Sigs, any reason you don't go stonks full time? Unlimited volume and you won't get banned.

cool, thanks for checking and pasting the link.

Nice to be able to line different model projections up side-by-side. Looks like Pythagorean Expectation and Elo agree on the favorite for more than half of these upcoming matchups, but a few big divergences too.

Moneylines and percentages are from the player 1 (left column) perspective.

I'm still paper trading this, but here is what frames based Elo would have bet into on BM. Of the 6 bets that overlap with JSkelt's Pythagorean Expectation model, there were 2 cases where the Elo model took the opposite side:
maflin v williams, williams -163
williams v mcgill, williams -157
yan v robertson, yan +162 (opposite side of pythag.)
perry v gilbert, perry +122
yan v perry, yan -157 (opposite side of pythag.)
williams v gilbert, williams -127
extra bets:
maflin v yan, yan -168
gilbert v mcgill, mcgill +108
robertson v williams, williams +162
maflin v mcgill, mcgill -117

Was Yan some prolific player ~10 years ago? Hes the guy we seem to disagree on. The data I used for PageRank scores and Pythag WRs was what Diced scraped, which starts in 2014.

Even by using just pure PageRank, which is I guess my control for player quality, I have Yan as the 42nd best player since 2014, and Perry as the 14th best. Most likely reason (that I can think if) that we'd differ so much is that Yan had a lot of success prior to 2014.

Just as an edit, I scraped the last 20 years of data from professional tournaments to see how much it changes things - and I still have Perry as 14th, but Yan as 78th. Does PageRank suck? Does Elo?

Good question. I checked and I don't have any data for Yan before 2014 either. Wonder what we are missing... let's see what happens with today's matchups.

Truly the Yan difference today - I wonder if I'm going too far back with the last 7 years of results. Maybe something shorter term matters more, or weighting more recent results somehow.

+4.67 units on 12.89 risked

I'll try and post what I have as the fair lines for tomorrow's matchups in a few hours. We're still dealing with a very small sample size, so I'd like to keep comparing projections. Maybe that will help us figure out why the models agree on most but not all plays.

Just noticed that there's 118 matches in the csv file where he score of both players is tied. Not sure what's going on with that.

Example:


How do they grade that? Tie? Higher points wins? I have no idea.

They're also all in 2020 which is suspicous and also the year I am using for backtesting which is nice.

If you scrape far enough back you'll run into some matches where they've put text in the frames field too - which is also very nice of them.

Also, I'll post my fair lines in a little bit - I just know I'm fading Yan again and I hate it

Loaded up on TME after Archegos forced-selling ended on Monday. Up 8.4% in 2 days, let's get that 300% for 2021 sigs.

I'm still down if you're still down.

I left tied games in the oddsportal dataset because Elo can use them as is. If I'm using an algorithm that needs to have a clear winner and I don't want to remove the tied games from the dataset for whatever reason, I duplicate the tied game and record it once as a player A win, and once as a player B win.

Here's what I have as the fair lines for tomorrow, and bets I found on BM:
mcgill v perry -+175, mcgill -119
yan v gilbert -+215, yan -122
robertson v gilbert -+187, no bet
williams v perry -+253, williams -152
robertson v mcgill -+127, mcgill +177
yan v williams +-167, williams -111
gilbert v maflin -+126, no bet
perry v maflin -+124, no bet
yan v mcgill -+148, no bet

Tried to sneak in 1 bet against Yan today but he wasn't having it.
today:
+1.47 units on 6.04 risked
total:
+6.14 units on 18.93 risked

*(I'm not including the projections I posted on 3/19 in the total here because those were pure Elo projections that only accounted for matchups won or lost. The current projections account for frame win rates and how many frames a player has to win to win the matchup, and I consider these two approaches separate models. If you want a strict evaluation of my entire track record of snooker bets, feel free to include the pure Elo projections from 3/19 as well; the return is still positive until now, but with a lower ROI percentage.)

who's got the info?

Snooker fair lines for tomorrow, with bets:
higgins v williams -+149, no bet
zhou v trump +-303, trump -269
wilson v carter +-118, carter +133
zhou v dott -+185, zhou -135
higgins v dott -+290, higgins -220
trump v carter -+163, carter +218
higgins v zhou -+249, higgins -184
wilson v williams +-127, williams +111
trump v higgins -+137, no bet
williams v dott -+219, williams -160
trump v wilson -+235, trump -184
zhou v carter +-144, carter -117

So between 9 and 9:05pm Pacific time (about 30min ago), the BM snooker lines all moved in the direction of the sides I posted here, and the two matchups I skipped were left as is. (I happened to have the menu open as the screen lit up). Doesn't seem like a coincidence, so I'm just curious if a silent 2+2er fired off some limit bets? If not, I wonder if some of the trading teams are following these threads...

far more likely the former than the latter (and no it wasn't me lol)

I don't mind if it was a 2+2er that moved the lines; just don't want to be doing some lazy sports trader's job for him while I'm over here working on commission lol.

well I haven't been adjusting lines for it at the book I work for, but 1) we aren't open to Americans 2) our snooker lines are more heavily juiced than BM (at least I think so) 3) our limits are most likely much smaller than theirs 4) I have far more important things to do lol

Oh! Well I appreciate the honest yet good humored feedback.

After you posted the whole BM line move thing, I went and checked out lines super quickly vs what you had and some def wouldnt't plays with us..although I didn't actually check if our lines had already moved and will not cuz I'm lazy haha

It looks like BM sets at least a 20 cent line when close to 50/50, and increases the cents as the size of the favorite increases (as would be expected). The Max to win on minus odds is $250, and the max to risk on plus odds is $250. This surprises me because the closing line metrics (accuracy & log loss or brier score) on the snooker market are about the same as the darts market, but although the limits on darts are lower, I've found it harder to beat than snooker.

so you actually made me check and I was wrong about my assumptions w the limits as ours actually around 1.4x theirs although ours are to win and not risk limits so it rly depends on the odds, but we have 6% margin which is roughly 25 cent lines. we also had less than 1k in total bets on ALL of snooker yest (I'm not revealing some confidential info there in case you're wondering since we show much is bet on the main lines for any event)