from Amazon
Saturday, November 21, 2009
Suppose you had a game-by-game Cy Young award
And it was Clemens pitching this game (20K, 0 BB, 1 HR):
http://www.baseball-reference.com/boxes/BOS/BOS198604290.shtml
Against Murray Dickson (shutout, with 15 baserunners, including 4 doubles):
http://www.baseball-reference.com/boxes/CHN/CHN195406091.shtml
Clemens gets the loss of course. Suppose you had alot of those games, maybe not so extreme, but lopsided at least. In the NHL, they have a “three stars of the game” award, which goes back several decades. It’s one of the great traditions of the game. In their case, they’d likely give one star to Clemens (definitely), one to the guy who hit the HR, and MAYBE to Dickson. It could probably go to a fielder on Dickson’s team. It could go to whoever might have gone 3-4 even though Dickson shut him out. And really, that’s what you would do, isn’t it? You’re not going to rely on the fact that Dickson gave up no runs in the game, are you? That’s because you know that it wasn’t just Dickson, but his team with Dickson on the mound.
What’s the point of looking at extreme games like this? It’s to make you think in terms of highlighting the strengths and weakness of a system. It is NOT designed to VALIDATE the system. There’s no specific reason that a system must work at data points outside what the system will ever get. You validate for Barry Bonds, but you don’t need to validate for Superman.
The system therefore should be able to capture the reality of the game. You need some model to represent that reality. W/L records don’t do that. ERA likely doesn’t do that. WPA probably won’t do that. FIP might do that. WPA/LI might very well do that.
And then problem comes in when you lose the quality of the one game, and start aggregating things at the seasonal level. Then, all of the little things that happened in one game gets swept away. All of a sudden, if Dickson give up 6 runs in 4 innings on 9 baserunners, he ends up with a 2-game total of 13 innings, 6 runs, and 24 baserunners. He stinks. But he has a 1-1 record.
So, exactly how is it that you want to evaluate your player’s contributions to winning each game? By just ignoring the context? Or, are you going to rely on the oldest standy bullsh!t nonsense of all time: “I consider everything”, which means you have no idea how to consider each part of everything: you can’t consider everything until you know how much to weight each part of everything.
To me, the answer is clear: you need some sort of “Three Stars of the Game” approach. And, just on a gut level, WPA/LI (which is nothing at all like WPA) might give you that.
Recent comments
Older comments
Page 1 of 320 pages 1 2 3 > Last »Complete Archive – By Category
Complete Archive – By Date