From: John Kasupski on 4 Mar 2010 14:41 On Thu, 4 Mar 2010 11:06:06 -0800 (PST), Ron Johnson <ron7941(a)gmail.com> wrote: >> I'm not at all surprised that RC is better than OPS for this purpose. > >Yeah but you have to understand that runs created isn't >great at the team level and flat does not work properly >at the individual level. > >A minor problem is that some provably incorrect weights >are used (just as a for instance, sac hits and sac flies >are treated as having equal value. A regression based >study will demonstrate that this is not the case) > >A major problem is that runs created is multiplicative. >This has the odd effect of asserting that a home >run hit by Frank Thomas is a fair amount more valuable >than one hit by Joe Carter. > >Don't believe this? Try adding an 1-1, home run to either >player in any given season.It turns out that runs created >asserts that a HR hit by a player with an OBP of .445 is >worth 1.78 runs while a HR hit by a player with a .306 OBP >is worth 1.22 runs. > >There is a work-around but it's extra work (Calculate >team runs created. Then remove the player's stat line and >recalculate. Credit the player with the difference. >A decade after Dave Tate came up with this Bill >James came up with it on his own.) > >And there's another big issue with runs created per out. > >runs created per out is at heart (OBP*SLG)/(1-BA) > >(Basic runs created starts out as AB*OBP*SLG. More >advanced versions have various correctors thrown in >but this still remains the core) > >While runs created per out makes sense as a concept >it doesn't work particularly well in practice. OK, apparently somebody changed the formula for runs created and forgot to insure that I received the memo. :-) What I was talking about with RC was the much simpler definition from a couple of decades ago: (Runs Scored + RBI) - HR, basically the number of runs a player scored plus the number he drove in, with HRs subtracted so that he doesn't get credited for two runs for a homer (one for the run he scores and one for the RBI he gets credited with). It would not have surprised me - and still would not - if this number turned out to be a better indicator than OPS - for an individual player. A guy who hits a double with the bases empty and then gets stranded will drive up his slugging percentage, but his team hasn't scored any runs as a result. The guy who singles with the bases loaded gets the same benefit with respec to his OBP (though not to his SLG), but has probably just driven in two runs. As for the value of those runs, I'd say that depends on the game situation. If his team is ahead or behind by ten runs, to me that's not the same as if he just broke (or created) a 2-2 tie, or if his team was one run behind when he did it. JK
From: Ron Johnson on 4 Mar 2010 15:09 On Mar 4, 2:55 pm, John Kasupski <w2...(a)spamfilter.verizon.net> wrote: > On Thu, 4 Mar 2010 09:45:07 -0800 (PST), HTP <tmbowma...(a)yahoo.com> wrote: > >We're covering specifically stats as > >a measure of a players offensive capabilities. > >I've been on record > >forever as being skeptical about the reliability of fielding stats. > >Its still a relatively young field. > > >Just because you arent aware of stats pertaining an outfielders > >throwing arm or range doesnt mean that they dont exist. > > I think you'll find that most serious sabermetricians will admit that they still > have a long way to go with respect to defense. Actually Bill James for instance said recently that we can now evaluate a season's worth of defense as accurately as a season's worth of offense. What's unclear is how well we can do projecting the future. That is to say James believes we can get a good sense of how valuable (say) Marco Scutaro was last year but we're really not sure how that will translate to a new park/ staff/coaching etc. I think that's a tad optimistic. For one thing, nobody's demonstrated that the defensive numbers properly reconcile at the team levels. From what I can tell, current numbers tell us a tad more about how good a player is defensively than their batting average tells us about their offense. But they are good enough that you should be able to point to a specific problem (EG, ZR doesn't properly account for the monster when it comes to Boston LF rating. And if the manager is fond of the shift it's going to affect infielder ratings differently depending on system design) > He does his job based on an entirely > different set of principles and, in all likelihood, he always will. > > So, to you, and to many others who are into statistical analysis, he is > therefore an idiot. No. I think he's just mailing it in. > Well, I'd suppose he'd say the same about you or me if we > got in his face with all this saber stuff. He's doing the same thing that's been > done in the game of baseball for decades upon decades. It worked then. It still > works now, that's why guys like Leyland or Torre still show up with their teams > in the playoffs every year. Well Leyland doesn't. He's below .500 for his career, has made the playoffs 5 times in 18 years and has never managed a surprise winner. I'm at a loss to see a year when you can point to his team and say without Leyland's decision to do X they don't win. And Torre didn't win for a damned long time. I mean Torre's winning percentage with the Mets was .405 In Atlanta he made the playoffs his first year and went downhill every year after that (and there are reason to assign some of the blame to his decisions). And that one year was his lone playoff appearance before going to the Yankees. (He was 109 games below .500 before going to the Yankees)
