PennStateHoops.com Discussion Forum

Eyeballs-Stats general thread

Love the economist Tyler Cowen:

Stephen Curry and the duration of the great stagnation

In other words, this “technology” has been legal since 1979, yet only recently has it started to come into its own. (Some teams still haven’t figured out how to use it properly.) And what a simple technology it is: it involves only placing your feet on a different spot on the floor and then moving your arms and legs in a coordinated (one hopes) motion. The incentives of money, fame, and sex to get this right have been high from the beginning, and there are plenty of different players and teams in the NBA, not to mention college or even high school ball, to figure it out. There is plenty of objective data in basketball, most of all when it comes to scoring.

How about this for some advanced analytics: You are what your record says you are - at least in basketball:

Statistics-Free Sports Prediction

We use a simple machine learning model, logistically-weighted regularized linear least squares regression, in order to predict baseball, basketball, football, and hockey games. We do so using only the thirty-year record of which visiting teams played which home teams, on what date, and what the final score was. No real "statistics" are used. The method works best in basketball, likely because it is high-scoring and has long seasons. It works better in football and hockey than in baseball, but in baseball the predictions are closer to a theoretical optimum. The football predictions, while good, can in principle be made much better, and the hockey predictions can be made somewhat better. These findings tells us that in basketball, most statistics are subsumed by the scores of the games, whereas in baseball, football, and hockey, further study of game and player statistics is necessary to predict games as well as can be done.

Baseball:

ESPN: In the age of analytics, putting the focus back on scouting

Baseball, but on topic

Review of the great Michael Lewis’s new book, which apparently digests for laymen my favorite book of last decade or so, Thinking Fast and Slow (which doesn’t need digesting - read it!)

I just ordered that book (“Thinking Fast and Slow”), and it’s supposed to come tomorrow! I was having a discussion with someone last week about the stock market (he’s new to it), and I had recommended that he read “A Random Walk Down Wall Street” as a primer on investing. He, in turn, recommended “Thinking Fast and Slow,” saying that he likes to read books that smart people he knows are reading, and that was one of those books. I had forgotten which book it was that he recommended to me until I saw the title in a review of Lewis’s book. After I read that, I may have to take a run at Lewis’s book, too.

Slate actually printed a good sized section of the Michael Lewis book today, you can read here. Very nice read, going to pick up the whole book now.

1 Like

Listening to repeated declarations of “find/go to the hot hand” had me looking for more on the hot hand controversy.

This paper is good: Sloan Conference: The Hot Hand : A New Approach to an Old “Fallacy”

Finding: when adjusting for difficulty of shot, there is a hot hand. An NBA player who makes two or more of his last four shots is 2.4% more likely to make his next when controlling for shot difficulty (players who think they’ve a hot hand tend to take more difficult shots!)

In other words, a 45% shooter becomes a 46.1 percent shooter with a hot hand. So likely to make 1 more three-point shot of the next 100 than he had been previously if his shot selection is good.

To add to this, I don’t think the average basketball analyst is thinking “go to the guy likely to make one more shot than usual in the next 100!” when they urge seeking out the hot hand.

I suspect their intuition is more like “the hot-hand guy is now a 70-80% shooter”’, not “the 33-percent three-point shooter is now a 33.8% shooter - get him the darn ball!”

Moneyball ruins everything

http://marginalrevolution.com/marginalrevolution/2017/04/las-vegas-average-no-arbitrage-condition.html

Sports books have capitalized on big events, too. During March Madness, a five-person booth at the Harrah’s Las Vegas sports book cost $375 per person, which included five Miller Lite or Coors Light beers a person. In the past, seating at most sports books was free and first-come, first-served, even during big events. Placing a small bet or two could get you free drinks.

Ha, they got ripped off.

Spent this past March Madness opening weekend in Vegas. We did the Cosmo Hoops and Hops event, it was like 4 giant ballrooms opened up to form one big space that had HD projectors covering every wall the whole way around. There’s a video of it here to get an idea of how it looked: If you missed our exclusive college... - The Cosmopolitan of Las Vegas

It was between 150-250 a day (depending on single day vs multi day and VIP or not) and regular tickets didn’t include food, but was open bar all day and they were selling food inside the place all day with the menu changing from breakfast to lunch/dinner. You could reserve one of about 80 leather couches in the front for a bit more and be a VIP which DID include food. Othewise you could sit at high tops or giant wedding-style round tables that seated like 12 with decent enough chairs.

