Final game of the World Series. Johnson, Mathewson, Cy Young?

[rats60]

Originally Posted by darwinbulldog

You'd want, ballpark (see what I did there), a sample of 100 games to assess how good the pitcher is likely to be in games outside of that sample. Obviously you can't get that just from World Series games, so you should look at how they do in the regular season over their career rather than how they did in the postseason over their career. The differences between pitching in the regular season and pitching in the post-season are essentially negligible, whereas the difference in predictive validity between a sample of say 6 games versus 600 games of data are enormous.

Don't believe me? Pick a number between 100 and 500. 220 let's say. Then look up how the pitcher performed in games 220 through 225 of his career. See if you can tell which numbers belong to Greg Maddux versus Jamie Moyer versus Dennis Martinez vs. Randy Johnson. You could probably match Randy Johnson with the right K total, but with W/L%, WHIP, ERA+? You can't tell much from 6 games.

I take it you studied Psychology by your blog. I studied Math and Statistics. You want to dismiss small sample size, but you can't do that. With a pitcher like Matty, his 101 WS innings are backed up by his career. Any random sample from a regular season is irrelevant to postseason. Now if you had an average player that had a great postseason, Larsen perfecto, you can dismiss as it is out of his normal range.

Statistics uses small sample sizes all the time. The bigger the better, but they don't ignore small ones, it just leads to less confidence in the result. When there is significant regular season performance to back up that sample size, it produces a higher confidence. Your flaw is your opinion that postseason games aren't significant and no different than regular season ones.

Pollsters use a small sample size of 1000 to predict an election of over 120 million. They do it very accurately. Even the last election when they missed the result of the electoral college, the result of the general election was right on as well as most individual states.

[Peter_Spaeth]

But you still have the question whether Johnson's 6 games (at the end of his career) are really representative, in order to make a comparison with Mathewson.

[Koufax32fan]

Ok, so he is post-war, but Koufax.

[clydepepper]

Pre War: have to go with Matty - Three Straight Shutouts!

Post-War only:

Based on how they pitched in the World Series (not the regular season),

I'd have to go with Bob Gibson followed by Sandy Koufax and then
Johnson - but Randy, not Walter.

Schilling and Smoltz would be runners up (and I also thought about adding
Jack Morris too).

[pitchernut]

1916-18 era Babe Ruth would be my choice.

[btcarfagno]

Pedro Martinez is my choice. Circa 2000. I'm not sure there is a close second.

Tom C

[ksfarmboy]

Talking about the three greatest pitchers ever and not a single card shown? You guys are losing it. Here's my choice.

[Sean]

Originally Posted by clydepepper

Pre War: have to go with Matty - Three Straight Shutouts!

Post-War only:

Based on how they pitched in the World Series (not the regular season),

I'd have to go with Bob Gibson followed by Sandy Koufax and then
Johnson - but Randy, not Walter.

Schilling and Smoltz would be runners up (and I also thought about adding
Jack Morris too).

How is it that no one thinks of Bumgarner among the best modern pitchers?

[darwinbulldog]

Originally Posted by rats60

I take it you studied Psychology by your blog. I studied Math and Statistics. You want to dismiss small sample size, but you can't do that. With a pitcher like Matty, his 101 WS innings are backed up by his career. Any random sample from a regular season is irrelevant to postseason. Now if you had an average player that had a great postseason, Larsen perfecto, you can dismiss as it is out of his normal range.

Statistics uses small sample sizes all the time. The bigger the better, but they don't ignore small ones, it just leads to less confidence in the result. When there is significant regular season performance to back up that sample size, it produces a higher confidence. Your flaw is your opinion that postseason games aren't significant and no different than regular season ones.

Pollsters use a small sample size of 1000 to predict an election of over 120 million. They do it very accurately. Even the last election when they missed the result of the electoral college, the result of the general election was right on as well as most individual states.

Some of that is correct.

[lug-nut]

And you can't compare pitchers of today with these guys. Too many variables. these guys get 5 days rest and with the exception of a handful of guys, no one ever goes 8 innings, let alone 9...todays pitchers learned from Matty, Young and Johnson...different era, different pitchers, no comparison. Same goes for hitters. It irritates me when someone compares Aaron, Bonds or ARod to Cobb, Gehrig or Ruth... If Ruth didnt pitch for the first 5 years, he'd have 900 homeruns...and Aaron would still have 755