Stat Abuse

[wyvern37]

A lot of baseball fans are really stuck on using either OPS or Runs Created to compare players. (Although, ironically, also skeptical of new "all-in-one" stats like Win Shares.) Many people believe that these conceptual stats are the definitive answer in rating a player. But consider the following two players (from the 2001 season):

Player A had an .838 OPS which was 27% better than the league. This would be an average year for Mark Grace or Keith Hernandez.

Player B had an .930 OPS which was 42% better than the league.
This would be an average year from Harry Heilmann or Hank Aaron.

Looking at this, Player B looks a lot better than Player A. Of course, you all know that I am leading you on, so here is some more information:

Player A created 119 Runs in 157 Games. Player B created 107 Runs in 146 Games. (Player A also has the better rate .757/G versus .732/G.)

Most of you (I think) would now say that Player A is a better hitter than Player B. And maybe that's true. I don't know. If you are wondering, Player A is Ichiro Suzuki and Player B is Fred McGriff.

The point I wanted to make is that conceptual stats are tools to understand a player's performance but they are not facts. When we look at OPS, it seems a fact that McGriff was better. And when we look at Run Created, the reverse seems true. But Suzuki's 119 Runs Created is only evidence of itself. It does not prove that Suzuki would have created 119 runs in Tampa. And it does not even prove that he created 119 runs in Seattle. (In reality, he contributed to 188 runs scoring.) All it proves is that there is a conceptual world in which Suzuki created 119 Runs.

[satchel]

I may be about to restate the point you were making. To me, your comments highlight that the holy grail quest for a single stat that summarizes all a player's abilities is pretty much a fool's errand. You always have to look at more than one number. Players have different roles to play and a single number can never tell you, where you are comparing players with different roles (like McGriff and Ichiro) which one is better for his team.

I say this as a person who loves stats as much as any other fan. I haven't read Bill James's Win Shares book yet but I am skeptical of the method because all his career James has been looking for stats that are independent of what the rest of the ballplayer's team is like. Win shares naturally depend upon how many wins the team can put together.

[Craig S.]

Originally posted by satchel
I may be about to restate the point you were making. To me, your comments highlight that the holy grail quest for a single stat that summarizes all a player's abilities is pretty much a fool's errand. You always have to look at more than one number. Players have different roles to play and a single number can never tell you, where you are comparing players with different roles (like McGriff and Ichiro) which one is better for his team.


I think stats are always going to be the most critical tool in player evaluation, but they're certainly not the only one. Things like leadership and other "intangibles" will never have ant standard to be measured by, but they're always be important.

[VNV Nation]

Originally posted by satchel
I say this as a person who loves stats as much as any other fan. I haven't read Bill James's Win Shares book yet but I am skeptical of the method because all his career James has been looking for stats that are independent of what the rest of the ballplayer's team is like. Win shares naturally depend upon how many wins the team can put together.

This is a misconception. The quality of the team does not impact an individual player's Win Shares. A player who hits .300 with 30 homers on a team that wins 50 games will have the same number of wins shares as a player on a team that wins 100 games,if the two players are equal.

[MattNYY]

Take Carlos Delgado, for example compared to Jason Giambi.

[Max Power]

To me any stat works, just as long as you:

1. Adjust it for era
2. Adjust it for home park
3. Only use it as a measure over a long period of time - like a season; and,
4. understand exactly what it measures and use it just for that purpose

[SmedIndy]

OPS and runs created are much better indicators of a players value than the "triple crown stats". If I could pick just one, its OPS, but hopefully no one here picks just ONE to evaluate a player.

[moose]

Originally posted by SmedIndy
OPS and runs created are much better indicators of a players value than the "triple crown stats". If I could pick just one, its OPS, but hopefully no one here picks just ONE to evaluate a player.
I use only the unwritten rules to evaluate players.

[duh!]

