HFBoards

Go Back   HFBoards > NHL Western Conference > Pacific Division > Edmonton Oilers
Notices

Bill James and hitter/pitcher matchups ...

Closed Thread
 
Thread Tools
Old
08-19-2004, 05:55 PM
  #1
igor*
 
Join Date: Mar 2002
Posts: 3,276
vCash: 500
Bill James and hitter/pitcher matchups ...

It's baseball I know, but this stuff has so much in common it's eery ... so if you're interesting in the moneyball-type stuff as it applies to the NHL game this still might be worth you reading this little ramble. And the general opinion of anyone who is into baseball would be appreciated.

I'm not into baseball ... but Bill James is one interesting guy. Lowetide has mentioned him enough times that I've looked him up ... and quite frankly I can't see the sense in much of what he does, though it seems to work so how can you really argue?

Anyhow, I read an interview with him a while ago, about speeding up the game ... and he really came across as a remarkably sensible guy. So I revisited some of his stuff.

Bill James credits Dallas Adams with the "log5" equation (he comes up with the most catchy, unrelated names for his stuff :lol ) for predicting the probabilities with different hitter/pitcher matchups ... be it for OBP, batting average, slugging%, whatever.

This is his equation:

(PitAvg x BatAvg)/LgAvg
__________________________________________________ ___
(PitAvg x BatAvg)/LgAvg + ((1-PitAvg) x (1-BatAvg))/(1-LgAvg)

Complicated stuff.

And this is the analogy that it is EXACTLY equal to:

* 3 guys (a pitcher, a batter, and a 'completely average' pitcher) play roulette on a special table with 100 numbers on it.
* All three take a spin.
* The batter has a batting average of .320 ... so if he gets 1 to 32 with his spin ... it counts as a win.
* The pitcher has an opposing batting average of .300 ... so if he gets 1 to 70 with his spin ... it counts as a win.
* The 'completely average' pitcher has an opposing batting average of .260 ... so if he gets 1 to 74 with his spin ... it counts as a win.
* Bill James sits in the balcony and keeps a tally of these two scenarios:
A. The rounds when the pitcher lost, the batter won, and the average pitcher won.
B. And the exact opposite ... pitcher wins, batter loses, and the average pitcher loses.

.
.
.

He notes that A/(A+B) will be the expected batting average of this picture against this batter ... and he'll be surprisingly accurate. There is no arguing that.

But how in Hedes does this even make a shred of sense? Seriously, I'm not really into baseball ... but how is that even remotely similar to a real hitter/picture matchup? Seriously ... maybe I'm missing something.

.
.
.

Wouldn't this analogy make more sense ...

* The roulette wheel normally rewards 1 to 26 with a win.
* A batter with a .320 average gets (32-26 = 6) six extra squares, courtesy of the baseball-loving casino owner So 1 to 32
* The pitcher is betting against him is kind of terrible (.300 opposing batting average) ... so the baseball loving casino-owner punishes him by giving the batter (30-26 = 4) four extra squares. So now the batter wins with 1 to 36.
* The hockey oddswriter sits in the balcony and observes that these odds (.360) should be the expected batting average of this batter vs this pitcher.

It boils down to:
= PitAvg + BatAvg - LgAvg

Uber-simple stuff. Hockey guy calls it the "Socrates7" equation ... the idea doesn't sell.

And while it is more accurate than 'log5' for the American League ... it's doesn't work for the National League.

Why? Seriously.

igor* is offline  
Old
08-19-2004, 06:27 PM
  #2
Cloned
Sexy Genesis
 
Cloned's Avatar
 
Join Date: Aug 2003
Posts: 21,852
vCash: 500
I would think the DH has a role to play in the difference between the AL and the NL.

Cloned is offline  
Old
08-19-2004, 06:34 PM
  #3
igor*
 
Join Date: Mar 2002
Posts: 3,276
vCash: 500
Quote:
Originally Posted by Cloned
I would think the DH has a role to play in the difference between the AL and the NL.
Ya, must be, that's the only real difference between the two leagues (and it is a whopper of a difference).

