Sports Statistics

[Jason Strasser (strassa2)]

Im really sick of the results oriented nature of sports and sports statistics. Part of what makes sports fun is the ability to be results oriented--to balance it with poker is very fun for me because I totally shun the results oriented side of my brain when im gambling, but when I watch sports I let myself get carried away.

However, there is a need for some better, more complicated statistics. The gist of what I want is to see more efficiency-based calculations. There are so many skewed stats that could be replaced by more meaningful efficiency stats:

Football is full of crap stats, almost every stat is meaninless. The XYZ team is 5th in rush defense, for example, is generally meaningless. Rush defense as a function of total yards is more of a function of how good your team is. If you get up big, the other team must throw the ball a lot. Comparing one team to another in their rush defense rankings is fairly useless, except for extreme examples. If there was a way to express rush defense as a function of: number of yards/carry, strength of opposing backs, strength of opposing D line, etc.

How about total rushing yards for a running back. I find this pretty useless as a judge for RB skill. How about some sort of efficiency rating that incorporates skill of the O-line, yards/carry, ability to break tackles, strength of the system, etc. ?

On ESPN there is some ranking system for the NBA and I find it very interesting. Allen iverson and Arenas are both way down the list.

I think more advanced calculations (like Passer Rating) would be great for sports fans.

[JaredL]

Strassa,

I'm just getting into a lot of sports stats myself. I eventually would like to do an analysis of play calling in the NFL. Run vs pass on 1st and 10 being the most simple. It appears that most teams pass too little (the expectation from passing is greater than running), but I think something like the oft discussed reducing variance in tournaments may make running just as good.

I would highly recommend that you read Baseball Between the Numbers if you haven't already and care about baseball, which you didn't mention. It's a book full of good statistics and strategic stuff. Some chapters are written by 2+2er Nate tha Great (or however he spells it). The most interesting IMO was, and discussed whether Ortiz is clutch. Basically, he created a stat that uses probability of winning through various batting outcomes to determine whether some hitters do better in more important situations.

For football, www.footballoutsiders.com is pretty good. I haven't read everything, but there is a lot there. They developed some fairly deep statistics that seem pretty good. I would reccomend reading The Hidden Game of Football as well. It was published I think in 1987 and apparently is out of print but your local library probably has it.

[Kneel B4 Zod]

Strassa,

Football Outsiders

check out DVOA and DPAR among other things. also read The Hidden Game of Football

Baseball Prospectus

VORP, .EQA, etc

also read Baseball Between the Numbers

basketball is a bit behind (at least to the public), but check out 82 games and read about PER. also see this thread here

[ QUOTE ]
I think more advanced calculations (like Passer Rating) would be great for sports fans.

[/ QUOTE ]

Passer Rating is actually pretty poor

[w_alloy]

I too wish there were more good statistics, but at least it is getting better.

Moneyball kind of sparked a "stats revolution" that has been happening over the last four years in the major sports, but especially baseball. I know there are some great baseball stats websites out there (I think nate tha great wrote for one for a while). Too bad I'm not a big baseball fan.

The NBA has come a long way in the last couple years, with some of the top stat measuring guys in the media getting jobs with NBA teams. www.82games.com is an awesome resource for meaningfull stats and discussion on stats in the NBA, but could use a lot of work.

I would love to see some good work devoted to the NFL and various soccer/football leagues, although the NFL can be really tough because of small sample sizes and soccer can be tough because you need to make up a lot of more subjective stats that are only now starting to be kept track of (like keeping track of touches and individual possesions).

Edit: I see im slow to the post. Baseball prospectus was the site i was thinking of, and thanks for the NFL link.

[7ontheline]

John Hollinger and David Berri are doing some interesting work on the NBA in terms of statistical analysis. John Hollinger is on ESPN and David Berri wrote Wages of Wins.

For football, check out footballoutsiders.com, as previously suggested. Excellent site. KC Joyner on ESPN is interesting, but I often feel after reading his articles that a lot of what he does ends up based on a fairly small sample size.

Baseball Prospectus, of course. Duh. Baseball lends itself better to this sort of sabermetrics because it is so much more of an individual's game. The team aspects of basketball and football make it more difficult, which is why this statistical stuff is less developed in those sports.

[UMTerp]

Jason, I used to keep a site with some ACC basketball efficiency ratings that I contrived that I bet you'd find interesting, but I took it down a year or so ago. Now I just have the data saved as a bunch of spreadsheets on my computer. Seemed to work out well though.