From: Ron Johnson on 4 Mar 2010 15:33 On Mar 4, 3:41 pm, John Kasupski <w2...(a)spamfilter.verizon.net> wrote: > On Thu, 4 Mar 2010 11:06:06 -0800 (PST), Ron Johnson <ron7...(a)gmail.com> wrote: > >> I'm not at all surprised that RC is better than OPS for this purpose. > > >Yeah but you have to understand that runs created isn't > >great at the team level and flat does not work properly > >at the individual level. > > >A minor problem is that some provably incorrect weights > >are used (just as a for instance, sac hits and sac flies > >are treated as having equal value. A regression based > >study will demonstrate that this is not the case) > > >A major problem is that runs created is multiplicative. > >This has the odd effect of asserting that a home > >run hit by Frank Thomas is a fair amount more valuable > >than one hit by Joe Carter. > > >Don't believe this? Try adding an 1-1, home run to either > >player in any given season.It turns out that runs created > >asserts that a HR hit by a player with an OBP of .445 is > >worth 1.78 runs while a HR hit by a player with a .306 OBP > >is worth 1.22 runs. > > >There is a work-around but it's extra work (Calculate > >team runs created. Then remove the player's stat line and > >recalculate. Credit the player with the difference. > >A decade after Dave Tate came up with this Bill > >James came up with it on his own.) > > >And there's another big issue with runs created per out. > > >runs created per out is at heart (OBP*SLG)/(1-BA) > > >(Basic runs created starts out as AB*OBP*SLG. More > >advanced versions have various correctors thrown in > >but this still remains the core) > > >While runs created per out makes sense as a concept > >it doesn't work particularly well in practice. > > OK, apparently somebody changed the formula for runs created and forgot to > insure that I received the memo. :-) What you're talking about has never been called runs created. I've seen it called runs participated in and a few other things. (I've also seen TBI, teammates batted in) > What I was talking about with RC was the much simpler definition from a couple > of decades ago: (Runs Scored + RBI) - HR, basically the number of runs a player > scored plus the number he drove in, with HRs subtracted so that he doesn't get > credited for two runs for a homer (one for the run he scores and one for the RBI > he gets credited with). Here we go again. It's half runs contributed and it's butt stupid to subtract the home runs. Think about it. Triple + Sac Fly. One run scored, One RBI Home run followed by flyball out. One run scored, one RBI Any metric that asserts that the second case is less valuable is fundamentally broken. Yes, there are runs scored without an RBI. Not a heck of a lot. > It would not have surprised me - and still would not - if this number turned out > to be a better indicator than OPS - for an individual player. A guy who hits a > double with the bases empty and then gets stranded will drive up his slugging > percentage, but his team hasn't scored any runs as a result. The guy who singles > with the bases loaded gets the same benefit with respec to his OBP (though not > to his SLG), but has probably just driven in two runs. Probably would be true if such a player existed. Feel free to name him. Situational hitting basically comes out in the wash over time. The best example that I've found is Paul Molitor. Over a 15 year period Molitor hit 32 points better with runners in scoring position than overall. Best I can tell, conventional metrics underrate Molitor by about 29 runs. If you're off at the absolute extreme by less than 2 runs a year then it's really not worth sweating. Situational stats simply aren't terribly important in team run scoring. Roughly 90% of the variation in team runs scored is explained by OBP and SLG. You just can't build a good offense without taking care of those elements. > As for the value of those runs, I'd say that depends on the game situation. If > his team is ahead or behind by ten runs, to me that's not the same as if he just > broke (or created) a 2-2 tie, or if his team was one run behind when he did it. Again, feel free to name the players who were substantially more (or less) productive in high weight situations. There is a method to check this (Called WPA or win probability added). Basically it turns out to be a great deal of work and everything turns out to pretty much come out in the wash. The most conveniently accessible article on game state analysis is Tom Ruane's at retrosheet. http://www.retrosheet.org/Research/RuaneT/valueadd_art.htm Actually the whole research papers section at retrosheet is worth a look. In particular Tom and Dave Smith's papers. http://www.retrosheet.org/Research/Research.htm