So why doesn't "hockey-guy's" analogy work as well for the NL ... in terms of roulette ... what's the real difference there?

igor* is offline  
Old
08-19-2004, 06:55 PM
  #4
Lowetide
Registered User
 
Lowetide's Avatar
 
Join Date: Feb 2002
Country: Canada
Posts: 13,281
vCash: 500
Whenever there is a different outcome between AL/NL, the list of things to start looking for begins with the DH. AL lineups have one more hitter per turn through the lineup, 4 a game x162 per year and that will skew the number heaviliy. The AL also has the Yankees and the BoSox, and since they are no longer concerned with competitve balance I'd suggest some of the skew comes from that too.

As for James, I have a few questions. When you list Pitching avg/hitting avg, what are you talking about? Is that straight Era versus batting average? I suspect it isnt.

What James' does is take Ted Lilly's era (3.88) and opponent's ba (.232), runs it through every other player in the league and hones it into a specific ratio.

He then takes Vlady Guerrero's slugging pct, obp and runs that through every AL hitter and hones that into a similar (opposite) number.

He THEN allows for ballpark effects (are they playing in Anaheim?) and then runs the program 100 times to see the results. In 100 plate appearances against Ted Lilly, my guess is that Vlad Guerrero would hit about .340 (higher than his .325 season average due to platoon advantage and clubbing the ball into the gap), hit about 8 home runs and drive in about 25 since the Jays are a poor team and Lilly doesn't have an out pitch Vladdy can't handle (is there a pitch Vladdy can't handle?)

The key to Bill James is that he tries to take everything into account, and then come up with a logical number. He THEN applies it over several seasons to attempt some kind of predicting tool. He's been quite successful, although as someone who first read his stuff 20 years ago he has had some moments of stupidity.

But I'll tell you something: he predicted things decades before they happened (he once said that the 1986 baseball rookie crop would hit like thunder but die with the leather and baseball people laughed at him but he was right. All the glove men couldn't hit a lick and all the bashers were Rob Deer's and Jose Canseco's).

The thing that remains the most impressive about him is his major league equivancies. He takes the minor league talent pool and puts it on an approximately level playing field (with hitters) and it is an exceptional tool. He also talked about k/w ratios long before people became obsessed (did a stunning study on k's per 9ip accross baseball history, and found the number of k's per 9ip a pitcher must manage and still be effective. Jim Abbott was below that number and he was out of baseball within a couple of years. It was kind of thrilling to watch him predict that stuff 20 years ago. He also predicted that someone like Randy Johnson could pitch well into his 40's if his k/w per 9ip was above 6 and his k/w ratio maintained a 3/1 ratio. It's all true).

igor, you should take your stuff and apply James' marketing strategy to it. There's a market for it, I swear.

Lowetide is offline  
Old
08-19-2004, 07:20 PM
  #5
Cloned
Sexy Genesis
 
Cloned's Avatar
 
Join Date: Aug 2003
Posts: 21,852
vCash: 500
Quote:
Originally Posted by igor
So why doesn't "hockey-guy's" analogy work as well for the NL ... in terms of roulette ... what's the real difference there?
The only thing I can think of besides the DH would be the fact that there is much more strategy in the NL. Managers routinely ask for sac bunts, double switches, etc. in the NL whereas in the AL most managers would rather sit back and let the batter do the talking. I would think that even though a sac bunt doesn't officially count as an at-bat, that it must have some kind of effect because you're now moving up a runner for the NEXT batter, and so on, kinda like a domino effect. The same thing with double switches... we switch players and pitchers and that must have some kind of effect.

Cloned is offline  
Old
08-19-2004, 07:57 PM
  #6
mudcrutch79
Registered User
 
Join Date: Jul 2003
Location: The Big Smoke
Posts: 3,903
vCash: 500
Quote:
Originally Posted by lowetide
igor, you should take your stuff and apply James' marketing strategy to it. There's a market for it, I swear.
Couldn't agree more. For those who aren't aware, Baseball Prospectus is sort of the modern day inheritor of the legacy of the Bill James Baseball Abstracts. I'd love to see something similar for hockey, and think that there are tons and tons of interesting things that could be looked at from a statistical point of view, creating interesting essays for all 30 NHL teams. Sadly, I'll likely never see it, but if one of the smart guys around here wants to publish it...guaranteed sale right here.

mudcrutch79 is offline  
Old
08-19-2004, 08:11 PM
  #7
igor*
 
Join Date: Mar 2002
Posts: 3,276
vCash: 500
Quote:
Originally Posted by lowetide
Whenever there is a different outcome between AL/NL, the list of things to start looking for begins with the DH. AL lineups have one more hitter per turn through the lineup, 4 a game x162 per year and that will skew the number heaviliy. The AL also has the Yankees and the BoSox, and since they are no longer concerned with competitve balance I'd suggest some of the skew comes from that too.
Does James address the NYY, BOS factor?