Brief overview (and I realize the formula's not perfect from a statistical standpoint, it's just something I created as a kid and tweaked over the years):

Offensive Efficiency
Factors are points, field goal attempts, and free throw attempts. Basically a gloried field goal percentage stat.
OEFF = Pts / (2*FGA + FTA)

Total Offense
Includes Points, assists, and turnovers. Rating is based completely on a per minute basis. An attempt to demonstrate a player's total offensive contribution within the contest of the number of minutes played.
TOFF = (2*AST + PTS - TO) / Minutes

Possesions Gained
Again, all per minute. Includes rebounds, steals, blocks, and personal fouls. Stats are weighed diffferently. This area collects the so-called 'defensive' stats.
POSS = (REB + 1.5*STL + BLK - PF) / Minutes

Win Rating
This is an effort to measure intangibles. It's not easy, since there are no stats for them. Minutes played for your team and your team's winning percentage are the main factors in this area. Individual defense obviously gets lost a little in this formula, but this component is an effort to reward the players a little bit if their coaches see fit to have them on the floor.
WIN = [(Player Minutes / (Team Games * 40)) * Team winning percentage] / 2

Rating = (OEFF+TOFF+POSS+WIN)*50

-------------------------------------------------------

This year's Top 20:
Rank Name Team Pos Class OEFF TOFF POSS WIN Rating
1 HansbroughT UNC F Soph 0.591 0.629 0.276 0.305 90.054
2 LawsonT UNC G Fr 0.577 0.748 0.171 0.262 87.892
3 DudleyJ BC F Sr 0.645 0.584 0.221 0.278 86.422
4 SingletaryS UVa G Jr 0.570 0.762 0.117 0.273 86.087
5 WrightB UNC F Fr 0.632 0.553 0.261 0.272 85.839
6 ThorntonA FSU F Sr 0.611 0.596 0.252 0.245 85.210
7 RiceT BC G Soph 0.581 0.684 0.105 0.291 83.063
8 CrittentonJ GT G Fr 0.548 0.702 0.151 0.245 82.272
9 DowdellZ VT G Sr 0.568 0.642 0.160 0.273 82.126
10 StrawberryD MD G Sr 0.548 0.618 0.168 0.293 81.320
11 ReynoldsJ UVa G Sr 0.565 0.703 0.084 0.264 80.836
12 GistJ MD F Jr 0.574 0.485 0.293 0.255 80.340
13 McRobertsJ Duke F/C Soph 0.538 0.497 0.273 0.294 80.038
14 VisserK WF C Sr 0.600 0.560 0.252 0.175 79.332
15 TerryR UNC F Sr 0.603 0.523 0.228 0.219 78.638
16 MaysJ Clem F/C Jr 0.499 0.526 0.286 0.245 77.750
17 RiversK Clem G/F Soph 0.586 0.564 0.149 0.253 77.577
18 McCauleyB NCSt F Soph 0.596 0.529 0.182 0.240 77.352
19 EllingtonW UNC G Fr 0.543 0.612 0.146 0.244 77.235
20 HammondsC Clem G Jr 0.585 0.545 0.115 0.290 76.732