They had betting stations right outside the room with tv’s out there as well, never had to wait more than about 2 minutes to get a bet in if you timed it right.

I’ve been out there for MM before and tried to just get seats wherever we could find them with no real gameplan and it’s just not worth it. Having a comfortable seat with perfect views of all the screens is the way to do it, even if it’s going to cost you.

This was also one of the pricier events from what I saw, at least for general admission. But one of the best ran events as well, highly organized and very effectively managed. Was a lot of fun as well, whole place was packed but not crowded and everyone was living and dying on every point spread and total.

Missed this a couple of days ago. Wonder if Franklin does this stuff, thinking of his fourth-down tries in last year’s Big Ten championship.

I would welcome at this point a study of correlation between “bad body language” and poor performance.

He used KenPom final as the metric, because I guess he had to pick one as the standard.

Penn State: Final KenPom rank 43

  Mas FSH DII DOK Sag Kenpom YAG PGH PIG BWE RTP BPI TRank MOR Hess Comp Gasa Norlander
Final Rank 18 20 28 57 64 32 28 48 15 56 37 49 50 8 50 35 46 164
KenPom error 25 23 15 14 21 11 15 5 28 13 6 6 7 35 7 8 3 121
All teams MAE 47.39 42.98 47.01 47.72 42.07 41.03 43.47 48.11 50.66 49.82 51.01 43.54 40.31 49.12 41.97 42.16 40.40 45.16

Legend

Abbrev Name Analytics/Eyeballs
MAS Massey Analytics
FSH ERFunction Ratings Analytics
DII Donchess Inference Analytics
DOK Doktor Entropy Analytics
Sag Sagarin Analytics
Kenpom Pomeroy Analytics
YAG Yale USAG Analytics
PGH Pugh Analytics
PIG Pigskin Analytics
BWE B Wilson Empirical Analytics
RTP RT Power Analytics
BPI ESPN BPI Analytics
TRank T-Rank Analytics
MOR Moore Analytics
Hess ? ?
Comp Composite Both
Gasa John Gasaway / ESPN Eyeballs
Norlander Matt Norlander / CBS Eyeballs

Same numbers, ranked by Mean Absolute Error in Ranking

Name Analytics/Eyeballs Mean Absolute Error
T-Rank Analytics 40.31
John Gasaway / ESPN Eyeballs 40.40
Pomeroy Analytics 41.03
Hess ? 41.97
Sagarin Analytics 42.07
Composite Both 42.16
ERFunction Ratings Analytics 42.98
Yale USAG Analytics 43.47
ESPN BPI Analytics 43.54
Matt Norlander / CBS Eyeballs 45.16
Donchess Inference Analytics 47.01
Massey Analytics 47.39
Doktor Entropy Analytics 47.72
Pugh Analytics 48.11
Moore Analytics 49.12
B Wilson Empirical Analytics 49.82
Pigskin Analytics 50.66
RT Power Analytics 51.01

To many people’s question, here’s a subset ranked by predictive accuracy:

System Predictive Accuracy*
TeamRankings Pred 71.7
ESPN BPI 71.1
Pomeroy 71.1
Lefevre 71
Sagarin 71
T-Rank 70.9
Dokter Entropy 70.8
Donchess Inference 70.5
Yale USAG 70.4
Massey 70.2
Pigskin 70.1
B Wilson Empirical 69.9
Moore 69.8
Zamstat 69.1
Pugh 68.9
RT Power 67.9

* “Since the rankings used for this analysis are only updated once a week and do not take into account home court advantage and other factors that ranking systems may adjust for, this analysis is not fully representative of the included systems’ accuracies.”

MAEs of 40+ seem relatively large with a range of 353. Would be interested in the opinions of professional statisticians or other knowledgeable practitioners.

Great podcast that I was not aware of: Wharton Moneyball.

In this episode, Todd Golden of USF. His interview starts at about 59 minutes..

He said USF has five areas of emphasis in-game this season.

  1. Winning transition battle
  2. Scoring more than other teams in first 10 seconds of the shot clock. Plenty of data shows that you get a better shot, probably because the defense isn’t set.
  3. Shot selection: “Last night against Sonoma State, we took 73 shots over the course of the night. Only one of those shots came outside the paint and inside the three-point line. So that was something that was really, really good.”
  4. Winning rebounding and turnover
  5. Getting to the line more than the other team

(Interviewers said Princeton wants 95% threshold - only 5% of shots outside the paint or inside the three-point line.)

1 Like