Quote:
As for James, I have a few questions. When you list Pitching avg/hitting avg, what are you talking about? Is that straight Era versus batting average? I suspect it isnt.
Nope. It is the batting average against the pitcher. So if hitters avg .300 against a pitcher over the course of a season ... then James defines that as a "Pitching Average" of .300. Same goes for OBP, slugging%, doubles%, walk%, etc.

Also his "average pitcher" is mythical, it's the true average. i.e. total ABs in the AL divided by total HITs in the AL.

Quote:
What James' does is take Ted Lilly's era (3.88) and opponent's ba (.232), runs it through every other player in the league and hones it into a specific ratio.

He then takes Vlady Guerrero's slugging pct, obp and runs that through every AL hitter and hones that into a similar (opposite) number.
Could you explain that further LT. If I understand you correctly then he does it the same way hockey odds (for a player or team) are calced ... but I didn't see anything that indicated James did this ... granted the stuff I did find was just terrible adaptations on the original idea ... not from James himself.

But I'll wait to see what you mean.

Quote:
He THEN allows for ballpark effects (are they playing in Anaheim?) and then runs the program 100 times to see the results. In 100 plate appearances against Ted Lilly, my guess is that Vlad Guerrero would hit about .340 (higher than his .325 season average due to platoon advantage and clubbing the ball into the gap), hit about 8 home runs and drive in about 25 since the Jays are a poor team and Lilly doesn't have an out pitch Vladdy can't handle (is there a pitch Vladdy can't handle?)
What do you mean by "runs the program 100 times to see the results" ? ... you've skipped a step on me

I take the point on ballpark effects. Not something I care about ... but its just obvious that it would really matter in predicting performance with a ball player.

Quote:
The key to Bill James is that he tries to take everything into account, and then come up with a logical number. He THEN applies it over several seasons to attempt some kind of predicting tool.
Actually I'm not sure that I agree. I think James' talent is his ability to recognize what really matters and what really doesn't. And he recognizes that patterns repeat ... and sees why ... and tries to find a model to fit it. (damn the mathematicians!! ) He always seems to be aiming for the middle as well, and I'm not so sure about that.

Quote:
But I'll tell you something: he predicted things decades before they happened (he once said that the 1986 baseball rookie crop would hit like thunder but die with the leather and baseball people laughed at him but he was right. All the glove men couldn't hit a lick and all the bashers were Rob Deer's and Jose Canseco's).

The thing that remains the most impressive about him is his major league equivancies. He takes the minor league talent pool and puts it on an approximately level playing field (with hitters) and it is an exceptional tool. He also talked about k/w ratios long before people became obsessed (did a stunning study on k's per 9ip accross baseball history, and found the number of k's per 9ip a pitcher must manage and still be effective. Jim Abbott was below that number and he was out of baseball within a couple of years. It was kind of thrilling to watch him predict that stuff 20 years ago. He also predicted that someone like Randy Johnson could pitch well into his 40's if his k/w per 9ip was above 6 and his k/w ratio maintained a 3/1 ratio. It's all true).
Again, this is sort of outside my area of interest. But I think that this is probably doable. There have been some academic runs at this with hockey ... but it's all so terribly misguided IMO ... head-banging stuff.

I think one thing is clear ... NHL scouts do a remarkable job of picking the right 18-year-old. On the whole and on the average. And on the whole ... the teams of scouts from different teams aren't really THAT much smarter than each other. But what do scouts really look for in a player that they think will translate to the NHL. Outside of the fuzzy terms and endless details about aspects of a guys game ... what do they really see.

I mean you used to always here "the guy won't be able to hit a major league curveball" or some such. But James' et al have pretty much minimalized that type of fuzzy argument.

I mean the 20-year-old AHL thing works well for reasons ... its the same league, their the same age, and you're looking at the same kind of player (scoring forward). And if you had used that ridiculously simple tool in the 90s as a GM in the NHL ... you would look like a genius right now for Christmas sakes, how whack is that? But you don't find the next John Madden that way ... and that's the kind of player that helps you win at a low cost (undervalued by the market) ... the exact type of guy Beane would be loading up on if he were a hockey GM, and the kind of guy Lamouriello has been digging up for a couple of decades.