Top 20 Seasons since 1980-81:
Rank Name Year Team Class OEFF TOFF POSS WIN Rating
1 DuncanT 1996-97 WF Sr 0.621 0.653 0.460 0.355 104.415 (NPOY)
2 SampsonR 1982-83 UVa Sr 0.625 0.614 0.427 0.362 101.386 (NPOY)
3 LaettnerC 1991-92 Duke Sr 0.673 0.685 0.291 0.370 100.946 (NPOY)
4 WilliamsJ 2000-01 Duke Soph 0.594 0.936 0.131 0.357 100.896
5 AndersonK 1989-90 GT Fr 0.580 0.872 0.188 0.377 100.860
6 MayS 2004-05 UNC Jr 0.613 0.682 0.415 0.299 100.444
7 FerryD 1988-89 Duke Sr 0.589 0.867 0.238 0.314 100.377 (NPOY)
8 BrandE 1998-99 Duke Soph 0.641 0.619 0.390 0.347 99.843 (NPOY)
9 GrantH 1986-87 Clem Sr 0.668 0.703 0.294 0.328 99.658 (ACCPOY)
10 SmithJ 1994-95 MD Soph 0.627 0.644 0.404 0.312 99.341 (NPOY)
11 WilliamsS 2005-06 Duke Sr 0.621 0.556 0.415 0.370 98.082
12 LaettnerC 1990-91 Duke Jr 0.640 0.681 0.327 0.310 97.885
13 BattierS 2000-01 Duke Sr 0.619 0.632 0.305 0.392 97.431 (NPOY)
14 DuncanT 1995-96 WF Jr 0.589 0.580 0.397 0.379 97.268 (ACCPOY)
15 ScottD 1989-90 GT Jr 0.589 0.747 0.208 0.391 96.724 (ACCPOY)
16 DaughertyB 1985-86 UNC Sr 0.654 0.672 0.278 0.329 96.663
17 JamisonA 1997-98 UNC Jr 0.601 0.669 0.299 0.361 96.511 (NPOY)
18 WilliamsJ 2001-02 Duke Jr 0.575 0.842 0.137 0.372 96.279 (NPOY)
19 BoozerC 2001-02 Duke Jr 0.688 0.654 0.265 0.314 96.027
20 PerkinsS 1983-84 UNC Sr 0.647 0.569 0.321 0.375 95.573

FWIW, Redick's senior year is 21st on this list. Steve Francis' year is 24th, Juan Dixon's senior season 26th, Michael Jordan's junior year 29th, and Hurley's senior year 33rd.

It's fun to break out when you get into those "Is Josh McRoberts better than Cherokee Parks?" arguments that are impossible to answer...

[krupa-]

The NFL QB Rating is a system that is constantly used as a determination of a quarterback's talent on ESPN and FOX and the other networks, and it is a perfect example of a results-oriented statistical system.

[Kneel B4 Zod]

[ QUOTE ]
The NFL QB Rating is a system that is constantly used as a determination of a quarterback's talent on ESPN and FOX and the other networks, and it is a perfect example of a results-oriented statistical system.

[/ QUOTE ]

except it actually doesn't measure results that well.

Situation 1: 3rd and 10, QB makes a completion of 8 yards
Situation 2: 3rd and 3, QB makes a completion of 4 yards

Situation 1 would give a QB a better rating than Situation 2, even though the play was a failure, and Situation 2 was a success. even some completed passes for negative yards can boost a QB rating.

another example I found:

Imagine two quarterbacks -- Super Joe and Broadway Joe -- who both drive their teams 30 yards to a touchdown in three plays. Super Joe does it with three 10-yard passes. His completion percentage is 100, and for the drive his rating is 147.9. Broadway Joe throws two incomplete passes, then on a clutch third and long he finds a receiver in the end zone -- touchdown! For the exact same result, his rating is 111.1.

it penalizes guys who throw long, overly credits short passes, and has too big a penalty for INT's. it doesn't include sacks or rushing.

basically, what's wrong with it is the same thing wrong with many other football stat's: it doesn't take into account the situation.

[Dids]

Given Moneyball credit for the stats revolution is pretty incorrect. It's more a combination of Bill James, Rob Neyer, Fantasty Baseall, The Internet and publications like Baseball Prospectus. Moneyball was written after a lot of this stuff got off the ground.

[Kneel B4 Zod]

[ QUOTE ]
Given Moneyball credit for the stats revolution is pretty incorrect. It's more a combination of Bill James, Rob Neyer, Fantasty Baseall, The Internet and publications like Baseball Prospectus. Moneyball was written after a lot of this stuff got off the ground.

[/ QUOTE ]

here is short timeline of the above events (and then some)

# 1940's and 1950's - Branch Rickey hires Allan Roth as team statistician.

# 1964 - Earnshaw Cook, Percentage Baseball

# 1960's-1980's - Earl Weaver builds a three-run home run offense

# 1971 - Bob Davids founds SABR (Society for American Baseball Research)

# 1970's-1989 - Bill James, working at Van Camp pork and beans as a night watchman starts writing the Baseball Abstracts (1980-1989) (coins sabermetrics)

# 1985 - John Thorn and Pete Palmer, Hidden Game of Baseball

# 1998 or so - Billy Beane promoted by A's to GM

# 2003 - Michael Lewis writes Moneyball

# 2003-present - Probably 20+ teams now have a statistical consultant on staff

http://www.sju.edu/~sforman/research...-21/node2.html