From: John Kasupski on 4 Mar 2010 16:38 On Thu, 4 Mar 2010 12:33:34 -0800 (PST), Ron Johnson <ron7941(a)gmail.com> wrote: >What you're talking about has never been called runs created. I've >seen it >called runs participated in and a few other things. You're right. What I was talking about is called Runs Produced. My bad. >Here we go again. It's half runs contributed and it's butt stupid >to subtract the home runs. > >Think about it. Triple + Sac Fly. One run scored, One RBI >Home run followed by flyball out. One run scored, one RBI The idea of Runs Produced was to measure how many runs scored as a result of a player's actions. The reason HRs are subtracted from the total of the player's runs scored and RBIs because a player who hits a home run is credited with both a run scored and an RBI on the play - as a result of the fact that he himself crosses the plate on his own homer. If he belted a solo shot, his team scores one run on the play, not two, so he is credited with one run produced, not two. JK
From: John Kasupski on 4 Mar 2010 17:20
On Thu, 4 Mar 2010 12:09:42 -0800 (PST), Ron Johnson <ron7941(a)gmail.com> wrote: >But they are good enough that you should be able >to point to a specific problem (EG, ZR doesn't >properly account for the monster when it comes >to Boston LF rating. And if the manager is >fond of the shift it's going to affect infielder >ratings differently depending on system design) That's one reason I'm not so keen on the current methods of quantifying defense. You mentioned LF in Boston - Jason bay really gets hosed by current metrics. The guy logged over 1200 innings in one of the toughest left fields in the game and played errorless ball, put up a better range and range factor than the average of all left fielders, and contributed 15 assists to boot. His reqard is that he gets a -11.2 UZR and everyone arguing about his defense in LF based on a stat that has less than a decade of data since it wasn't invented until...what, 2003? Don't you guys always talk about small sample size? Similarly, stats analysts panned the Phillies when they signed Ibanez because of his -23.5 UZR in 2007 and -10.5 in 2008, but last year he put up +10.7 (a 200% improvement). Amazing what playing next to a Gold Glove centerfielder will do. No, I still think there's a long way to go WRT defense. >Well Leyland doesn't. He's below .500 for his career, has made the >playoffs 5 times in 18 years and has never managed a surprise >winner. I'm at a loss to see a year when you can point to his >team and say without Leyland's decision to do X they don't win. > >And Torre didn't win for a damned long time. > >I mean Torre's winning percentage with the Mets was .405 > >In Atlanta he made the playoffs his first year and went downhill >every year after that (and there are reason to assign some >of the blame to his decisions). And that one year was his >lone playoff appearance before going to the Yankees. >(He was 109 games below .500 before going to the Yankees) Well...there are a lot of other considerations that I think come into play, including the players they had to work with, and how much cooperation they got from the front office, to name two. Also, on-field strategy isn't the only thing that goes into being a successful manager. You said yourself that "don't screw it up" isn't as easy and people think, and working for Steinbrenner...yet Torre didn't screw it up four times during the kate 1990s. Leyland started out with some truly awful teams in Pittsburgh that are mainly the reason he's under .500 for his career...and when he took the Tigers team that went 71-91 in 2005 to the World Series in 2006, it sure surprised me! McKeon going to Florida and taking the Marlins to a WS win, how can that be anything but a surprise winner? Bobby Cox is another old school manager. He and Torre are tied with 15 playoff appearances, more than any other manager in the history of the game. Piniella, another traditionalist, is also in the top ten along with Lasorda, Sparky, and a bunch of guys who managed in the 60's or earlier. The only guy in the bunch you can say is a stats-oriented manager is LaRussa, and he's needed more than 600 more games than Torre to get 300 more wins - and 300 more losses. He also needed more than 400 more games than Cox to about 150 more wins and over 300 more losses...his winning percentage is lower than either of the other two. I suspect that arrangement may change with time. Cox is retiring after this year and Torre may do so in a few years, but LaRussa's not done yet and Piniella's still at it. After that...Scioscia has six playoff appearances, Gardenhire and Francona each have five, all still active and currently with pretty good teams. JK |