But if you have all the event data (who was on the ice for every goal scored, and the game situation) ... for every single player in the league in question (and you would need ALL of it ... if you only had 90% of it ... useless IMO[/quote]


Quote:
igor, you should take your stuff and apply James' marketing strategy to it. There's a market for it, I swear.
I'm flattered. But I don't think so. I'm kind of in a different direction with it anyways. Besides which the NHL is funny with their data, some legal issues there. (Me posting here as the same difference as videotaping a hockey game ... as long as you're not making hundreds of copies for resale they don't care.)

Gotta go now ... but i'll get back to it.

igor* is offline  
Old
08-19-2004, 08:43 PM
  #8
igor*
 
Join Date: Mar 2002
Posts: 3,276
vCash: 500
Quote:
Originally Posted by Cloned
The only thing I can think of besides the DH would be the fact that there is much more strategy in the NL. Managers routinely ask for sac bunts, double switches, etc. in the NL whereas in the AL most managers would rather sit back and let the batter do the talking. I would think that even though a sac bunt doesn't officially count as an at-bat, that it must have some kind of effect because you're now moving up a runner for the NEXT batter, and so on, kinda like a domino effect. The same thing with double switches... we switch players and pitchers and that must have some kind of effect.
Yup, really good points. Thanks.

igor* is offline  
Old
08-19-2004, 08:54 PM
  #9
mudcrutch79
Registered User
 
Join Date: Jul 2003
Location: The Big Smoke
Posts: 3,903
vCash: 500
The point about the sort of guy that Beane or Lamoriello identifies is bang on. K-Lo has half of it down, in that the Oilers seem to have become extremely proficient at identifying defencemen who can eat minutes (LT's immortal comment about finding them in a snowstorm at midnight springs to mind), with guys like Staios, Cross and Ulanov, but he then has this seemingly overwhelming urge to overpay them, or at least pay them market value, rather than either move them, or let them go. It blows my mind that he's found something he's good at, and then essentially made this a moot point in that he signs them long term.

The same philosophy is present with the contracts given to guys like Isbister and Moreau. I'd like to go back and look, and I will someday when I have the time, but I wonder what people think Isbister can do at this point? I'd suspect around here, the expectation is still a 30 goal, 70 point ceiling. Personally, I think Lowe just locked in 3 mil over 2 years to a guy who's going to score 10-15 a year, and not really be an asset otherwise. Waste of a spot when there are guys every year who can't get deals that could provide the same thing.

It's kind of depressing to admit this, but Lowe, while being in my mind a passable hockey man, has made a series of decisions that I think are absolutely indefensible in terms of trying to build a winner in a small market. He's committed the Oilers to a guaranteed level of mediocrity.

But I digress...

mudcrutch79 is offline  
Old
08-19-2004, 09:33 PM
  #10
Lowetide
Registered User
 
Lowetide's Avatar
 
Join Date: Feb 2002
Country: Canada
Posts: 13,281
vCash: 500
Quote:
Originally Posted by igor
Does James address the NYY, BOS factor?
No, that's just my feeling about baseball. It has fallen into the wrestling/boxing realm of theatre that pretends to be sport. When a sport asks you to believe that a team with 190 million dollars worth of players playing a team with 30 million dollars worth of players is within the window of fair play they are asking too much. Baseball is no longer sport, it's George Steinbrenner's roto league.



Quote:
Originally Posted by igor
Could you explain that further LT. If I understand you correctly then he does it the same way hockey odds (for a player or team) are calced ... but I didn't see anything that indicated James did this ... granted the stuff I did find was just terrible adaptations on the original idea ... not from James himself.
Well, this is going way back into my memory banks (bouncing around in there with "One Night in Bangkok" and Peter Straub's Ghost Story) but as I recall james always started by asking a question, like this:

Who is the best hitter in the minor leagues?

He would then take hundreds of minor league seasons, say the two years before a player's rookie major league season, and start nicking the offense 40 percent. Then he'd fiddle with ballpark effects and see what was happening. He would take maybe 300 examples and ask himself if maybe the number should be 35 per cent instead of 40, that kind of thing. Eventually he landed with a system that docks a specific number from each category (takes away three doubles in 10, that sort of thing).

