Monday, August 31, 2020
Sunday, August 30, 2020
Saturday, August 29, 2020
Friday, August 21, 2020
One year wonder
https://www.baseball-reference.com/players/w/wittge01.shtml
9-2, 1.61 ERA in 1958. 2-14 for rest of career.
Thursday, August 20, 2020
Tuesday, August 18, 2020
What in the world?
A pedestrian group of relievers can undermine a playoff-worthy team's blueprint for reaching the postseason, while a few gifted arms can elevate a mediocre team to championship-caliber.
from ESPN article
from ESPN article
Monday, August 17, 2020
Sunday, August 16, 2020
Uncanny
markdiane34
Christian Yelich has an MVP type season for the Brewers
Christian Yelich has an MVP type season for the Brewers
BJO comments from 2018. Top one is Dave Fleming replying on his 10 bold predictions
Tough-luck pitchers
https://www.baseball-reference.com/players/g/garvine01.shtml
124 ERA+, 58-97 career win-loss.
108 ERA+, 73-113
112 ERA+, 40-68
112 ERA+, 129-157
102 ERA+, 86-119
Saturday, August 15, 2020
There are four things that we are dealing with here: there are SKILLS, there is PRODUCTION, there is RANDOMNESS, and there are WINS. That which creates sabermetrics is the realization that skills lead predictably to production, and production leads predictably to wins, albeit with some variation based on randomness. In the pre-sabermetric baseball world, people understood in a vague way that good numbers sometimes led to success, but they had no concept of how that process worked. The fact that people did not understand what the predictive value of each skill set was allowed sportswriters and baseball people to (a) greatly underestimate the value of specific skills, (b) greatly exaggerate the effects of randomness, and (c) invent cause-and-effect relationships which simply did not exist, and use them to explain the success or failure of each team. There were generally accepted beliefs about baseball which were wildly at odds with the actual relationships—such as "baseball is 90% pitching", to pick the easiest one, but actually there were hundreds of them. They generally believed that a walk was something that a pitcher did, and that the batter was in most cases an accidental recipient of the pitcher’s failure, or that the pitcher walked the batter because he was reluctant to pitch to him. They significantly over-valued batting average, somewhat under-valued power, and tremendously under-valued drawing a walk. They greatly over-valued speed. In 1976 the Oakland A’s stole 341 bases. Their manager, Chuck Tanner, said. . .not an exact quote, but close. . . "we were next-to-last in the league in batting average, but fifth in runs scored. Why do you think that is?"
He meant that it was because of the base stealing. It was, in fact, mostly because, with Gene Tenace, Sal Bando and Bill North, who all walked a lot, they had led the league in walks, by a wide margin. They were fifth in the league in on base percentage and fourth in homers, so finishing 5th in the league in runs scored isn’t really a surprise. The base stealing probably helped a little, probably about 23 runs, but since they were 67 runs ahead of the team which was 6th in the league in runs scored, they almost certainly would have been fifth in the league in runs scored even if they hadn’t stolen a base all year.
You can’t fault Tanner for saying that, because you don’t know what you don’t know. All of us carry around a thousand misunderstandings of the world around us. But this misunderstanding of the offense in effect transferred credit for the A’s relatively good offense from Gene Tenace, who had a .373 on base percentage and hit 22 homers, to Don Baylor, who had a .329 on base percentage and hit only 15 homers that season, but who stole 52 bases, and to Phil Garner, who had a .307 on base percentage with 8 homers, but stole 35 bases.
The BIGGEST misunderstanding, actually, was timing. Baseball people in that era deeply believed that what mattered most in an offense was coming through in the clutch, getting your hits when a hit was needed, and that some players had an ability to do this, and other players did not. NO ONE actually believes that anymore, no one. Of course, some people still believe in a clutch ability, and that is fine; there may in fact be a clutch ability. But no one believes in it NOW the way EVERYBODY believed in it then. They believed that it was the central skill that separated winners from losers. They believed that having a championship team depended absolutely on it, and that therefore the key to winning was to acquire the players who had it. No one believes that anymore, just as no one believes anymore that Venus is inhabited by gorgeous women or that dreams are messages from the future.
Sabermetrics united the PERFORMANCE of the players with the WINS of the team in an organized and systematic way. But the essential problem is that, with our present understanding of the relationship between skills and wins, our ability to predict the number of wins resulting from a certain set of skills is limited and imperfect. We would sometimes predict, based on a team’s home runs and walks and double plays, etc., that a team might win 90 games, but in fact they might win 96, or 84. Team A might have an expectation of 90 wins but might win 96, while Team B might have an expectation of 98 wins but might win 95. The key question, then, is "what do you do with that that disparity?" That is the central question that drives sabermetrics forward, in 1976 and now: How do you explain what our present understanding of the relationship between skills and winning does NOT explain? What do we need to add to our current understanding in order to make sense of this?
There are two camps. One camp says, in essence, that everything we cannot explain is merely random, and therefore of no real significance. It should be ignored; players should continue to be evaluated based on our present understanding of the relationships of production of wins. How they can fail to perceive that this is defeatist attitude which obstructs the development of understanding puzzles the hell out of me, but. . . that is probably the dominant attitude.
The other camp—which we will call the winner’s camp—says that "there are values there that we have not measured, or, having measured them, have not incorporated into our evaluative systems. If a team wins 96 games when we would expect them to win 90, they have done so because they have done something well that we have not included in our prediction system. What is it?" That is what I said to the Chuck Tanner baseball world in 1976; that is what I say now.
It is not difficult to imagine what the answers are. If a team which should have won 90 games wins 96, it may be that they have homered more often with men on base than with the bases empty. It may be that they have played well in the late innings of one-run games. It may be that they are better than other teams at going first-to-third on a single, or that they are better than other teams at not getting a runner thrown out on the bases in a meaningful situation.
The Loser’s camp says "Well, it’s all just random; we don’t care what it is." The Winner’s camp says "If it contributed to the wins, then it matters." The reality is that some of it IS random, and some of it is unmeasured skills. It’s a mix.
But my attitude is, "If it happened, it matters." I don’t care if it was random; that doesn’t invalidate it. It only invalidates it for predictions. For understanding the past, if it happened, it matters.
OK, so you’ve got this bull-in-an-analytical china shop notion about ignoring what happens in the games that were lost, and focusing only on the games that were won. The essential problem with that is that it separates SKILLS from PRODUCTION, and, by so doing, greatly amplifies the role of RANDOMNESS. Does that make sense? For a team that went 60-102, you would be basing your assessments on only 60 games. On average it would mean basing your evaluations on one-half of the games. This makes the assessment of skills much less accurate, thus inevitably increasing the role of randomness in the predictions.
The thing is, we don’t NEED to do that. The systems we have predict wins from production with a high degree of reliability. What we want to do is to INCREASE that predictive connection. In doing that, we hold on both to wins, and to skills. We hold on to both sides of the rope bridge. Throwing away half of the data would not lead to a BETTER understanding of the long bridge between skills and wins; it would quite certainly lead to less understanding. That, at least, would be my expectation.
8-14-20 Bill James
Friday, August 14, 2020
Thursday, August 13, 2020
Saturday, August 8, 2020
Monday, August 3, 2020
Flaw in WAR
https://www.baseball-reference.com/players/b/beschbo01.shtml
Because caught stealing totals were so high then, Bob Bescher's base running RAA was much higher than when they did have data, because he stole so many bases.
Saturday, August 1, 2020
Subscribe to:
Posts (Atom)
3:41 PM Mar 22nd