Weather is warm in Philly so the Nola over Ks project is back on, sigs.

I know nothing about the NFL draft but I just looked around at some lines and tried to find ones that were super off compared to others, how'd I do draft gurus?

Trey Lance u6.5 -115
Najee Harris u30.5 -135
Penei Sewell o6.5 +100
Rashawn Slater u9.5 -115
Jaylen Waddel o11.5 -106
Kyle Pitts u5.5 -133

Solid prices you got, I would say. Lance keeps going up and up on the U6.5, now to -165 I see this morning.

The juice still heavy, but the O4 CB's 1st round still seems like a good investment for at least push. Seems like a lot of books have shifted to 4.5 or even as high as 5.5. Surtain and Horn likely both gone top 15. Farley still floating with chances at some books as first DB off the board and an O/U of 14.5 (slam the over if you can get that number). Greg Newsome O/U 25.5. Asante Samuel Jr showing 32.5 and 52.5. Four to happen seems pretty safe (Surtain, Horn, Farley, Newsome, then might get a 5th in there).

Any thoughts on the sewell line? After betting these I asked some football friends and they made me feel like a massive donk, essentially saying that Sewell to dolphins/bengals at 5/6 was a virtual lock. I looked at a few mock drafts and they all have him going there as well.

My general sense has been that the 49ers ultimately take Fields or Lance.

If it's the former, the Falcons new GM has reportedly liked Lance for a while. Everybody wants to write "Matt Ryan has two years no matter what so they won't draft a QB," but it's not like Ryan is a superstar at this point. Not that he's had a "choice," but the Falcons new HC also has only worked at a high level in Tennessee, with mobile guys like Tannehill and Mariota. Also, if you "deep-dive" the situation, the Falcons have the lowest number of players under contract right now and their current cap space can only afford their 1st rounder anyways (this is without knowing CPatt's contract too).

So, what can the Falcons do? Are they honestly getting better from last year without the ability to sign any new players just because of a regime change? If you like Lance at all and believe he can be a franchise guy, then should be stashing him to learn.

Anyways though, if that happens, then as I've said before, does feel like it's some sort of Penei/Chase combo at 5/6 with Pitts as the only potential one to ruin it. But, if Falcons do decide to stay put and go Kyle Pitts as most suggested, then that opens up the chance Bengals move down because Lance is on the board.

It's an educated guess thing right now. Like I said, just unsure what the Falcons are supposed to do? You have a new regime you believe in, so hypothetically should be telling yourself "we'll never be this bad again. Ryan isn't going to be around much longer, might as well get our future QB now while we can." But hey, maybe they hate Trey Lance and/or NFL teams have done some pretty stupid **** before.

Hey crowd syndicate.

Who is ready to really help me tackle a challenging market, the Brazilian Big Brother, who be kicked out this week?

From what I can tell with my rough translating skills, it's between musician Fiuk, and normal ppl Caio and Gil, both from poor areas of Brazil.

Gil is super popular so has almost no chance of exiting (+10,000)
Fiuk is pretty popular and just survived last week so looks to go on (+350)
Caio looks to be on the outs (-666)

So, where is the value? Let's hammer this book.

There's also over under on what percent of votes they get, 60%.

Is there some way we can game the system? They have a simple captcha for votes, and on average get only a few hundred mill votes. An amazon virtual box could probably cheaply at least get the ratio under 60% for Caio, if that's what we needed.

Is the voting process a live event or has this been recorded a while ago and a certain number of folks already know the outcome?

It's a continuous live thing ongoing till the Tuesday night show at approx 10PM when the voting will be concluded and whoever has the highest number of votes will exit.

Caio up to -1000 to exit.

Fiuk +500

Saturday through today:
2-1 +1.00u on 9.6u risked
total
31-32 -4.59u on 96.3u risked, -4.8%

Spent most of the weekend tinkering with the current and some new models. After looking closer at the backtest results I posted the other day, I'm concerned about how slightly changing settings starts to cause -60 unit or worse drawdowns and periods of 200 and 300 bets where the model stays underwater, even if the final ROI is positive. Fractional Kelly makes it worse. I was able to get a machine learning model configured to beat the closing line's prediction and probabilistic accuracy more than Elo does and post a +ROI, but with swings so bad the bankroll gets as low as 20 units remaining at one point.

