Botty Challenges

[gutuami]

if bf notices something like AI on their markets they will introduce a 90% premium charge . They like charges

[mcgoo]

(Just an exercise, extract you bet history to excel for this bot and see what profits you would have made if they were straight backs or lays). This will give you an idea if you are entering at a value position and visa versa
Regards
Peter

How very interesting My back bets profit twice as much as lay bets -over 250 trades only though..hopefully that continues..back bets only return 2.5% on turnover but is interesting
Nigel's spreadsheet analysing by track and race is also very insightful. One track already shows comparatively large Lay bet profit but time will tell.The good thing is I am building data up pretty fast (perhaps nearly 700 trades a week) so hopefully something really stands out over time and its not just another break-even bot Thanks again for the pointers.

[mcgoo]

if bf notices something like AI on their markets they will introduce a 90% premium charge . They like charges

No doubt

[PeterLe]

(Just an exercise, extract you bet history to excel for this bot and see what profits you would have made if they were straight backs or lays). This will give you an idea if you are entering at a value position and visa versa
Regards
Peter

How very interesting My back bets profit twice as much as lay bets -over 250 trades only though..hopefully that continues..back bets only return 2.5% on turnover but is interesting...

OK, I haven't read the full thread so not sure if you are backing first...either-way, you may want to look at how you place your lay bets...One thing you can do via excel is create a way if scenario; you can calculate what would have happened if you had placed the lay bets at one tick lower, two ticks lower etc and define at what point you would have made a profit...this is a good yard stick
If over a long period (250 is no where near enough), you find that the back bets are much more in profit than the lays, you can drop the lays and just have the backs going forward. You will have much longer periods of losing bets (as you are not greening up) but over time it will maximise your profits
(also when you have enough of your own data, you can then look at other possibilities...ie what would have happened if you had backed all the lays etc)
Im keen to know if you tried switching cross matching off too. (it will be less apparent pre off than in play, but over time will make a difference)

Regards
Peter

[bennyboy351]

if bf notices something like AI on their markets they will introduce a 90% premium charge . They like charges

No doubt

IMO, Betfair are just greedy b*stards! I mean, they are GUARANTEED a winner/profit on EVERY trade aren't they? I wish there were many more trading platforms 'out there' so that Betfair's virtual Monopoly could be reigned in! Any chance, Peter?

[mcgoo]

OK, I haven't read the full thread so not sure if you are backing first...either-way, you may want to look at how you place your lay bets...One thing you can do via excel is create a way if scenario; you can calculate what would have happened if you had placed the lay bets at one tick lower, two ticks lower etc and define at what point you would have made a profit...this is a good yard stick
If over a long period (250 is no where near enough), you find that the back bets are much more in profit than the lays, you can drop the lays and just have the backs going forward. You will have much longer periods of losing bets (as you are not greening up) but over time it will maximise your profits
(also when you have enough of your own data, you can then look at other possibilities...ie what would have happened if you had backed all the lays etc)
Im keen to know if you tried switching cross matching off too. (it will be less apparent pre off than in play, but over time will make a difference)

Regards
Peter

Not tried switching cross matching off yet but will try that. (All the trades are pre-off btw) I made a nice little profit last night over 68 markets, partially due to a green up not happening fast enough on 1 race (glitch in my timing that I have only seen fail this once) but it still did well enough/made more on the trading alone which is great. I am trying to get my head around how I can going to find the time to analyse the data properly but have a few ideas (adding yours ). There is a lot of info there, that's for sure. Fun times Thanks.

[mcgoo]

So I am at a bit of a loose end here. I have run my horse and dog bots over what must be thousands of markets now. They started out well and slowly got worse over time. I watched many hundreds of automatic trades and started noticing a performance issue which was clearly causing inconsistent results. I have mentioned the BA clock issue before before but noticed recently again that the VPS I was using would appear to get slower and slower..the BA clock would 'stick' (not countdown) and then suddenly catch up-sometimes up to 8 seconds. The CPU was running at 60%. I would restart and things would be ok for a while then it would start happening again. I decided to try another hosting provider and installed the latest BA version on a clean machine. The VPS was running windows 2016., 2 cores and 4GB ram on SSD. After a time..same problem so I thought right it must be the bot being too complex( it has 43 rules) I upgraded to 4 cores and 8 GB of ram. Same thing-runs for a while and then slows down. Tonight the clock sticking was worse than ever. Refresh is 20ms and restricted to 5 minutes. I checked the task manager and CPU never went above 40% and RAM utilisation never above 20% even during the worst of the "sticking", which doesn't make sense to me . The clock was stuck for up to 8 seconds again. Continuous pings reach Google in less than 1ms consistently-even during the clock issue.
The countdown sticking always seems to be more pronounced in the last minute before the race regardless of bets matched or not.The rules are not set up to change as things get closer to race jump time.
Removing the automation file/rules from the race immediately shows results in terms of the clock resumes normal countdown..re-applying the rules file causes the issue again. I suppose this would make sense in terms of a complex bot causing performance issues except that the VPS doesn't appear to be struggling for resources . I have 149 markets added to Guardian with the bot applied to all of them but restrict refresh is on as I said. I don't recall this being an issue with previous versions of BA but could be wrong there.Has anyone else seen anything like this or similar as I am not entirely sure how to progress from here.Thanks
PS: I tried polling but apart from the fact that I need a faster refresh..I noticed the same issue from what I recall(it was when the issue first appeared)
PPS: Just tried polling-same issue
PPPS: tried removing suspended markets..all good until last 45 seconds and then stuck clock for 8 secs again -this time with a ""Not responding" message at the top of the BA One click window

