The BTC Horse Racing Thread
-
Yesterday i made a small change to import the min/max prices. I didnt remove the selection condition with the prices (I thought it shouldnt do any harm). But,
Its still not matching up. BFBM has too many bets.Example. 16:15 Sandown.
The backtest says it should have the following bets:
08/07/2023 16:15:00 0.24 -5.60 -1.34 Sandown Lord Protector JustLay
08/07/2023 16:15:00 0.24 1 0.24 Sandown Aramaic JustLay
08/07/2023 16:15:00 0.24 1 0.24 Sandown Paradias JustLayIt actually took these plus
Majestic (got odds of 6.2)
Mustazeed (got odds of 7.8)Now those do sound reasonable. The odds are in the range. But, The backtest doesnt have them and my Fairbot Live doesnt have them (BFBM is running in sim mode so not sure if thats the problem).
The backtest tick chart says Majestic never traded at 6 or above till 16:17 (the tick chart doesnt show seconds) which is a full 2 mins after the official start.
Also Mustazeed never traded as low as 10 at all according to the backtest charts.Is it possible the BFBM differences are down to Sim Mode?
I cant take credit for that strategy. Soemone else posted it here a while back. I picked it up, thought it was interesting and added it to my watch list. Its had a couple of good live months now (Live performance is matching its performance when it was created) so i have added it to my running list.
-
Can I also suggest that you all change your refresh rate to 300ms as I have done here TOOLSANDSETTINGS>Monitoring Settings. Its the fastest BFBM can go. Ignore my start monitoring thing. It all depends on when your bot needs to start monitoring a market for your purposes. That's for mine.
-
@Joshua-Varley said in The BTC Horse Racing Thread:
@Jonathan-Jones how did the bot go yesterday? The one I made yesterday morning to test the strategy you posted, fired 41 times, with 36 winners, all in the odds range specified. I checked against the results this morning and the bot only missed firing on 4 selections, all of which were at the extreme high end of the odds range.
Did your results match up?
That sounds more promising.
-
@Jonathan-Jones said in The BTC Horse Racing Thread:
For today i have imported the min and max in the tipster service. I have left that selection criteria where it is but have unchecked the box in the staking to ignore tipster min/max. Hopefully they will work together.
Anyway. These are the files i am trying to get to work. The strat is one someone poster here. I watched it live for a couple of months and it seems to work very well (though has no real reason why). I can replicate the backtest results 100% in Fairbot but BFBM is miles off.
See if there is anything i have missed.
I cannot see why you are getting such different results. I run two very similar bots in that it takes selections from the software and then filters according to price only like yours.(wider ranges but still should be the same. The literal only differences I have are as I mentioned I do not have a stop placing bets rule(but that should result in less bets not more), my ticks to get matched are 2 rather than 4. That might make a difference. The only other thing I can think of is the software has a minimum amount matched at a price point. Stumped otherwise. How did it do yesterday?
Nice strategy by the way
-
Just downloading the backtest data to compare but in BFBM it did 43 bets. Its not possible to compare by strategy with Fairbot unfortunatly. Only way i can do that is to compare Fairbot to the aggregated race totals then investigate any differences
-
@Martin
Betfair does have status updates so i guess you could use those but they dont seem accurate enough. see a few races bouncing between ATTHEPOST and GOINGBEHIND etc. If you were scalping a horse market with big money would you trust them or would you be watching the horses get loaded to decide when to get out? -
@Jonathan-Jones how did the bot go yesterday? The one I made yesterday morning to test the strategy you posted, fired 41 times, with 36 winners, all in the odds range specified. I checked against the results this morning and the bot only missed firing on 4 selections, all of which were at the extreme high end of the odds range.
Did your results match up?
-
@Martin
If you are watching the race then you could make a pretty accurate guess i think. If your criteria is something like last 10 seconds or so, watching the horses get loaded etc then no issues.I am running several strats so even if i was able to watch the races to manually execute i doubt i would be able to keep up with all the different rules for the different horses.
-
@Jonathan-Jones said in The BTC Horse Racing Thread:
For today i have imported the min and max in the tipster service. I have left that selection criteria where it is but have unchecked the box in the staking to ignore tipster min/max. Hopefully they will work together.
Anyway. These are the files i am trying to get to work. The strat is one someone poster here. I watched it live for a couple of months and it seems to work very well (though has no real reason why). I can replicate the backtest results 100% in Fairbot but BFBM is miles off.
See if there is anything i have missed.
Thanks for sharing!
-
@Daniel-Mills
Its not LTP. I tried the default on yesterday and it didnt help. Exactly the same issue.
