Hi All,
Peter's latest video "Why do betting systems fail the moment you start using them" (https://youtu.be/QoWg97WGZ8I?si=rkzi9tzOLXoYv4Yt) was very interesting, and is something I've experienced many times.
I'm not particularly guilty of 'backfitting' results so far that I come up with rules like 'back trap 2 at Monmore on a Tuesday if it's a black dog' but I suspect I do tinker/filter enough that I'm backfitting to a degree.
I'm OK with Excel and use it to place bets via Guardian, so I'm pretty good at creating rules etc. in the software but I can't help thinking I'm missing something when it comes to data analysis.
I'm not a statistician by any means, and some of the maths is a bit beyond me (although I try my best to learn) so my question is do people have any tips or tricks when looking at data. I'm sure we all looks at different stuff, but in terms of general analysis, is there any useful rules you have or any Excel functionality you make use of?
Cheers
Data Analysis
-
paspuggie48
- Posts: 674
- Joined: Thu Jun 20, 2013 9:22 am
- Location: South-West
For me I try to keep it as simple as possible i.e. Hcaps, Card #1, no AW racing...that's about it.
Agree with your sentiments about micro-filtering data "to make it fit". So many times I've seen people say don't select, for example: John Gosden horses on good ground, only 5f, 7f or 1m3f, between May 12th and Aug 27th, not on a Tuesday afternoon, not at Chester, Doncaster & Lingfield, racecard number between 7-12 and there has to be minimum 8 runners and above 3/1 30 seconds before the off. It's just bonkers !!
No poisson, no special heatmaps or formulas for me, I just look for consistency over a period of 1, 2 or 3 years of data and if it generally makes long term profits with few losing months then I'll add it my system portfolio and try "live". So for me, it's a case of having tonnes of records and applying simple filters.
However, my past data and actual selections were consistently profitable in the past 2 years with minimal losses on just a few of those months with one large losing sequence last September. That changed between Jan-March this year when it lost 5x more than that previous largest losing month. I don't mind equal the losses or twice as much but this was 5x more
Lesson learnt and I now know that it is possible, in future, that there may be a chance of a large losing month or two. Hopefully not though
Agree with your sentiments about micro-filtering data "to make it fit". So many times I've seen people say don't select, for example: John Gosden horses on good ground, only 5f, 7f or 1m3f, between May 12th and Aug 27th, not on a Tuesday afternoon, not at Chester, Doncaster & Lingfield, racecard number between 7-12 and there has to be minimum 8 runners and above 3/1 30 seconds before the off. It's just bonkers !!
No poisson, no special heatmaps or formulas for me, I just look for consistency over a period of 1, 2 or 3 years of data and if it generally makes long term profits with few losing months then I'll add it my system portfolio and try "live". So for me, it's a case of having tonnes of records and applying simple filters.
However, my past data and actual selections were consistently profitable in the past 2 years with minimal losses on just a few of those months with one large losing sequence last September. That changed between Jan-March this year when it lost 5x more than that previous largest losing month. I don't mind equal the losses or twice as much but this was 5x more
Lesson learnt and I now know that it is possible, in future, that there may be a chance of a large losing month or two. Hopefully not though
Excel graphs have interesting and useful feature - "trendline" with option “display equation on chart” for modeling linear or non-linear relations (polynomial is the most interesting).
Correlation function can be used to measure relations (e.g. SP vs "some rating") with a "correlation is not causation" in mind.
ps. paspuggie48, dont post anymore or ww3 will start (665 posts)
!
Correlation function can be used to measure relations (e.g. SP vs "some rating") with a "correlation is not causation" in mind.
ps. paspuggie48, dont post anymore or ww3 will start (665 posts)
!
Last edited by napshnap on Sat May 11, 2024 2:05 pm, edited 1 time in total.
-
getagrip68
- Posts: 44
- Joined: Thu Oct 14, 2021 5:56 pm
Thank paspuggie48, your approach sounds similar to mine. I had a shocker the early part of this year too, wiped out a third of my bank, fortunately I stake conservatively (% wise) thesedays, in the past it could've been worse. It's tough going through a bad run but my records show that if I can stick it a good run usually follows, reversion to the mean I guess.
Cheers napshnap, I've tinkered with graphs and regression analysis but will revist and have a look at polynomial. I'll also have a look at the CORREL function and see if I can make it work in my dataset.
Cheers napshnap, I've tinkered with graphs and regression analysis but will revist and have a look at polynomial. I'll also have a look at the CORREL function and see if I can make it work in my dataset.