Measuring sports performance

Yesterday I learned the following:

1. Team payroll and team wins are not strongly correlated in the major U.S. sports.

2. Labor disputes and lock-outs do not have a long-run negative effect on attendance and receipts.

3. Problems of competitive balance come from the distribution of playing talent, and sports leagues do not much remedy the problem by salary caps and the like.

4. In the NBA, team wins attract crowds more than does star power.

5. The great NBA players do less to make their teammates better than is often supposed; in fact many great players make their teammates worse.

6. In statistical terms, the better players do not play much better, if at all, during the NBA playoffs.  (TC: But for sure they try harder, especially on defense.)

7. NBA decision-makers do not seem to understand the value of players.  In particular they tend to overvalue scoring.

All of that is from the new The Wages of Wins: Taking Measure of the Many Myths in Modern Sport, by David Berri, Martin Schmidt, and Stacey Brook.  Here is the book’s home page.  Here is the book’s blog.  Here are my early season NBA predictions.

Addendum: Here is Malcolm Gladwell’s review.


"In statistical terms, the better players do not play much better, if at all, during the NBA playoffs."

Most players' per game averages are very similar for their regular season careers and for their playoff careers. What probably happens is that the quality of competition goes up enough during the playoffs to cancel out any statistical advantage they would get from playing harder or from being in the game for more minutes.

My impression is that, in contrast, famous baseball players' statistics tend to be worse in the World Series than in the regular seasons when they got to the World Series. This would be due to more competition, which is hard to compensate for by more effort. Since baseball doesn't require as much effort as basketball, players typically give a relatively high degree of effort most of the time.

By the looks of his blog this fellow seems to far more confident than knowledgeable, particularly about the NFL. A look at 2 of his posts:

1. A post on parity in the NFL:
The idea behind this is "look[ing] at competitive balance with the Noll-Scully measure; which is simply the ratio of the actual standard deviation of wins or standing points to the standard deviation that would exist if a league consisted of equally competitive teams." He purports to use this to show that the NFL is no more equally competitive than the European soccer leagues. But the goal of NFL parity has never been to create a league of 8-8 teams - that would be wallowing in dull mediocrity. The idea behind parity is that any given year any team might emerge as an unheralded champion, while a former winner can fall into ignominy. Examples of the former include last season's Bears and the season before's Chargers; the latter is best illustrated by the sad state of the 49ers today. Now, whether or not the NFL league has more of that type of parity than any other sports league I don't know. But these folks completely misunderstand the argument and consequently miss the mark with their "debunking," and look like arrogant fools in the process.

2. His post detailing some of his new stats:

I can't comment on his NBA metrics, since I don't follow basketball. But his QB Score metric (Yards - 3 X Plays – 50 X Turnovers) is honestly worthless. A player-rating is meant to measure the value of a player in a way that is at least nominally independent from the performance of the team as a whole. Your measure doesn't even attempt to seperate the QB from the rest of the team - you don't even seperate passing yards (which the QB has SOME responsibility for, though much still depends on his recievers, his O-line, and the defense) from rushing yards, which have almost nothing to do with the QB. And it isn't exactly the QB's fault when his running back fumbles.

For real football metrics, go to Their stats are great at seperating out the value an individual (skill-position) player creates, and at creating team ratings that accurately _predict_ performance, instead of just reflecting stats back. Their off-season prediction models predicted the resurgence of Tampa Bay last season and the emergence of San Diego the season before. In other words, their site is more than just a gimmick by an economist who hasn't bothered to do any background research on NFL football.

Frankly, I'm not impressed by the book, judging by the review. Iverson, the two-hundred-and-twenty-seventh-best player in the league in 03-04? Sorry, it doesn't pass the laugh test. Basketball statsheads have thought of Iverson as overrated for a long time, but that Win Score number is simply absurd. It's not very surprising since one of the authors once claimed that Rodman was the 97-98 MVP, but please!!. For a good book on how many wins does an NBA player generates, check "Basketball on Paper" by Dean Oliver.

there's likely something wrong with this analysis, but i have trouble imagining iverson isn't great (far better than 33rd in the league at best) because: iverson took his team to the finals (and a game off the mighty lakers). the NBA playoffs let everyone in, and then force you to win a seven game series to move on (first round likely 5 games when sixers went to finals). so going far in the playoffs in the NBA means that your team really is better than the other teams. even assuming the east was really weak, who else on that philly team was responsible for philly getting to the finals? was the D really that good? it seems almost certain that you take iverson off that team and they go nowhere. if being almost singlehandedly responsible for taking your team to the finals don't correlate with being a great NBA player, what could? put another way, i think there's likely some real problems with their model.


I think for the quarterback score, they include only passing plays and QB rushes. Running plays aren't included unless the QB is the runner.

Comments for this post are closed