-
@John-Folan said in The BTC Horse Racing Thread:
@Jonathan-Jones said in The BTC Horse Racing Thread:
I think your problem is using LTP. I just use back and lay price(both, not either, or) and my stuff seems to match the software 98% of the time.
daft question, where are you setting this? is it this?
-
For today i have imported the min and max in the tipster service. I have left that selection criteria where it is but have unchecked the box in the staking to ignore tipster min/max. Hopefully they will work together.
Anyway. These are the files i am trying to get to work. The strat is one someone poster here. I watched it live for a couple of months and it seems to work very well (though has no real reason why). I can replicate the backtest results 100% in Fairbot but BFBM is miles off.
See if there is anything i have missed.
-
That didint work. Exactly the same issue. I have more bets than i should.
Going to push the min max odds in with the tipster service and see if thats any better.
Will upload both the strat and the BFBM file in a bit
-
@Jonathan-Jones that makes sense. Let me know how you get on. Particularly with the price change. Good point on the pre race thing. I usually find I get the odd extra bet and that explains it. I think it will be impossible to get an exact match so as long as I am close I am good. Especially if a strike rate is matching on the strategy.
-
If you set up a strategy in the software to run relative to 'Official' start then the software will stop placing bets at the official start time. Unlike Fairbot (and probably most other trading software) this software cant specify a different value (Official/Actual) for start and stop (@Adam maybe thats a possibility for the future).
As all my strats are set up like this my BFBM setup i think is correct to maintain consistency between actual results and backtest results. Either losing the stop betting from BFBM (thus creating a betting window of unknown size) or changing the software to use 'Actual' (creating an impossible to implement criteria) means i am introducing a set of circumstances that cant be modelled (like the favourites issue) and a total lack of trust in the backtest results.
I have changed the price criteria to default as you suggest but thats not how the backtest executes and a big reason for this change is to improve execution consistency. Will see how that performs today.
-
@Jonathan-Jones I would be happy to have a look at it for you if you like (a second pair of eyes) if you send me both the BFBM bot and the BTC strategy files.
-
@Jonathan-Jones said in The BTC Horse Racing Thread:
Its an interesting one. The strat lays between 6 and 9.8 (so thats LTP is in that range inclusive). I have loads of extra bets taken. 2 from yesterday at Yarmouth 16:30.
Apex
More Than a GreyBoth have odds Req 7.2 and avg Price of 6.4 (i set to ask for 4 ticks worse to be sure i get matched) so they look fine right?
But, They are not in the software backtest this morning (BFBM took 4 bets for this strat on that race, The Software and Fairbot both only have 2). If i look at the chart in the software neither horse ever gets near 6 in the time window. The fact Fairbot also skipped them tells me BFBM is wrong in this case.The only other suspicious thing is BFBMs time window:
'before in play', is that before the official start or what because its impossible for BFBM to know when 10 seconds before actual inplay is. So, is BFBM using the official start for the before and the actual start for the to?
#Edit#
I dont think its the times. I just ran the software with a few variations and those bets dont show. I also checked the wording on BFBM and though the start is a bit odd the stop says 'start time'.Lose the stop betting bit for starters. As long as you don't have the "keep betting inplay" box ticked then it will automatically stop wen it goes into play. BFBM will stop placing bets at the official start time the way yours is set up I think. I could be wrong, but either way it will make the bot slightly more efficient as it is one less variable to calculate.
I think your problem is using LTP. I just use back and lay price(both, not either, or) and my stuff seems to match the software 98% of the time.
@Adam does a certain amount need to be traded at a price point in the software or is it just if it reaches that price regardless of the amount matched? Timeform for example needs £100 matched at a price for it to count.
-
Its an interesting one. The strat lays between 6 and 9.8 (so thats LTP is in that range inclusive). I have loads of extra bets taken. 2 from yesterday at Yarmouth 16:30.
Apex
More Than a GreyBoth have odds Req 7.2 and avg Price of 6.4 (i set to ask for 4 ticks worse to be sure i get matched) so they look fine right?
But, They are not in the software backtest this morning (BFBM took 4 bets for this strat on that race, The Software and Fairbot both only have 2). If i look at the chart in the software neither horse ever gets near 6 in the time window. The fact Fairbot also skipped them tells me BFBM is wrong in this case.The only other suspicious thing is BFBMs time window:
'before in play', is that before the official start or what because its impossible for BFBM to know when 10 seconds before actual inplay is. So, is BFBM using the official start for the before and the actual start for the to?
#Edit#
I dont think its the times. I just ran the software with a few variations and those bets dont show. I also checked the wording on BFBM and though the start is a bit odd the stop says 'start time'.