Hello there
For all those experienced in the automation side of things, what would you say is a good sample size when deciding when to promote an automation from practice into live?
I appreciate that practice mode and live mode will have their differences but was wondering if anyone had any thoughts?
Sample size
rich - i have pretty much the same puzzling thoughts and have concluded that there definitely needs to be a good long trawl against as many suitable markets per day for a period of a good few weeks. Of course, during this time, often things get tweaked a little, so the jury is out as to whether you restart the process on any change, no matter how minor. However, as I'm sure you'll have grasped, this could last forever. SO - I've tended to strike a balance based on it being as good as I feel it can reasonably get, whilst keeping an eye on the avg P/L over the period of a few weeks. Then, I put that particular rule into live use and monitor it's performance with small stakes for a while and if it still shows promise, will then promote it gradually to larger stakes if appropriate.ricardodeano wrote:Hello there
For all those experienced in the automation side of things, what would you say is a good sample size when deciding when to promote an automation from practice into live?
I appreciate that practice mode and live mode will have their differences but was wondering if anyone had any thoughts?
TBH, it's pretty hit or miss and I'm sure you'll have had great days, follwed by poor days with little or no visible reason as to why. I guess a bit of faith is required along the way
[edit] -as an aside, I'm more and more distancing myself from IR scenarios, in favour of pre-race strategies. I'm also tending to look (where possible) to lead with back bets, rather than lays. just a little re-evaluation of risk/reward ratio over the past wee while.
Rich, after 3 weeks of testing various strategies I took the plunge and used "real money" tonight for the first time. Was pig sick after the first 3 races but stuck with it to the end of the night. It turned out ok. Looks not to work on shorter length races so can pick my targets next time. Good Luck
G
G
You do not have the required permissions to view the files attached to this post.
I would nt say there is a right or wrong time to go from practice to live mode, just when you do make sure you stick to a stake size where any run of losses are acceptable, One of the worse things to do is start in live mode with a larger stake then reduce it if you encounter a few early losses.
Athos55 screenshot is a perfect example of this (by the way well done) but had he reduced his stake after those first 3 losses (something a lot of users would of done) then its highly unlikely he will still of been in profit at the end of the night. So always start with what your comfortable with and dont need to reduce and build it slowly over time once the rule has built up a profit bank
Athos55 screenshot is a perfect example of this (by the way well done) but had he reduced his stake after those first 3 losses (something a lot of users would of done) then its highly unlikely he will still of been in profit at the end of the night. So always start with what your comfortable with and dont need to reduce and build it slowly over time once the rule has built up a profit bank
Stake is only £2... 
Still got a long way to go to get back to profit. Speaking as a newbie to all this all I can say is I picked up snippets from the posts of others, put together a load of potential strategies of sorts and ran them in practice mode. It only costs you time...
Still got a long way to go to get back to profit. Speaking as a newbie to all this all I can say is I picked up snippets from the posts of others, put together a load of potential strategies of sorts and ran them in practice mode. It only costs you time...
I personally think that if you are in live mode, it focuses the mind more!
I find that in practice mode I feel more relaxed and therefore less focused. However, when in live mode I find that I concentrate more and I study things much closer. OK I use the smallest stakes possible, but like most of us, I really hate to loose.
One very useful tool that I have found is a little piece of software that records my screen. So I set it up to capture BA from start to finish. Once I have done this I then replay the video loads of times and check the action closely at certain moments of a race. I am then able to see if timings and scenarios can be tweaked to give optimum results.
The software that I use is called CamStudio.
I find that in practice mode I feel more relaxed and therefore less focused. However, when in live mode I find that I concentrate more and I study things much closer. OK I use the smallest stakes possible, but like most of us, I really hate to loose.
One very useful tool that I have found is a little piece of software that records my screen. So I set it up to capture BA from start to finish. Once I have done this I then replay the video loads of times and check the action closely at certain moments of a race. I am then able to see if timings and scenarios can be tweaked to give optimum results.
The software that I use is called CamStudio.
-
ricardodeano
- Posts: 205
- Joined: Tue Oct 06, 2015 2:03 pm
Thanks for the input everyone, really interesting to get people's views on this.
Jimbit - Thanks for the pointers, I may have to consider pre trade automation.
