Can you say histogram.

nice histogram!

Is it somewhat hypocritical of us, in an age where we are quick to chastise athletes for putting themselves ahead of their teams, to care who wins these individual awards?

That's not a rhetorical question. I'm really curious what people think.

Not very helpful data without knowing at least how many plate appearances the two players had in each situation.

I like the histogram and I appreciate the data and work that went into this.

I gotta say, though, that the big Papi argument was always based on a Win Expectancy premise that accounts not only for breakdown in production by score, but by overall game situation including inning and base-out situation, and by number of chances in given situation.

One of the reasons I'm perfectly fine with ARod winning the MVP is that such a Win Expectancy methodology is not the only reasonable way to measure retrospective value - I think I like it best, but I've heard enough smart people disagree that I don't think it's the only way. And Papi's MVP argument is based on stats like Rhoids that use that exact methodology.

So, my feeling is that this article absolutely does not discredit that WE method and thus does not discredit Papi's MVP argument. On the other hand, it is a reminder of the alternative methodologies that look better for ARod.

And, of course, it doesn't matter what you think. David Ortiz is the greatest hero in American history.

Is it somewhat hypocritical of us, in an age where we are quick to chastise athletes for putting themselves ahead of their teams, to care who wins these individual awards?

My response would be that we're not "in an age" where athletes are chastised for such behavior. If you look at any newspaper from 1911, you'll see the same chastisements. The exaltation of the individual and the bitterness at those athletes who remind us of that exaltation has been a fundamental conflict in the response to sport for as long as I know.

Is it somewhat hypocritical of us, in an age where we are quick to chastise athletes for putting themselves ahead of their teams, to care who wins these individual awards?

That's not a rhetorical question. I'm really curious what people think.

I root for teams, and for the most part the debate over individual awards leaves me cold. It only interests me to the extent that the debate turns on questions like "what constitutes value," and when that devolves into a shouting match between microstatmongers and Red Ruffinsore sportswriters it gets old pretty quickly.

But I do tend to believe ARod when he says he'd gladly trade his two MVPs for one World Series ring---I know I would. For that matter, I'm sure he'd gladly trade his MVP for that last at bat of his producing a broken bat double down the right field line. In his place, would't you?

Yesss... histograms... that's great thanks.

The real MVP was Crosby though. (/half joking)

If you look at the data more closely, it's really pretty obvious.

I have a crackpot theory about why Ortiz did so well in Win Expectancy numbers. I started a long post, but my browser got shut down when I tried to do some hyperlinks. Basically, I theorized that he may matchup better against the type of pitchers that are used in late and close situations. I don't think that this is something that you could prove statistically (probably not enough data), but you'd have to look at scouting info.

Happy Thanksgiving everybody.

A run in a game that ends close is worth more than a run in a game that ends in a blowout, but when in the game it is scored doesn't change its value -- at least not by that much.

According to WE, a sacrifice fly in the ninth inning of a game that ends 1-0 is worth three times as much as a home run to lead off a game that ends 1-0. THREE TIMES. A FLY BALL OUT IS WORTH THREE TIMES AS MUCH AS A HOME RUN.

And the pitcher's performance is worth less in the game that is won by the SF than the one where it's ended by the HR -- despite doing the same exact thing, and pitching in a tied game the whole way rather than with a lead.

WE is a fatally flawed stat.

I'm guessing WE is a typical "clutch" stat. Useful for examining if 'clutch' really exists year to year, but not sure it's a 'value' stat in telling the true value of the player.

Has anyone ever done a study of WEA as a percent of possible WEA, so every at-bat would give you a score between 1 (doing the most possible to improve your chances, usually a home run) and -1 (K or GIDP) ? Or something like that. I realize that would remove *some* of the usefulness of WEA in evaluating what is typically called clutch, but it would solve some of the problems like the one Larry mentions, and retain some measure of clutch -- moving runners over would be better than gb outs; sac flies in "clutch" situations would still have a lot of value, and so on.

The simple version I mention above is probably not the way to optimize the sort of measurement I'm thinking of, but it'd be a start.

I thought that was how WE was measured. It compared the difference between what was expected and what really happened.

According to WE, a sacrifice fly in the ninth inning of a game that ends 1-0 is worth three times as much as a home run to lead off a game that ends 1-0. THREE TIMES. A FLY BALL OUT IS WORTH THREE TIMES AS MUCH AS A HOME RUN.

Yeah, Larry, but that sac fly "locks the state" ;). Even though a leadoff homer is better, statswise, the opposing team still has nine innings to overcome it.

They had 9 innings to overcome the sac fly. They just happened to be before it.