So I'm not convinced that I actually have a Snooker model that can beat the juice long term, and I'm going to put a pause on real money wagering for now. I want to study some more matchup video (was pleasantly surprised to find them quite entertaining to watch) and real-time market moves to see if I notice a feature that provides a bigger edge than what I've found so far.

Sounds like an interesting prop market, but not sure how much progress we'll be able to help you make by tomorrow night. I'm not sure what media source you were translating, Portuguese Twitter or Facebook perhaps, but maybe if you can find a few polls of the local viewership on those or other sites, you can use the wisdom of the crowd to try and price the possible outcomes. Assuming the bookmaker who opened the lines didn't already think of the same thing (which I doubt), maybe there is a little value left on some of the contestants with that approach.

Is this a Nola only thing? Is warm weather that much of a difference for certain pitchers or something?

This is from 2006 but it says strikeouts go up in cold weather https://tht.fangraphs.com/temperature-effects/

Question: what K value are you using for your Elo model? I was going to fit the optimal K value that yields the best EV for the 10 years of back data. Also considering parameterizing K for individual frames based on recency + frame results. Either way assigning some K value would be arbitrary and affect the results.

I agree about watching Snooker. The game is very strategic, and although not requiring as much mental space as chess it requires infinitely greater physical skills. Guys like O'Sullivan appear to be hard wired to play the game in such rapid fashion it seems they don't put forth any mental effort at all, yet they are fully aware of the myriad of possibilities.

Hey guys I'm new to sports betting. What you guys like this week?

For the basic implementation of the algorithm, I found 4<=K<=8 was the sweet spot for calculating the frames-based Elo score. Even when adjusting K dynamically based on the various parameters I could think of, the range I was using was 2<=K<=10. Anything over 10 and the over-reaction to recent results would start to get severe enough that after running the frames-based win probability through the Binomial distribution to get best-of 11 or 19 win probabilities for example, the model would be overshooting the market consensus vig-free line by 20% or more for some matchups. It's certainly possible that there is some situation I didn't think of that can take a higher K value and squeeze another 0.5 to 1% of accuracy out of Elo though; please let us know if you find something promising.

It's fascinating to see a talent like O'Sullivan in action. I was watching an old video of one of his 147 breaks in the Championship deciding frame, and it made me want to try and tackle some of the snooker prop markets like Centuries over/under eventually; want to understand snooker better before I start trying to model stuff like that though hah.

I see. So it seems your model assumes independence in frames during a match. Seems reasonable, but I'm not sure based on watching 100s of matches this is actually the case.

Meanwhile Trump is running the table in his first frame looking quite unbeatable. Let's see how he performs in the later frames.

Yes, I used that assumption as it greatly simplifies the modeling task. But speaking of oversimplifying the problem, I was surprised to find in some cases the opening lines for matchups between newer players was almost dead on the number derived from taking each players matchup win-loss ratio and running that through the Binomial cumulative distribution function to account for the number of frames needed to win the matchup. Obviously there's a more complicated process involved in setting the opening line for a matchup between O'Sullivan and Trump for example, and I'm not sure I was able to accurately reverse engineer the odds maker's process in those cases. Also, when I tried to account for the results of the recent matchups between 2 specific players or recent performance in a best-of-n setting etc, all it seemed to do was add noise to the model and degrade accuracy & ROI.

Still, I didn't do any rigorous quantitative analysis of questions such as "what is the importance of winning the first frame in a best-of-n matchup?" or "is the momentum and player streakiness observed between frames of a matchup or between matchups in snooker statistically significant?" etc. so there's definitely ideas left to explore in the search for more predictive features. We just need to make the time to go through them one by one and properly assess for statistical significance.

Selt gives away the first frame to Hawkins after being up 60-0. So Selt follows that up and gives away the 2nd frame as well. Some of these guys seem to be mental wrecks.

Higgins gave me a good sweat over the weekend for the 6.54 units I bet on him against Pengfei. After he lost frames 2 through 5 in a row to Pengfei, I mentally prepared myself for the loss. Didn't really get his **** back together until the last few frames.

The line is messed up (both listed at -200), but seeing Alabama 1st rounders listed 5.5.