He would then use it as a future predictor. The last baseball book I ever purchased was his Stats Inc Minor League Handbook 2000, from which he suggested these MLE's for Nick Johnson: .319, 11hr, 72rbi as a 20 year old. I don't follow baseball anymore, but I do know that Nick Johnson just won the player of the week honors in the NL and looks like a player. James didn't just say it 4 years ago, he had a system (proven over many minor league seasons) that suggested Johnson would be a player.

If he was going forward (as you had him doing in your example in the original post) he would use the same basic rules: he'd get the info, tally it, nick for ballpark effects or the fact the guy was a fastball pitcher and this was July (fastball pitchers get better as the season wears on) and then run hundreds of baseball seasons trying that out. Once it was done, he'd fiddle some more.

MOST of Bill James theories and predictors have gone through that kind of process, or at least they did 20 years ago.

Lowetide is offline  
Old
08-19-2004, 10:43 PM
  #11
igor*
 
Join Date: Mar 2002
Posts: 3,276
vCash: 500
Quote:
Originally Posted by lowetide
No, that's just my feeling about baseball. It has fallen into the wrestling/boxing realm of theatre that pretends to be sport. When a sport asks you to believe that a team with 190 million dollars worth of players playing a team with 30 million dollars worth of players is within the window of fair play they are asking too much. Baseball is no longer sport, it's George Steinbrenner's roto league.
Good intuit Lowetide, or at least I think so. Must be all the hockey viewing making you think that way.

James' eq'n, and "hockey guy's" will get worse in recent years. And the reason is that the "average player" that everything is in reference too doesn't really work as accurately any more.

Because the pitcher from the AL East has gotten his .280 opposing batting average numbers NOT vs the mythical average .260 player ... but against an extra alotment of Yankees and BoSox ... so maybe a .270 average opposition.

And the batter he's facing from the AL central didn't get his .290 average against Opp.Avg. .260 pitchers ... but maybe .270.

So its muddied up a bit now. Just like hockey And you have to sum up every single at bat in the league to get a fair view and correct for "toughness of at-bats". And I wonder if MLB gms are taking that into account? (Nielsen was 20 years ago in hockey ... but its a heckuva lot more obvious in our game) And I wonder if players on a Team like the BlueJays don't have worse numbers than they would on a team like the Twins? I really don't know. Really don't care much either, just curious.

Quote:
Well, this is going way back into my memory banks (bouncing around in there with "One Night in Bangkok" and Peter Straub's Ghost Story) but as I recall james always started by asking a question, like this:

Who is the best hitter in the minor leagues?

He would then take hundreds of minor league seasons, say the two years before a player's rookie major league season, and start nicking the offense 40 percent. Then he'd fiddle with ballpark effects and see what was happening. He would take maybe 300 examples and ask himself if maybe the number should be 35 per cent instead of 40, that kind of thing. Eventually he landed with a system that docks a specific number from each category (takes away three doubles in 10, that sort of thing).

He would then use it as a future predictor. The last baseball book I ever purchased was his Stats Inc Minor League Handbook 2000, from which he suggested these MLE's for Nick Johnson: .319, 11hr, 72rbi as a 20 year old. I don't follow baseball anymore, but I do know that Nick Johnson just won the player of the week honors in the NL and looks like a player. James didn't just say it 4 years ago, he had a system (proven over many minor league seasons) that suggested Johnson would be a player.

If he was going forward (as you had him doing in your example in the original post) he would use the same basic rules: he'd get the info, tally it, nick for ballpark effects or the fact the guy was a fastball pitcher and this was July (fastball pitchers get better as the season wears on) and then run hundreds of baseball seasons trying that out. Once it was done, he'd fiddle some more.

MOST of Bill James theories and predictors have gone through that kind of process, or at least they did 20 years ago.
Good stuff. Makes sense.

With hockey ... (and bear in mind I am nowhere even remotely close to being a scout, I have no idea how they do what they do, I don't even watch CHL hockey or care about it) ... but with hockey I think a players ability to help his team outchance the opposition at 5on5 hockey, at any level, will probably translate to higher levels. One way or another ... speed, size, stick skills, whatever ... if they found a way at one level, they'll find a way at another. Same goes for a players ability to help his team create PP chances.

If a guy scores a lot ... it may or may not matter. If he's cheating to create that offense then either that's the way he'll always play or the offense will drop when he plays a more complete game.