How does it fix the situation, Gary? ;-)

Here's my proposal for a better offesive stat:

Value each event on the Base-Out state when it happens. In this manner, a Grand Slam homer is more valuable than a solo homer, as it of course is. Then re-value all the performances of the game by the final score of the game (I'm not sure exactly how to do this, but I'm sure there's a relatively simple way that someone smarter than me can figure out).

This would devalue runs in blowouts, increase the value of runs in close games, and more accurately value each event, rather than assigning them a value on the base-out state neutral standard.

And it would, I'm certain, be a 1% improvement for 500 times the work.

Not really worth it.

Also Gary, why is the pitcher's performance worth less in the former scenario? I've yet to hear a good argument for that one.

Another problem with stuff like Rhoids: how does it account for defense? Surely not very well.

Consider: road team leading by one run, runners on second and third with two out -- batter hits a line drive to third, third baseman makes a diving catch. Now, and MGL-type stat would probably have given that a very very high percentage of being a hit, which would have scored two runs and won the game for the home team. By catching the ball the third baseman ended the game, winning it for the road team -- through WE, that play right there was worth almost an entire win.

Rhoids, I would assume, credits pitchers for outs, and debits fielders for errors -- which of course would give a huge edge to DHs, since it would be penalizing fielders rather than rewarding them. Not counting defense at all also unfairly benefits designated hitters, since it gives zero credit for defense -- and if you're going to adjust offense based on situation it's patently unfair to just add on an unadjusted defensive statistic -- especially when they're working on two different scales with a volitile link - runs saved for defense and wins added for offense.

Lots of problems.

Before anyone says that the line drive caught by the 3B should be worth .217 wins, that's the cumulative value of the play. The pitcher gave up the line drive, which is worth probably about -.7 wins, the positive value is entirely the fielder's.

Another problem with stuff like Rhoids: how does it account for defense? Surely not very well.

I don't think Rhoids takes defense into account at all. Even the Mills Brothers type stuff just used errors and nothing else. That's true.

Actually I was thinking on my way over the woods to grandma's house about the two batters you mentioned: the leadoff homer guy and the walkoff sac fly guy. If they had similar batting lines the rest of the game, except for those two events, who would come out on top with in Win Expectancy Added for the whole game? It might be the guy that hit the homerun, I honestly don't know. FWIW, there have been only two 1-0 games in the last ten World Series. Neither GWRBI was from the first or last inning, one was in the fourth (Cecil Fielder in Game 5 of 96) and one was in the eighth (Jermaine Dye in this year's Game 4). Both had identical batting lines 3-4 with a double, but Dye's PGP (Jay Bennet's stat) was 15.8 while Fielder's was 6.8.

Also Gary, why is the pitcher's performance worth less in the former scenario? I've yet to hear a good argument for that one.

He put his team in a hole to start the game off. That changes the complexion of the game.

Before anyone says that the line drive caught by the 3B should be worth .217 wins, that's the cumulative value of the play. The pitcher gave up the line drive, which is worth probably about -.7 wins, the positive value is entirely the fielder's.

This is interesting, Larry, and one of the flaws of Win Probability type metrics in their current conception. Other flaws include stuff like the following: runner on first - batter hits a single that moves the runner to third. How much credit should the batter get? The runner? Should the fielder get negative credit? What about the pitcher and/or catcher?

The play that you mention sounds like one where it is easy to assign negative credit to the pitcher and positive credit to the fielder, but what about those tweener plays that aren't routine, but not Web Gems quality?

Value each event on the Base-Out state when it happens. In this manner, a Grand Slam homer is more valuable than a solo homer, as it of course is. Then re-value all the performances of the game by the final score of the game (I'm not sure exactly how to do this, but I'm sure there's a relatively simple way that someone smarter than me can figure out).

This sounds like something that Bill Janes was trying to do in the 80's when he came up with Victory Important RBIs. I thought that it was a step in the right direction, but that you may also want to count Victory Important Runs as well. That eventually led me towards the path towards win probability.

Sorry for the rambling, but I think that I am on the cusp of a turkey coma.

BTW, just so you don't think that I view the world entirely the Red Sox colored glasses, I am perfectly fine with Rodriguez winning the MVP. My attitude towards awards is actually fairly similar to Andy's anyways.

If they had similar batting lines the rest of the game, except for those two events, who would come out on top with in Win Expectancy Added for the whole game? It might be the guy that hit the homerun, I honestly don't know.

I checked -- the guy who hit the sac fly.

He put his team in a hole to start the game off. That changes the complexion of the game.

No, you're looking at the wrong guy.

The guy who pitched the shutout. In both scenarios.

