week or early next week which will allow you to do this. I'll publish this on the Changelog thread when its out.
-
@Tom-Williams The favourite setting in the betting rules does exactly what you're describing - tells you which horse was the favourite, 2nd fave etc at the exact bet time instead of at BSP.
You just need to use the "Favourites" section in the Betting Rules instead of the BSP favourite rule in the Horse Selection Rules.
-
@Adam Thanks alot for the reply, my biggest issue really is the timings as i'm trying to mainly go off the tissue price as i'm trying to see who's accuracy is more successful the bookies or the bettors.
Is there anyway of getting the timings to line up directly with the favorite at the time and the odds at the time? even if these odds change before the very start of the race and say the favorite at 8am is now 2nd favorite can the backtest still register these results?I understand the variables you must have to go through with all this so i do respect it, however is the backtest effectively only accurate for strategies that are based around the starting prices and figures given?
-
@Tom-Williams I was about to reply via PM to your support ticket but will reply here.
There are a couple of issues with your strategy I think.
You mentioned the runners issue, whereby the software is recording some trades in the backtesting based on the final number of runners and not the number of runners at the time of the bets.
We're not currently storing the number of runners at every timestamp, so unfortunately we can't create a query based on "number of runners at 7am", for example. This might be adding some phantom trades to your results to be fair, so I do understand the issue. We are going to add some granularity in future for these occasionally changing values like number of runners, going changes etc.
At the same time though, the number of runners at any given timestamp isn't actually going to be a meaningful rule in terms of the success of the strategy, like the final number of runners would for example, or to some extent the original number of declared runners.
We can easily add a "declared runners" rule, which will at least let you compare versions of the strategy based on original vs final number of runners. You could also maybe incorporate the non-runners rule to limit how much the number of runners has changed over the course of the market.
The bigger issue in the strategy is that you're using the BSP favourite strategy rule but betting way before the race. Obviously you can't see into the future, so if you change this for the "favourite" setting in the Betting Rules instead, you'll get much more accurate results. The difference is that the Betting Rules version picks whichever horse is the favourite at the time of the bet, not just at BSP.
The other minor thing, which is just organisational and doesn't have any effect on your results, is that you don't really need all the rule groups you've created in the builder, as the match type on all of them is "all". You could just have simple list of rules without the need for additional groups.
I've created a new version of the strategy in your strategy bunker for you (the "Adam edit" one) with those last couple of suggestions. Take a look