Comments on: Worst Opening Day performance ever? Not quite. This and that about baseball stats. Tue, 16 Jul 2013 17:01:55 +0000 hourly 1 By: Whiz Mon, 04 Apr 2011 16:28:08 +0000 Neil, you should mention that this is for 1950-2011, since it requires p-b-p data.

Pujols' opening day was the 263rd worst by RE24 overall (not just for opening day games). He has the 813th, 871st and 918th best RE24 in a game. Aramis Ramirez has the most RE24 games above 4.9 with 7 (I took 4.9 to compare with Pujols, who had 3 games over that).

RE24 also has context; if the bases are empty, the most you can increase RE is 1.000. Not surprisingly, Mark Whiten's 4-HR, 12-RBI game in 1993 has the best RE24, 9.495.
Mike Cameron hit 4 solo HR in a game in 2002 and had only 4.064 RE24 (the extra .064 came from a HBP minus an out). From a linear weights standpoint, Cameron had the better game because of the HBP (they both went 4 for 5).

By: Neil Paine Mon, 04 Apr 2011 13:47:46 +0000 Wow, I didn't expect this to be so controversial.

Again, it all depends on your definition of "worst". Here, I'm defining worst as: "players who decreased their teams' chances of winning the most within the context of each specific game".

If you think that clutch considerations like leverage index should not be taken into account, you can re-run the above list but sort by RE24, which shows how much a player decreased his team's run expectancy without considering how "crucial" the situation was. That's another way to define "worst". There's no right or wrong answer, it just depends on whether you're focusing on the value of each performance in a vacuum, or within the context of the actual game itself.

By: Jim Dunne Mon, 04 Apr 2011 13:38:47 +0000 @19

I agree with everything you said, with the exception of "If someone uses WPA in a way that distorts its meaning, that doesn't devalue the stat itself." While the statistic itself may remain useful in its own way, misuse of that statistic will result in people mistrusting it and stop using it altogether. Think of it this way - is RBI a meaningless stat? I would say no - it is a very simple function of how many runs a batter drives in, and how many opportunities he had to drive them in. However, it was misused for decades as a way to determine what player was "best," when it isn't. Now, anyone who quotes RBI in an argument gets a funny look from anyone in the statistical analysis community. It's not the meaninglessness of the stat, but the years of misuse, that brought us to that point.

I'd hate to see the same thing happen with WPA, which, as you said, is a useful stat for its purposes. Dusty Baker did not have a worse game than Albert Pujols. Claiming that WPA proves he did, when that's not what WPA is meant to measure, will simply turn people off of WPA.

By: Rich Mon, 04 Apr 2011 04:09:09 +0000 Dusty still did more to help his team win that day than at any point as a manager

By: John Autin Mon, 04 Apr 2011 03:54:48 +0000 @17-18 -- I, too, have criticized WPA on a number of counts. I do not think it is a particularly good measure of "good" or "bad" games in the general sense.

At the same time, I don't think it's fair to say that WPA is bogus or gets it wrong. WPA is what it is, a purely probabilistic measure of how a player's offensive events change his team's odds of winning the game; it can be useful in various aspects of analysis and discussion, as long as we don't try to make it do too much. If someone uses WPA in a way that distorts its meaning, that doesn't devalue the stat itself, any more than batting average is devalued when someone argues that one player is better than another because of a higher BA.

In the case of this blog entry, I'm not jumping all over Neil. But I do think it would have been better to call it, "Worst Opening Day WPA ever?"

By: Blake Mon, 04 Apr 2011 03:03:46 +0000 Take a look at the play-by-play of the game Dusty Baker had that WPA calls the "worst opening day ever." He was having an ordinary 1-for-4 game, nothing particularly significant, until he came up in the bottom of the 9th with the bases loaded and the A's down by a run. He batted into a game-ending double play. That must have been crushing, but it was his only big negative; he played DH so he made no fielding miscues, and he made no base-running mistakes. So was a game-ending double play even the worst opening day at-bat ever?

Until now I never realized how bogus WPA is.

By: Jim Dunne Sun, 03 Apr 2011 20:13:05 +0000 Neil,
You hit the nail on the head. "Value within the context of the game" can be misleasding, so, in my mind, it's a very poor way to measure best and worst. I mean, if Player A gets hit by a pitch with the bases loaded in the bottom of the ninth, one could easily argue that he has more "value within the context of the game" than a guy who hit three homers in a game that his team loses 10-5. But did the guy who gets hit by the pitch have a "better" game? That's awfully hard to defend.

It's possible that sabermetricians don't do enough to consider the importance of game context. By the opposite token, I feel that considering WPA to determine "best" or "worst" on the basis of a single game set us back to the days of sports reporting where players were considered "clutch" or "goats" simply on the basis of what they did in high-leverage situations.

I hope I'm not coming off as condescending here - I just think it's important to have a discussion on what our interpretations of "best," "worst" and "value" are. My view is that WPA gets it wrong.

By: John Autin Sun, 03 Apr 2011 19:48:33 +0000 @11, Dave V. -- Tartabull came up through the minors as an infielder, mostly at 2B. Although he was error-prone and seemed not long for that job, early rotisserie players salivated.

BTW, check out the 1986 AL rookie class: Tartabull had 25 HRs, 96 RBI and a .270 BA -- and placed 5th in the ROY vote, behind Canseco (33-117-.240), Joyner (22-100-.290), reliever Mark Eichhorn (14-6, 1.72 ERA in 157 IP, with 10 saves), and Cory Snyder (24-69-.272 as an OF/SS).

By: Neil Paine Sun, 03 Apr 2011 19:34:31 +0000 Well, that's an argument about the difference between performance in a vacuum and value within the context of a game. Depending on which definition you use, you're going to get different results. Like #4 said, if you want to ignore the leverage aspect, you can sort by RE24.

By: Jim Dunne Sun, 03 Apr 2011 17:53:57 +0000 @10

Exactly. WPA really doesn't have anything to do with the player who had the worst game, it has more to do with how that performance lined up with what the rest of the team was doing. If you ground into a double play with your team losing 2-1 or winning 13-2, it's still just as bad of a play individually. However, the WPA in the 2-1 game will be significant, while in the 13-2 game it is negligible.

There was a post a few weeks back about best birthday performances, and I was surprised to not see Nomar's 3-homer game against Tampa in 2002 on there. I then remembered that the Red Sox ended up winning that game something like 22-5, compromising the quality of his performance in WPAs eyes.

WPA can be a useful tool to measure trends over a long period of time - for example, to measure a claim that certain players will be more "clutch" than their statistics would otherwise show. It can also be useful when analyzing a game, both real time and in hindsight, to identify critical points/plays. However, what WPA is not useful for is comparing games of two different players, because it is dependent not just on outcome but on situation.

I mean, just look at the results this generated. Which is a worse individual performance, going 1 for 5 with a GIDP, or 0 for 5 with 3 GIDP? According to WPA, it is the former. I don't need a complex linear weights formula to tell me that's wrong.