Atho55 - good first set of results, I'd happily take that all day long!!! Was that a singular £2 stake or firing lots of £2's into the market? i.e. low / high transactions?
Excellent advice on CamStudio Gundog, I was thinking of recording to see what went right/wrong so if it works for some, it can work for muggins here!
Jimbit - Thanks for the pointers, I may have to consider pre trade automation.
Atho55 - good first set of results, I'd happily take that all day long!!! Was that a singular £2 stake or firing lots of £2's into the market? i.e. low / high transactions?
Excellent advice on CamStudio Gundog, I was thinking of recording to see what went right/wrong so if it works for some, it can work for muggins here!
I use a similar tool called OpenBroadcaster: https://obsproject.com/, definitely a good additional aid to have in the kit when trying to figure what happened in the frenetic live market place. On that note, I did ask for a feature to be added to BA to allow historic markets to be pulled in and thus automation played out against those as many times as is needed to tweak the rules. The actual post is here: viewtopic.php?f=20&t=11983GunDog wrote: One very useful tool that I have found is a little piece of software that records my screen. So I set it up to capture BA from start to finish. Once I have done this I then replay the video loads of times and check the action closely at certain moments of a race. I am then able to see if timings and scenarios can be tweaked to give optimum results.
The software that I use is called CamStudio.
As an aside, one of BA's (automation) competitors uses a feature to perform exactly this function. I'm not promoting that product as I'm here to stay, but is well worth checking out this truly useful feature as this is where the real gains (IMHO) are to be had in terms of fine tuning those pesky rules that work 90% of the time [the other 10% being the unexplained losing runs]).
anyway..
I often have the following problem: I like to mess about with automation in the mornings, as I am an early riser. The trouble is that I am really eager to test them out immediately, but I have to wait till about 1400 to actually test them.
I think a really cool feature would be for BA to run some old races in a loop, so that any automation can be tested by linking into that loop??
I think a really cool feature would be for BA to run some old races in a loop, so that any automation can be tested by linking into that loop??
I agree and if you read my topic on this: viewtopic.php?f=20&t=11983, I even go as far to point out how BA could make this a revenue opportunity. So, win win I'd say!! We just need enough vocal support for this to push it fwd. It's not a new request by any means, but my slant on it may just allow BA to bypass the Betfair licencing issues by using a 3rd party to supply a BA bespoke fileset for historic races, at a small but reasonable price (i.e. 0.50-0.75p per meeting)GunDog wrote:I often have the following problem: I like to mess about with automation in the mornings, as I am an early riser. The trouble is that I am really eager to test them out immediately, but I have to wait till about 1400 to actually test them.
I think a really cool feature would be for BA to run some old races in a loop, so that any automation can be tested by linking into that loop??
Betfair require a different license when it comes to providing historical data, if you ask them for the details they reply with:
https://github.com/LiamPa/syntheticapi
I built my own 'synthetic API' that allows me to run markets that I have recorded with my own programs, not perfect but has saved and made me a few £'sPlease find that we plan to update historical data licensing soon as we have taken your details to contact you when details of the scheme are completed.
https://github.com/LiamPa/syntheticapi
this might be the time then to get into pythonLinusP wrote:Betfair require a different license when it comes to providing historical data, if you ask them for the details they reply with:
I built my own 'synthetic API' that allows me to run markets that I have recorded with my own programs, not perfect but has saved and made me a few £'sPlease find that we plan to update historical data licensing soon as we have taken your details to contact you when details of the scheme are completed.
https://github.com/LiamPa/syntheticapi
. Regards the licensing, I'm advocating that BA use a pre-existing authorised reseller of betfair data (Fracsoft) in order to keep sweet with betfair and at the same time, bag a little extra renewable revenue!! To be able to backtest the automation rules is almost (imho) an entry level requirement. I'm not saying that BA haven't mulled this over since adam was a boy, but am suggesting that there are alternative ways to achieve the same aim.
Having a backtest feature would enable so many rules to be versioned and *signed off* thus allowing for more productivity in terms of moving onto the next objectives (be that manual trading or further automation testing).
If I were polled as a customer (which I am), this would be at the top of the list of features that would help retain my (ongoing) interest.
already done so a few days backPeterLe wrote: Log it in the Suggestions thread so it dosent get lost
Peter
- take a quick look and cast your vote!! viewtopic.php?f=20&t=11983