His performance leading 1-0 is more valuable than the performance tied.

Makes NO sense.

Larry, I found the article on your blog that you are referencing in this thread (I can't seem to link it without shutting this browser down, though).

But if you are indicting a metric because it doesn't work in an extreme example, why not indict Runs Created and Linear Weights, too? If I read Chapter 8 of Curve Ball correctly, neither method works as well once you get beyond the normal range of player performance.

Anyways, FWIW, Keith Woolner may have come up with a better Win Probability model in the '05 BPro annual.

I realize that it is a holiday week, but I'm disappointed that no one responded to my #9 post. Tell my that I'm off my rocker. Tell me that I may be on to something.

I don't know GGC. Are you saying Ortiz hits better against better pitchers? Is it the types of pitches that closers throw (implying that more of them are more "fastball" pitchers than starters)?

How does his hitting break down against the quality of pitchers? Not sure if there's really anything there, but there may be something to the types of pitches he hits.

I agree with Larry, regarding the HR/SF thing. I think WE is absolutely useless for hitters.

The only time any kind of leverage index is useful is when you can 'spot' the player to take advantage of high leverage situations. That's why it makes sense to use it for relievers when figuring out how much retrospective value they had. Or for pinch-hitters. But ARod's 3rd inning HR is absolutely as valuable as Ortiz' 11th inning HR and the leverage doesn't matter, unless you can get Ortiz all 5 of of his PA in the 8th and 9th innings in a particular game that you know is tied at that point.

That being said, Mike Emeigh's idea of comparing how player do in close games vs. blowouts has merit, although you need to some how account for games where the player in question's performace causes the blowout. Like if ARod has 9-RBI in a 9-4 win, that should still count for something.

# I don't know GGC. Are you saying Ortiz hits better against better pitchers? Is it the types of pitches that closers throw (implying that more of them are more "fastball" pitchers than starters)?

How does his hitting break down against the quality of pitchers? Not sure if there's really anything there, but there may be something to the types of pitches he hits.

I looked at a scouting report here:

http://www.boston.com/sports/baseball/redsox/graphics/scouting_ortiz/

Basically, it says that he's a fastball hitter and doesn't do as well against offspeed or breaking stuff. Plus, most of the closers are RHP. I just opined that more of the late and close pitchers (closers and other elite relievers) might be the type of pitchers that he does well against; more apt to throw heat, et cetera.

I think you may have something there, GGC. I just wonder how significant (or predictive) it really is.

Value each event on the Base-Out state when it happens. In this manner, a Grand Slam homer is more valuable than a solo homer, as it of course is. Then re-value all the performances of the game by the final score of the game (I'm not sure exactly how to do this, but I'm sure there's a relatively simple way that someone smarter than me can figure out).


That being said, Mike Emeigh's idea of comparing how player do in close games vs. blowouts has merit, although you need to some how account for games where the player in question's performace causes the blowout. Like if ARod has 9-RBI in a 9-4 win, that should still count for something.

Joe and Larry, I'm not completely sold on a method like this being more accurate than WPA (or whatever you want to call it). However, as I ponder this more, I believe that it may be, in fact, a better method. It looks alot easier to calculate (splitting out stats by game type) and isn't probably much less of an accurate estimation a player value than WPA, and may be a better estimation than straight Extrapolated Runs, Batter Runs, VORP, et cetera.

I can be stubborn sometimes, but I do try to keep an open mind.

Larry,

I proposed something like this a while ago. From PBP data, replace all the players' PA with a random event from a league average distribution -- or other baseline distribution -- adjusted by quality of pitcher and ballpark and whatever else you want.

Calculate the total wins for the random doppleganger's team (when the number of PA in an inning changes, additional batters bat using their pitcher- and ballpark-adjusted seasonal stats and the next inning always restarts in the same place). Do this 100 times, take the average and subtract from the team's actual wins, and you've got a damn good wins-added metric.

The key is -- all the runs that a team scores in a game are worth approximately the same amount. I believe this strongly -- a 3-run HR that puts the home team up 3-0 in the first inning is just as valuable in retrospect as the walk off 3-HR in the ninth. Just because you couldn't know it at the time doesn't mean that it wasn't. We don't always know when the critical moments are going to be, even when they're happening. That's part of the beauty of the game.

There are some problems, namely how to deal with pitching changes. I say, use only the actual pitching changes that occurred, and any additional batters hit off the pitcher who finished the inning. Baserunner advancement for the simulated seasons is also an issue, but I think you can randomize it based on seasonal numbers and it'll wash out in a large sample.

In other words, the stat is:

Wins Added = Team Wins - E(Team Wins if he was an average player | performance of rest of team)