That seems like a very easy over.

Mac Jones, Patrick Surtain, Devonta Smith, Jaylen Waddle are 99% locks to go R1.

Christian Barmore and Najee Harris are 90% likely to go R1.

And then Landon Dickerson and Alex Leatherwood have outside shots too.

Am I imagining this, or do I correctly remember seeing a snooker stat like "pot success rate" on cuetracker or somewhere else? It might have been on Bovada, like when the match is live I think the game stats that Bovada provides via www.statscore.com includes that stat. Anyone seen that, or even shot-by-shot level data somehwere?

I was trying to look at more granular data as another potential method for beating the snooker moneyline market, but also in the interest of attacking snooker prop markets. Since ELO initially showed promise, I didn't bother to look at points-per-frame level scoring characteristics until this weekend, and that data looks hard to work with as it usually tends to be bimodal, with one peak at 0 points , and another peak close to each player's average points per frame. Might be able to approximate that with a Beta distribution, but I would also like to see shot-by-shot records. Several Google searches have been fruitless, but if anyone knows of a shot level database somewhere, please let me know.

Man Snooker scoring distributions are weird. Here's Judd Trump for example: https://i.imgur.com/1dCvOeR.png

And here's all snooker frames since 2014: https://i.imgur.com/aFOiiYe.png

Huge spike at zero, then it looks sort of uniform between 1 and 50 (with a slight bump near zero scores), and a thick tail normal past 50.

I have no idea if this is a thing, but maybe we parameterize a distribution made of three parts. Bernoulli for zero/not zero. Uniform between 1 and 50ish. Cauchy for the rest. We use our one parameter to scale the different parts of this Frankenstein distribution like the bernoulli parameter, the range of the uniform, and the scale & center of cauchy.

To investigate this I started working on the first part, bernoulli for zero/not zero. A really weird thing I found was that the top players tend to score more zero frames than the median player. Maybe it's a bug, but I don't think so. https://i.imgur.com/PgofdfN.png X is zero frame rate, Y is frame win rate (Blue ones are the top 10 on the WST rankings). The line is the median zero rate. This could also be due to the top players having a stronger field since I think Snooker tournaments have brackets.

This is probably all trash, but at least we have like 4-5 months until Snooker is back in full swing to figure it out.

Literally anyone winning over .550 frame % is above median in zero frame rate percentage. This is either be a bug or something interesting.

You probably already know this but 0s happen a lot because frames frequently get conceded after the other player has an unbeatable lead. So if its 80-0 when the guy gets his turn back he just concedes and they move onto the next frame.

No my dumb ass still doesn't understand how Snooker works. This is the problem with just look at numbers and not trying to understand the actual sport haha. So maybe good snooker players are less likely to waste time on lost frames and will take the zero and move on. Maybe the middle guys have bigger egos and just want to lose by less.

This gave me an idea though. What percent of the time does a player cause his opponent to rage quit an unbeatable frame and take the zero? This graph looks a lot nicer and makes more sense. X opponent zero rate, Y frame win rate.
https://i.imgur.com/NCVgO9W.png

From the official rules of snooker: https://wpbsa.com/wp-content/uploads...iards-2020.pdf

(b) A player shall not concede a frame in any game, or
match unless any player requires penalty points. Any
breach of this Rule shall be regarded as Unsporting
Conduct by the player concerned. If the referee has
already Warned the player under Rule 1(a) above, the
next frame shall be awarded to their opponent.

My thoughts, I am not a snooker expert:
frames are only conceded when the losing player is needing snookers, and
the player is only 'taking a zero' when his score is exactly zero when conceding.

Conceding is for situations where the winning player has made a single break of 74 or more, or a couple of breaks with 73 or less points remaining.
When the losing player is at 0 points, this implies 5 red balls left for 67 points remaining.

I suspect that for pro snooker players there is a tiny bit of difference in situations for when they concede instead of play on (Ronnie O'Sullivan might prefer to concede when needing only one snooker, whereas all the other players would not) , but more difference lies in in how often they get in 'concede situations'.

Better players face other better players more often, and deeper in tournaments the matches are played over more frames.

So the better players might find themselves in situations of being 100-0 behind more often than the middle guys.