The BTC Horse Racing Thread

Martin

@John-Folan think I need you to create some bots for the new strategies I just released!

John Folan

@Jonathan-Jones Fairbot is limited. It’s slightly faster than BFBM but the sheer scale of what is possible as filters on BFBM is brilliant. Ive fully automated my short odds lays so I don’t even need to import selections. (Still got a couple of courses to add to the course filters but that’s it. As they come up they will be added).

The other stuff is auto imported from the software and the filtering gets me a 99% accuracy with the software. Occasional differences but that’s life. I have thrown strategies away if I can’t get them close but that might be more down to being lazy rather than software. For inrunning I use bet Angel for speed. Its filtering however is terrible. Peter Webb has made it too complicated for most.(me included or I would use it for everything. It’s bloody brilliant for manual scalping tho).

BFBM will have the added advantage of automating football as well if you do it once the new revised software is out. Something to bear in mind. They are also pretty good at adding stuff to new releases if they think it’s useful. There’s a few of my suggestions that they have added over the last couple of years. A couple they haven’t. But I want the moon on a stick!

? Offline

@John-Folan
if you look at the 5th condition thats the favourite one. Fairbot does it by allowing you to set an order then do a compare. So its order by back odds then Back Price (named horse) <= BackPrice 4th horse. I think i have pushed fairbot a lot further than its meant to go and its well past its limits now. If BFBM can sort my issues with late non runners and allow me to run multiple rules against the same horse/ multiple horses in a race then i may just have to stump up the extra cash because there is no doubt my problems with executing according to the strategies are costing me a considerable sum at the moment.

Yes. Number of selections is number of runners

John Folan

@Jonathan-Jones thanks. I’ll have a play and see if I can get bfbm to play ball. I’m not a fan of fairbot as I just feel bfbm is better suited to what we do. Each to their own though. Some people swear by it.

There is no favourite condition though. That’s the only thing I could see throwing it out. But that doesn’t explain why the software is missing those you have said are in the top four that haven’t been taken.

I’m guessing selections is fairbots way of saying number of runners?

? Offline

@John-Folan

This is one of todays.

2 horses in the selections, a rule set for each.
You may wonder why the apparent duplication of the rules (back price and LTP). Its more of a thing on the lays, probably not needed on the backs but i found a lot of instances on the bigger lays where LTP and best avaible lay price were miles apart. So, it was a case of either go for entry at LTP (and not get matched) or go for entry at best lay and take a possible much bigger loss than intended. So, i added the check to make sure the softwares trigger was taken but i could also get a good price. Its not made any difference on the backs.

John Folan

@Jonathan-Jones can you show the conditions for your bot as screenshots?

? Offline

@John-Folan

That might explain ones that i dont have but the software does (not on my original selections in the morning but non-runners mean other horses move up a ranking then become selection) but cant see how it would explain ones i took but the software didnt.

@Adam
These are some from yesterday:

Doncaster 14:16 Taken Wannabe Brave Taken in reality, Not in the software
Listowel 15:00 Taken Karaoke @ 5.9 (but was triggered @6.1 so was valid) Taken in reality, Not in the software
Musselburgh 15:05 Taken Espressoo @ 10.5 Taken in reality, Not in the software
Musselburgh 15:40 Taken Gareeb @7.8 Taken in reality, Not in the software

All the above were taken but dont show on the backtest. All meet the odds critera, all are in the selections spreadsheet. All are in the top 4 at BSP. There were loads more differences but i can explain most of them by horses swapping places after official start but before actual start or not in the original selections (so probably moved due to non-runners).

I did have A fair few fairbot execution issues yesterday (some races dont look to of updated at all) so think i have had enough of that now (not the first time thats happened with Fairbot in streaming API mode). All in all a total points difference of 12 yesterday (not in my favour either - never is) which is becoming frustratingly common. Hopefully there is a very simple explaination for the ones above that i have missed. Then i can work to reduce/remove the issues. Think BFBM is probably an essential now as Fairbots limitations are becoming costly.

