rickroll, if i get your question right, ganch on sbr forum explained this quite good.

you have to calculate your z-score. some excelsheet will do it.

https://www.sportsbookreview.com/for...ss.html?slf=67

This is where your frame based Elo should work out better. These matches are much longer than matches from other tour events. The final match for example is best of 35.

You could do a bankroll growth calc from the results and see where you stand. Median should be the expected growth of each bet compounded. If you are below that, then either you are running bad or your model overestimates edge. Above that either you are running good or have a model that underestimates edge. Of course a bigger sample set will tend to rule out run bad/good.

If you flat bet everything you can assume a default edge of 3% or something representative of your model.

Kelly fractions naturally take care of the variance issue noted above.

I noticed on cuetracker that the Magic8Ball algorithm (with 50 variables) has a current success rate of "65.94% out of 9557 predictions" so I ran tests on the most recent 10,000 samples in the oddsportal database. Frames-based Elo (with 1 variable) adjusted for "best of n matchups" has an accuracy of 67.15%; closing lines from the same set are accurate to 68.24%.

Unfortunately, wagering results from that set have a -3.04% ROI over 8,386 bets. Adding constraints (like a minimum player sample threshold) to the frames-based Elo starts to beat the closing lines of the same sample set by tenths of a %, and pushes ROI to +7.5% over 6,928 bets. I'm still curious if a different model can hit 70% or better accuracy over 10,000 samples, so I'll keep running snooker data through machine learning models (e.g. gradient boost classifier as discussed in earlier posts) as time allows.

JSkelts were you able to pull any metrics from your pythagorean points scored/allowed model? As good as your recent run has been maybe you are already pushing 70% acc. over a thousand or so samples?

To be honest I tried using oddsportal for backtesting but they store names by using a players first initial + last name, which actually has a ton of overlap in player data. That + I'm also using pagerank (different than how the paper did it though) in my model and it was a pain to recalculate each players pagerank score at the time of the specific tournament meant I couldnt think of a good way to backtest properly.

If there was a way to grab historical odds with first name + last name I'd do it but I really don't know off the top of my head what my historical performance would be

Can you concatenate the 2 names in a match from oddsportal and look them up that way in cuetracker? I'm assuming cuetracker has mostly unique names.

Unfortunately that would require me to think of the most obvious and straightforward solution. Something apparently beyond my mental capabilities.

hey guys, really appreciate all the feedback, got a bunch of work coming in last minute and no time to implement yet, should be done this weekend, excited to work over the various options

Ronnie O'Sullivan v Neil Robertson +100 1.00u

I have fair as -102 so its probably too close to be betting, but whatever

Europe World Cup Qual. UEFA Group B, Round 2
Kosovo - Sweden

Sweden to win @ 1,65

I have let the syndicate down with my absence but I'm back. I'm hitting 67.98% accuracy just looking at match outcomes without considering how many frames the match is. Probably should break it down by frame.

7.8% are first to 3
49.2% are first to 4
18.5% are first to 5
12.3% are first to 6
rest are all over the place

I'll get datasnooker csv files updating everyday with cuetracker results soon so we can crush or go broke on these remaining 1.5 months of snooker.

Wow Ronnie gets crushed, how bad can we run?

Ronnie O'Sullivan v Neil Robertson L -1.00u

+6.06u on 11.69u risked


More seriously I ran my model on these upcoming World Qualifiers (or whatever theyre called) and I'm really not confident in long term profitability. It seems like there are quite a few spots I'm far off market and taking some sizeable underdogs, so I'd be pretty interested in what numbers other people come up with.

You were on the right side of the market though, I got it at +102 and last I checked it was -115

Crowdsource syndicate already controlling the international snooker market.

must of been my $100 that swung the whole market

Here's what frames-based Elo adjusted for best of 5 frames has as the fair lines for tomorrow. Close to market on most, way off on a few.