[calexpat]

Originally posted by wyvern37
The point I wanted to make is that conceptual stats are tools to understand a player's performance but they are not facts. When we look at OPS, it seems a fact that McGriff was better. And when we look at Run Created, the reverse seems true. But Suzuki's 119 Runs Created is only evidence of itself. It does not prove that Suzuki would have created 119 runs in Tampa. And it does not even prove that he created 119 runs in Seattle. (In reality, he contributed to 188 runs scoring.) All it proves is that there is a conceptual world in which Suzuki created 119 Runs.

Actually, stats are facts; it's a fact that Fred McGriff had a .930 OPS last year. Of course no existing stat is a complete measure of how good a player is, but nobody, not even Bill James, thinks that.

My problem with statistical analysis of baseball is a bit different. Usually, we use statistics to infer information about a population from a sample. But in baseball analysis, especially historical analysis, there is no underlying population. What happened is what happened. That's why I don't have much patience for arguments that dismiss facts--especially postseason facts--because they're attributable to random chance. That would be a valid argument if we were talking about samples of a population; a random property of a sample is trivial in relation to the population. But a random property of the population itself is not trivial. !Viva Bobby Thompson y Luis Gonzalez! !Viva la revolucion!

[SmedIndy]

Yes, they happened, but in such a small sample size anything can happen. Some statisticians think that even a full season of stats can't take away some of the randomness that can occur in baseball and a player needs to be viewed over two or more seasons.

See Doyle, Brian or Marquez, Gonzalo for post-season sample sizes.

[VNV Nation]

Originally posted by SmedIndy
Yes, they happened, but in such a small sample size anything can happen. Some statisticians think that even a full season of stats can't take away some of the randomness that can occur in baseball and a player needs to be viewed over two or more seasons.

See Doyle, Brian or Marquez, Gonzalo for post-season sample sizes.


Does this mean the Yankees really won the WS last year? please don't tell me that

[calexpat]

Originally posted by VNV Nation



Does this mean the Yankees really won the WS last year? please don't tell me that

Exactly.

Of course, if I took this "post-season-matters-sample-size-be-damned" argument too far, I'd have to say Mantle was better than Mays. A fate worse than death. So I won't take it too far. It depends on what question you're trying to answer.

[gyb13]

Originally posted by VNV Nation
Does this mean the Yankees really won the WS last year? please don't tell me that
someone always has to take the extreme view...:rolleyes:
the point is that the postseason stats should be used along with the rest of a player's career to evaluate him. It shouldn't be excluded or used exclusively. Doing so would constitute "stat abuse."
Some statistics tell you what happened - HR, RBI, R, K
Other statistics are tools to estimate a player's true level of ability - EqA, RC, Win Shares
If you want to argue "who played better on a given day", use your counting stats.
If you want to argue "who is the better player", use advanced sabermetrics.

[moose]

Originally posted by gyb13

someone always has to take the extreme view...:rolleyes:
the point is that the postseason stats should be used along with the rest of a player's career to evaluate him. It shouldn't be excluded or used exclusively. Doing so would constitute "stat abuse."
Some statistics tell you what happened - HR, RBI, R, K
Other statistics are tools to estimate a player's true level of ability - EqA, RC, Win Shares
If you want to argue "who played better on a given day", use your counting stats.
If you want to argue "who is the better player", use advanced sabermetrics. :hscheer: go, gyb go!

seriously, though, that says it all.**
**if you appended "and whichever method you use, mussina is great," then THAT would say it all :D

[wyvern37]

Actually, stats are facts; it's a fact that Fred McGriff had a .930 OPS last year.

Even traditional stats are flawed. Check out the latest article from Rob Neyer at http://espn.go.com/mlb/columns/neyer_rob/1388300.html. Apparently, we find new errors in the statistics all the time. The most famous one is when Cobb and Lajoie were fighting for the batting title in 1910. The title was given to Lajoie, but later research showed that Cobb actually was the winner. I don't know if the debate has ever been settled.

[SmedIndy]

But that's not abuse, that's statistical errors, and they're minor enough not to cause a great deal of change in how we look at those players.

And because baseball is loath to rewrite record books and award different batting titles (or not credit Cap Anson with 3,000 hits), we're stuck with some minor issues.

If someone unearthed play by play of old seasons, then we may re-look at some old timers, because many old records are shady, sketchy, and incomplete.