John Folan

@Jonathan-Jones said in The BTC Horse Racing Thread:

I am going to post this strategy here. Feel free to use it but thats not why i am posting it. What i need is some help. I created this one mid last year (so you can see its forward performance is very good) but i never traded it. I picked it up again last week and wondered why i never bothered with it. Found that one out pretty quick. It appears untradable. Its actual results NEVER match backtest results. Not one single day matches. Most are off by several or over a dozen points. I would like to know why. It does use the favourite criteria BUT it triggers in the last second and so far all of the trades it has taken in reality do have BSP in the top 4 (so its nothing to do with horses moving in the order). I did think its the pace rank issue. If horses are NR then that will bring horses in that were not originally in the selections but this would result in the software in backtest having trades that i missed in reality. For me to have trades in reality the software doesnt in backtest would mean horses have been entered into the pace ranking ahead of the original selections.

So, i am stumped. The differences are a mix of trades actually taken in live but missing on the backtest and trades on the backtest but missing in reality.

rules_export.json

It's because you are using pace rank. The non runners are throwing it out. It could potentially work in BFBM as you can get the selections to update hourly or even every minute in there so non runners would be a lot less of an issue.

? Offline

@Adam
Cheers. I will have a look in tomorrows results. I am sure there will be some.

Adam

@Jonathan-Jones Can you add a few examples of horses that traded IRL but don't appear in the backtest? I'll go through them (probably after the weekend) and clarify why they're not there.

Obviously avoid any examples from around 5pm on June 1st for now.

? Offline

@Toyin-Butler
Nope. All so far have been pretty much in the middle ground. No edge cases and its a lot of them. Forgetting about the 1st of the month which was a total mess because of the missing data yesterday resulted in an 11 point difference. It was still profitable but thats a lot of value to be missing out on. Some days i am negative by 5 or so points but should be almost 10 points up. Thats the scale of the problem. i just cant figure it out because there is nothing special about the strategy. If i was consistently missing trades only i would put it down to the pace ranking and non-runners but i cant see how non-runners could see me taking trades the software doesnt have in backtest (and so many of them). Maybe i am wrong and the issue is entirely down to non-runners but its a frightning difference if that is the case.

Toyin Butler

@Jonathan-Jones Are the trades that are missing outside of the odds range you have set? I've found that the odds filter isn't an exact science, for example I can run the same query for two different odds ranges and then another that combines them and I get a different answer.

i.e 2.0-2.99 and 3.0-3.99 does not equal 2.0-3.99.

When I drill down into the individual trades some appear in both odds ranges (at different odds) and as you would expect only once in the combined data.

? Offline

I am going to post this strategy here. Feel free to use it but thats not why i am posting it. What i need is some help. I created this one mid last year (so you can see its forward performance is very good) but i never traded it. I picked it up again last week and wondered why i never bothered with it. Found that one out pretty quick. It appears untradable. Its actual results NEVER match backtest results. Not one single day matches. Most are off by several or over a dozen points. I would like to know why. It does use the favourite criteria BUT it triggers in the last second and so far all of the trades it has taken in reality do have BSP in the top 4 (so its nothing to do with horses moving in the order). I did think its the pace rank issue. If horses are NR then that will bring horses in that were not originally in the selections but this would result in the software in backtest having trades that i missed in reality. For me to have trades in reality the software doesnt in backtest would mean horses have been entered into the pace ranking ahead of the original selections.

So, i am stumped. The differences are a mix of trades actually taken in live but missing on the backtest and trades on the backtest but missing in reality.

rules_export.json

John Folan

@Jonathan-Jones bfbm mate. Honestly apart from speed making it useless for inrunning(I use bet Angel for that). It is miles above any other bot provider for what we do. I’m more than happy to help set your bots up right too. I’ve done that for a few. You won’t regret making the switch.

Martin

@Ben-Ferguson-0 not that I know of. I just copy data over from downloads into a main sheet, delete results I don’t want.

? Offline

Is there any way to download/extract the results of more than one strategy at once for say the past day, or week?

? Offline

Does anyone know of a solution for strats that use rankings (like pace rank) with the obvious issue of late non runners messing up the order? Looking for an automated solution really. Its becoming quite a consistent pain in the arse and costing double digit points monthly. I use fairbot for automation and it doesnt have any facility to check stuff like that.

Adam

This is correct. It normally happens if you've imported the strategy from an old file or copied an old strategy.

When you re-select the columns you want and save the strategy (this being the key bit), the columns will be saved in the new format and shouldn't cause a problem again.

Nigel

@Jonathan-Jones thanks for reply

? Offline

@Nigel
Think its down to the version of the software the strat was originally saved in. I run a daily export and always use the same 5 fields. This was really frustrating me too. Just set it up how you want then save. You might also need to run it and save again. A couple of mine didnt stick but they are all ok now.