robertson vs perry
player 1 win%=0.6933, p1 fair ML=-226, player 2 win%=0.3067, p2 fair ML=226
yan vs maflin
player 1 win%=0.6650, p1 fair ML=-199, player 2 win%=0.3350, p2 fair ML=199
gilbert vs mcgill
player 1 win%=0.4920, p1 fair ML=103, player 2 win%=0.5080, p2 fair ML=-103
williams vs maflin
player 1 win%=0.7984, p1 fair ML=-396, player 2 win%=0.2016, p2 fair ML=396
robertson vs yan
player 1 win%=0.5674, p1 fair ML=-131, player 2 win%=0.4326, p2 fair ML=131
williams vs mcgill
player 1 win%=0.7360, p1 fair ML=-279, player 2 win%=0.2640, p2 fair ML=279
gilbert vs perry
player 1 win%=0.5387, p1 fair ML=-117, player 2 win%=0.4613, p2 fair ML=117
robertson vs maflin
player 1 win%=0.7349, p1 fair ML=-277, player 2 win%=0.2651, p2 fair ML=277
robertson vs williams
player 1 win%=0.5096, p1 fair ML=-104, player 2 win%=0.4904, p2 fair ML=104
yan vs perry
player 1 win%=0.7211, p1 fair ML=-259, player 2 win%=0.2789, p2 fair ML=259
mcgill vs maflin
player 1 win%=0.5830, p1 fair ML=-140, player 2 win%=0.4170, p2 fair ML=140
williams vs gilbert
player 1 win%=0.8155, p1 fair ML=-442, player 2 win%=0.1845, p2 fair ML=442
white vs hendry
player 1 win%=0.5650, p1 fair ML=-130, player 2 win%=0.4350, p2 fair ML=130

I really want to shop some lines and post some bets from this too but I have to leave for the day job in a few minutes.

For the matchups starting on April 5th, do you know how many frames they are playing? I'll check later as well and post the Elo based win probabilities in a day or 2.

football - Spain Second Division A (Liga Adelante)

CF Fuenlabrada vs RCD Mallorca

RCD Mallorca AH +0,5 @ 1,44

Looks like best of 11.

http://www.snooker.org/res/index.asp?template=24

I didnt even know there were bo5s tomorrow, heres what I'd have for fair lines and bets

Neil Robertson -272 [-191] v Joe Perry 221 [191]
Kurt Maflin 141 [130] v Yan Bingtao -169 [-130]
Kurt Maflin 129 [216] v Mark Williams -155 [-216]
David Gilbert -131 [-131] v Anthony McGill 110 [131]
Mark Williams -147 [-178] v Anthony McGill 123 [178]
Yan Bingtao 161 [239] v Neil Robertson -194 [-239]
Joe Perry 122 [-102] v David Gilbert -146 [102]
Kurt Maflin 234 [314] v Neil Robertson -289 [-314]
Yan Bingtao -156 [124] v Joe Perry 129 [-124]
Neil Robertson -193 [-143] v Mark Williams 160 [143]
Mark Williams -120 [-135] v David Gilbert 100 [135]
Kurt Maflin -102 [120] v Anthony McGill -116 [-120]

So I'd take these bets:
Mark Williams v Kurt Maflin -155 -1.55u
Mark Williams v Anthony McGill -147 -1.47u
Neil Robertson v Yan Bingtao -194 -1.94u
Joe Perry v David Gilbert 122 1.00u
Neil Robertson v Kurt Maflin -289 -2.89u
Joe Perry v Yan Bingtao 129 1.00u
Mark Williams v David Gilbert -120 -1.20u

What do we think? Also I'm going to lineshop these as best I can so hopefully I can get better lines, but these are what pinny has.

And yeah the ones starting the 5th are bo11.

Ok, found 1 or 2 better spots (in italics) so here's what I'm locking in.

Mark Williams v Kurt Maflin -155 1.55u
Mark Williams v Anthony McGill -147 1.47u
Neil Robertson v Yan Bingtao -190 1.90u
Joe Perry v David Gilbert 127 1.00u
Neil Robertson v Kurt Maflin -289 2.89u
Joe Perry v Yan Bingtao 129 1.00u
Mark Williams v David Gilbert -120 1.20u


Also I forgot to mention that I have a minimum threshold for how +ev a bet is for me to take it - so some of the bets I posted earlier are +ev but my script just didnt print them out for me. I only take those bets when its Ronnie in a finals and I see an opportunity to lose internet friends/strangers money

Hi guys I'm new to betting and am in the Aussie thread.

Decided to bet some NBA games and was good on the bulls then curry came back. So I had to cash out today. Still got to let 2 ride though.

Good to see Snooker betting though.. Might have to do some

300% in a year we did it

stock tout time

gsk is the same exact price as one year ago so that was a fail

Psshhh stocks, ez market. We're out here on the streets trying to solve snooker.

^lol

"Also I forgot to mention that I have a minimum threshold for how +ev a bet is for me to take it - so some of the bets I posted earlier are +ev but my script just didnt print them out for me. I only take those bets when its Ronnie in a finals and I see an opportunity to lose internet friends/strangers money [/QUOTE]

Challenge accepted ...lets gogogoogo!