Its funny that simming has worked well for me and LoL totals so thats what I gravitated towards.

I feel like knowing if a guy was on the break is a pretty big deal for that form of data since (I have now learned) the guy going first can get enough points (way less than 147!) that the guy going second just straight up concedes the frame - which might get recorded as a 0. I dont know how common that is, but if theres a way to generate a split of that data I'd be interested in at least exploring it at the beginning.

For the modelling purposes though it might not matter since each person probably breaks the same amount of times in a match.

I'm open to trying out as many ideas as we can come up with. We can always look at error metrics, or use other algorithms to decide if a particular idea is generating more noise than signal and scrap it before actually trying to bet it.

You guys heard of datagolf? Pretty cool site huh. Bet you haven't heard of datasnooker...

http://datasnooker.com/

Introducing your direct source to download CSV files with match results. No per-frame results yet, but I can add that.

If you use python, you don't even have to download the CSV. Thanks to Al Gore and the information superhighway you can get it directly:

pd.read_csv('http://datasnooker.com/static/matches.csv')
pd.read_csv('http://datasnooker.com/static/tournaments.csv')

with data for over 19,000 matches, that's a beautiful thing!

Awesome, thanks. Is all this data from cuetracker? do they have a way to easily download csvs of their data or did you scrape it yourself?

Scraped it from cuetracker.

Now here is something interesting. The cuetracker guy has a model page. We could try setting his model as the lines and try to beat it. He has "over 50 variables" and the model grows "smarter and more accurate" everyday. Are we smarter than cuetracker? To be determined...

That sounds like a fun challenge, let's do it.

Watch cuetracker aka Magic8Ball control the international snooker betting market.

Sigs had the same problem with not having historical lines on faceoffs. It didn't end up mattering since every game was -115. But another thing we can try before bankrupting the cuetracker model.

I'm no expert on Snooker but the player going first probably is at a disadvantage. This is because the exact position of the cue ball and red balls is somewhat unpredictable. The opponent will then have a chance to either make a red (called a "pot"), or play an effective defensive shot (called a "safety") if the opponent decides potting a red is too difficult.

The player with the opening break generally gets the advantage only when the opponent subsequently tries for a long pot and fails. More often than not the opening break player finds himself with a tricky shot just to avoid a foul, or falling behind when the opponent pots a red and builds a long break.

I don't know how the order is initially chosen, but each frame alternates order until the match is completed.

Another detail that probably doesn't matter too much: both players can agree to restart a frame, if the position is such that neither player wants to take a risk to benefit the opponent. Playing order is unchanged. I don't think restarts are ever recorded in the final result.

Also, I have never seen any player pot a red on an opening break.

Oddsportal has Snooker lines going back quite a few years.

i'm excited to see us crack the snooker code. i will happily contribute in an open and collaborative way.

Oh ok I thought the guy going first (i.e. the guy with the first realistic chance to sink a red ball) was considered on the break. Snooker is quite complex.

Anyway, is there anything like just pure match logs for the matches.csv where it has scores per frame?

I'm not seeing any opening lines on oddsportal. How can we make any money when cuetracker aka Magic8Ball (tm) has extracted all the value from the line?

Hopefully snooker can repay you a fraction of what you've gave us Fubster.

I can get you scores per frame. Give me a day or two. Just can't tell who went first in the data. Maybe we can reverse engineer the tournament structure rules to determine this. Or perhaps we'll have to crowdsource a google sheets document where everyone is watching Youtube videos.

Unpredictable shots (called "flukes") can play an important role in determining an outcome of a frame. I guess this would be like a fumble in American Football. If you could track those shots, and discount those wins it might lead to something.

Works for me, if you tell me what page specifically has it I can try my hand at scraping it too. Also dont worry about the who goes first thing, I no longer think it matters.

You need to setup an account. When I setup mine a few years ago it was free, not sure now.

It's just on the tournaments page you select a tournament and you get something like this:

7-61; 40-74(56); 78-75; 77-4; 111(111)-0

I don't understand what the numbers in parenthesis mean because I don't know what snooker is. But if you ignore them they add up to the scores row.

That's such a mess I changed my mind, I'll let you handle scraping that part

That said, I played around with the data you posted and have a dataframe of each players average points for/against per frame if anyone finds that useful.

I am no snooker expert.

Pretty sure that that number in parenthesis is the highest break of a frame, break being the number of points scored by a player in one single visit to the table.

I suspect that only 50+ breaks are mentioned, possibly all 50+ breaks? Pretty rare to have two 50+ breaks in a frame. Not 100% sure either if only the highest break of the player that has won the frame is listed.

In absence of many other player scoring statistics other than total points per frame scored, the 50+ break count might prove useful for some 'power indicator' per player? But the number of 50+ and 100+ breaks per player are probably registered somewhere, I will have a look if I can find those.

I think that the professional players should more or less be capable to clear the table when the balls are setup easily enough and therefore expect that the model needs more than only the scoring.

ELO style model may be useful: form factor for players is probably quite significant. Also I hunch that there are good opponents and bad opponents for players based on speed of play and style of play? (attacking or more risk-averse)

This is 3.05 now.
We don't really want this type of match go to the judges (same promotor as Mike Tyson - Roy Jones) and match might be rigged but the bout is scheduled for 8 rounds.

When checking boxing odds I found Antony Joshua - Tyson Fury unconfirmed fight, AJ for 2.75 for a fight to happen in 2021. This ties up some money for a while but I feel that this should be a coinflip at best for Fury because of concerns of him getting into shape.

Could it be that the (soft) bookies give too generous odds on the side they expect less money coming in on?

https://www.abc.net.au/news/2021-03-...belts/13252028

looks like june-july

AJ should be the fav but I see how they would make fury the fav with the hype

I googled on this. Being first seems to be advantageous and initial order is coinflipped. The guy from the article below used to be a top 10 player, died much too young.

http://news.bbc.co.uk/sportacademy/h...00/3986047.stm

If this forum can survive the loss of RickJ who posted winning plays each day for years I think we can survive without rsigley. You know, looking back I saw kids like greensmoke85 fresh out of college and I worried about the next generation of sports bettors. I don't worry any longer. The new generation is smart and hardworking. It's the books that should be worried.

News sites are reporting today that, surprisingly, John Higgins is the best snooker player ever according to a new study at University of Limerick.

https://academic.oup.com/comnet/arti...DKibfLGO8.link

This might be a good read also for other 1 vs 1 sports.
The link contains a link to a Github with tons of data in csv.
Source of most data is cuetracker.net

Nice that makes sense. I'll get that scraped as well.