The BTC Horse Racing Thread
-
@Myles-O-Bery I don't know much about t&p values, I need to read more into high level statisical analysis (wish I studied it at Uni instead of Sociology and Counselling!).
I think you are right but the only caveat is that sometimes getting 500+ results in one course could take a long time depending on the strategy. That said as @Adam is adding in a few more years soon that won't really be an issue moving forward.
-
Yeah it's a bit tricky. Do you have much experience with significance testing (t & p values)?
Based on my calculations, some courses with low trades (100 - 200) are showing high confidence levels (90%+) based on their results while some courses with 1000+ trades show confidence levels of 80%.
Naturally, the more trades we have the more accurate the results are so I'm disregarding significance tests with less than 500-odd trades as I feel that, with anything below this, a couple of big wins/losses could skew the results.
What are your thoughts?
-
@John-Folan yeah you are onto something there, also we are moving further and further away from it which is the good news!
-
@Myles-O-Bery it's so difficult to say really, I mean you want a clear trend so results per course I'm guessing if after 100+ you are seeing a massive reason not to use it then that could be one to avoid. I don't tend to remove anything unless it is a really clear negative trend though.
-
Read the first post on both threads.
Horse Trading Resources
https://forum.betfairtradingcommunity.com/topic/1111/john-s-horse-trading-resourcesBfBotmanager
https://forum.betfairtradingcommunity.com/topic/2915/bfbotmanager-automation-discussion -
How many trades do you guys like to see before removing a particular course from a strategy? 100? 1000?
Would appreciate any recommended reads about statistics that relate to optimisation!
-
-
-
Thought I would move my discussion to about 2023 figures being the only reliable ones due to Covid here.
I’ve mentioned it on the strategies megathread. It really does appear to be a thing.
It also explains why a few strategies fell apart as the year went on.
Especially those where the big profit was 2021 and early 2022
-
I wouldn’t read too much into changeover months unless it is consistent in the software that it affects it. Our only problem there is it’s really only the last 12/18 months where we can rely on uncorrupted data(due to all the Covid crap).
Would really need to see data from pre Covid to draw real conclusions, or wait a couple of years as we get a fuller picture of the new world order post Covid.
That said. Nothing wrong with a bit of caution and there will always be more racing tomorrow(or next week, month etc).
-
@Benny-Grahn Oct and End Mar/April
-
@Martin Yes, great insight. What months are the crossovers?
-
@Martin The strategies had a hard time
but just trust the process (and backtesting) it will recover for sure. Hopefully not with to much draw down. -
@Martin Depends on your strategy, October has been an absolutely cracking month so far for me and could be the best of 2023 (famous last words lol).
@Tony Hastie Thursdays has been the worst day by far for me but the back testing data does not support this theory, in fact it suggests its the best day by a country mile. Mondays are my best days.
-
@Tony-Hastie yes I know a lot of horse traders avoid those days as the quality of racing is usually low.
-
@Martin Ive started minimum staking my pre BTC DOBQOB strategy(for the records) on a sunday, monday and tuesday. Been dissecting my results and those days really stand out as quite rubbish for the horses! Gonna have to check out the strategies Im looking at here too...
-
-
@Joshua-Varley brutal!