[jimibt]

mcgoo - one thing that i pulled my hair out for a while over was the fact that if you have a lot of logging going on, after a while BA slows up considerably!! i think that every time a new log entry is written, it takes the entire previous chunk back up into memory. thus, if you have a large logfile, over many days, then it grinds to a halt. two ways to mitigate this:

1. switch off logging (on signals and stored values) once in production
2. clear the log entry (from the main Guardian screen)

these two steps sorted out what sounds like a similar issue.

[mcgoo]

mcgoo - one thing that i pulled my hair out for a while over was the fact that if you have a lot of logging going on, after a while BA slows up considerably!! i think that every time a new log entry is written, it takes the entire previous chunk back up into memory. thus, if you have a large logfile, over many days, then it grinds to a halt. two ways to mitigate this:

1. switch off logging (on signals and stored values) once in production
2. clear the log entry (from the main Guardian screen)

these two steps sorted out what sounds like a similar issue.

Thanks Jimibt. I'll try those.Cheers-I did use a lot of logging to troubleshoot.I'll let you know
PS: VPS is doing updates.. ..might be tomorrow but thanks again

[foxwood]

the BA clock would 'stick' (not countdown) and then suddenly catch up-sometimes up to 8 seconds

May be relevant to your situation (or not!) - using RDP and I always see stuff like that on VPS when my local internet connection gets a bit sick.

I usually have countdown timer floating on the VPS and it obviously just "hangs" for a few seconds.

In reality it is still running ok on the VPS - I'm just losing packets and not receiving updates of the picture of the screen.

Have got click happy before now when that happens and of course the VPS sees all the random mouse clicks and places unwanted bets all over the ladders

[Dallas]

Removing the automation file/rules from the race immediately shows results in terms of the clock resumes normal countdown..re-applying the rules file causes the issue again. I suppose this would make sense in terms of a complex bot causing performance issues except that the VPS doesn't appear to be struggling for resources . I have 149 markets added to Guardian with the bot applied to all of them but restrict refresh is on as I said. I don't recall this being an issue with previous versions of BA but could be wrong there.

Depending if you need historical info for your rules you could shorten the number of historical data points captured by BA, by default it stores the last 7200 data points for each selection of each market, this isn't in the main options but if you PM me I'll let you know how to change it

[poklius]

I run my own horse racing python bot on 1CPU, 1GB RAM, 18 strategies, quite a lot of complex calculations, moving averages, sorting, searching, loging, saving raw json data to files etc. And even at peak times bot is at 20-30% of resources. 4CPU's for a simple bot is ridiculous..

[mcgoo]

mcgoo - one thing that i pulled my hair out for a while over was the fact that if you have a lot of logging going on, after a while BA slows up considerably!! i think that every time a new log entry is written, it takes the entire previous chunk back up into memory. thus, if you have a large logfile, over many days, then it grinds to a halt. two ways to mitigate this:

1. switch off logging (on signals and stored values) once in production
2. clear the log entry (from the main Guardian screen)

these two steps sorted out what sounds like a similar issue.

Thanks Jimibt. I'll try those.Cheers-I did use a lot of logging to troubleshoot.I'll let you know
PS: VPS is doing updates.. ..might be tomorrow but thanks again

That seemed to solve it.Thanks again Jimibt

[mcgoo]

Removing the automation file/rules from the race immediately shows results in terms of the clock resumes normal countdown..re-applying the rules file causes the issue again. I suppose this would make sense in terms of a complex bot causing performance issues except that the VPS doesn't appear to be struggling for resources . I have 149 markets added to Guardian with the bot applied to all of them but restrict refresh is on as I said. I don't recall this being an issue with previous versions of BA but could be wrong there.

Depending if you need historical info for your rules you could shorten the number of historical data points captured by BA, by default it stores the last 7200 data points for each selection of each market, this isn't in the main options but if you PM me I'll let you know how to change it

thanks Dallas. I seem to be ok for now but good to know.Cheers

[mcgoo]

I run my own horse racing python bot on 1CPU, 1GB RAM, 18 strategies, quite a lot of complex calculations, moving averages, sorting, searching, loging, saving raw json data to files etc. And even at peak times bot is at 20-30% of resources. 4CPU's for a simple bot is ridiculous..

Wish I could code like that