The BTC Horse Racing Thread
-
@Dan-MacKinnon Yeh same. To be fair the beaten fav strat was absolute flying for a good 3-4 weeks so probably due a dip this week. Hope it picks up again shortly..
-
I've had an absolute 'mare these past couple of days with the horses (pun intended!)
But if being here has taught me anything, a good strategy isn't about avoiding all losers, a good strategy gets you back to square 1 asap.
-
@Joseph-Henderson I think you mentioned before the concept of using say one year of data (say 2022) for the initial strategy creation and then testing it against 2023, it’s definitely helped me with a new strategy I’m working on that is looking very promising so far. So thanks for highlighting this as a method.
Want to give my new strategy a few more weeks of forward testing and then will post here.
-
@Joseph-Henderson most of the member ones are from mid last year
-
That is useful thanks. Think I need to spend some time digging into the data.
Do the strategies say anywhere what date they were created? I know when you upload one to the software it says the date you added it, but I am talking about when originally created so it’s clear what historical data it was presumably based on?
-
@Joseph-Henderson
There is a concept of back fitting when using software like this. The software detects profitable patterns in the historical data. You then think thats wonderful and apply that to live betting. Sometimes the pattern is a genuine fundamental edge, sometimes its just a pattern that fits that particular range of data. The givaway that its the latter is that the pattern instantly fails when used on data not in its sample. Check the date created on back the fav and make up your own mind.The ways to avoid losing your shirt with that problem is to give it a few months after creation to see if the pattern holds before putting money on it or (when gernerating these yourself) dont use the full dataset to back test on. Develop your strategy on for example just 2022 data. When you are done then add 2023 and 2021 (there is some debate about how valid 2021 might be because of the hangover from covid but in any case i would prefer to see a parrter persist on recent rather than previous) and run the backtest again. If 2023 looks as good as the trained period (2022) then you very likely have a good one.
-
Morning All. Just wondered if anyone had any advice about analysing historical data and when to consider stopping/starting to follow a strategy. For example, the Back The Fav at the off (v3) strategy looks great on the all time data, but seems to have pretty much flat lined since December 22. Or am I over thinking it and this is just down to variance? Any thoughts/views would be appreciated. Thanks
-
yes it can and thats one of the things i was thinking about. It will always match at he touch point so for most cases i would still be getting the same price. Its strange how its really causing me problems this month in particular though. ~Will give it a go over the next week
-
Three winners on the bounce for the pace backs yesterday. Nice to see these performing as the flat gets going.
-
@Jonathan-Jones can fairbot ask for a worse price by a couple of ticks? If so do that and as long as you have Best Price Execution enabled in Betfair you should be matched everytime with one or two in a blue moon outside your odds range.
-
Wondering if anyone has done any analysis on the effects of not getting matched.
This month has been a pretty frightning wake up call to be honest. Its a totally crap month so far (unusually so) so i did a long and painful deep dive into my logs. I have in live lost 22.38 points this month (so month total current -22.38 and DD is far beyond the backtested limits). The software confirms its a bad start to the month but has only recorded -10.17 points (thats half as much and well within the backtested limits). Got me very curious as to what the hell was going on.After several hours i can see that the rules did execute for all the missing horses (accross all the strategies, including a very nice pace back that went on to win). The problem is the bets sat unmatched (despite being for less than £1 and being placed at best lay/back to maximise chance of getting taken).
I didnt check in April because like many people (probably) i only really start questioning and spending sill time on analysis when things dont go to plan (and April went very much according to plan).
Wondering if anyome else has noticed this and what they did to mitigate it. I can think of a possible mitigation but the way Fairbot automation works thats not going to work with running a portfolio. I requested a change from Bintenko a while ago (which they agreed to implement in the next version) which might (depending on how comprehensively they implement it) solve the problem but at the moment i am stuck.
-
Looks like things are getting back on track after an up and down winter. This game can try the patience of a saint. Which I am not.
-
@dan-mackinnon said in The BTC Horse Racing Thread:
@john-folan said in The BTC Horse Racing Thread:
Race Type
Handicap Y/N
Class
ClassificationThese are where I normally start. Especially the first one.
That's similar to my go-to list. Would you split those down further, for example Back the Fav in AW or hurdle only vs all race types that are profitable?
I’d go as far as to look at each one Individually. Mostly because after you have done that other filters have more relevance in some types than other(pace as an example).
Thank the Lord for the save as new version button!
-
@jonathan-jones said in The BTC Horse Racing Thread:
I have for Back C&D Winner. I have 3 seperate strategies (Chase, AW, Turf) and also for pac backs (only done on turf). I am sure i tried to do this on the lays but found they didnt pass out of sample testing or some reason like that for not keeping them.
Race type is the biggest one I think. You’ve deffo done the right thing there.
-
@jonathan-jones said in The BTC Horse Racing Thread:
@john-folan
Thing is.. Its already accounted for when you run the backtests and create the strategies. I admit it sounds completely reasonable but you would be acting on a variable with no empirical evidence in opposition to your backtest data.It is but imagine if you could take those days out. It would be interesting to see.
-
@john-folan
Thing is.. Its already accounted for when you run the backtests and create the strategies. I admit it sounds completely reasonable but you would be acting on a variable with no empirical evidence in opposition to your backtest data. -
I have for Back C&D Winner. I have 3 seperate strategies (Chase, AW, Turf) and also for pac backs (only done on turf). I am sure i tried to do this on the lays but found they didnt pass out of sample testing or some reason like that for not keeping them.