I'm sure a scout can tell this just through experience. Bias might come into it too, or they might catch the kid on an off-night or when he's recovering from the flu Or maybe he has crap linemates or the coach is matching him up against the top lines of the other teams and that makes him look worse? I dunno. Probably if enough different sets of experienced eyes see the guy play enough times ... they'll have a good idea of what they are getting. And hopefully they do some sort of personality profiling on these kids too, because that has to be a huge factor as well.

The style of play in the league and the quality of team-mates and opposition would be like the "ballpark correction" you talk about ... type and quantity of minutes a guy got is like "at bats". But without icetime and especially a list of all the players who were on the ice for every goal scored in the league ... I don't think its even worth taking a run at it.

Having said all that ... I couldn't imagine anyone less qualified to comment on the whole subject of prospects and scouting than myself. Just semi-random thoughts, that's all.

igor* is offline  
Old
08-19-2004, 11:21 PM
  #12
igor*
 
Join Date: Mar 2002
Posts: 3,276
vCash: 500
Quote:
Originally Posted by mudcrutch79
...
For those who aren't aware, Baseball Prospectus is sort of the modern day inheritor of the legacy of the Bill James Baseball Abstracts.
...
Good link mudcrutch.

I'm starting to think that baseball and hockey have a lot in common. If the MLB rules changed so that pitchers and batters could both be subbed in and out really frequently in a game, some variation or other of that sort of idea ... then I think the similarities would freak people out.

At a quick run, the "shyte happens" rule seems just as big. If a rookie hits .320 in his first two years in the league, then .280 in the third year ... enough theories on why it happened will be written by fans and pundits to wrap around the world twice. And if I bothered to read any of it (and I wouldn't ) I wouldn't understand much of it anyways. But he probably really didn't get much worse that quick, it was probably just shyte happening ... and he'll probably hit somewhere around .305 the next year. They're usually unstoppable, these things. IMHO.

igor* is offline  
Old
08-20-2004, 01:11 AM
  #13
YKOil
Registered User
 
Join Date: Apr 2004
Posts: 1,232
vCash: 500
Quote:
Originally Posted by igor
At a quick run, the "shyte happens" rule seems just as big. If a rookie hits .320 in his first two years in the league, then .280 in the third year ... enough theories on why it happened will be written by fans and pundits to wrap around the world twice. And if I bothered to read any of it (and I wouldn't ) I wouldn't understand much of it anyways. But he probably really didn't get much worse that quick, it was probably just shyte happening ... and he'll probably hit somewhere around .305 the next year. They're usually unstoppable, these things. IMHO.
Not in full agreement on that one igor (though I do agree in the main, I'm just wanting to point out the exceptions) - the world of sports is full of players who have a good first year or two and then start to fall off the map - after two years most players have a 'book' on them and thus their failings (slow glove hand, poor rebound control, etc - the Jim Carrey's and David Oliver's of the world) get magnified and exploited.

Some players can overcome that through training but most who fall into this trap cannot.

On the whole I have to say that this is one of the best threads I have had the pleasure of reading. My thanks.


YKOil

YKOil is offline  
Old
08-20-2004, 12:55 PM
  #14
Big T
Registered User
 
Join Date: Sep 2003
Location: LLoydminster, AB
Posts: 338
vCash: 500
Quote:
Originally Posted by Cloned
The only thing I can think of besides the DH would be the fact that there is much more strategy in the NL. Managers routinely ask for sac bunts, double switches, etc. in the NL whereas in the AL most managers would rather sit back and let the batter do the talking. I would think that even though a sac bunt doesn't officially count as an at-bat, that it must have some kind of effect because you're now moving up a runner for the NEXT batter, and so on, kinda like a domino effect. The same thing with double switches... we switch players and pitchers and that must have some kind of effect.

I have to agree that a pitcher in the AL or the NL has his skewed. The sac bunt and the stolen base and other 'small ball' strategies affect this not because they aren't counted in batting average. (though you're right they aren't, and your overall point here is a very good one.) But these stategies actually hurt NL pitcher's that see much more of them.

The reason being is they are getting outs and yet they are not being counted in their favour. A sac bunt is so inneffective it's ridiculous that MLB managers still use the strategy. They use it because it's 'thought' to be good, not because it's a good way to produce runs. The same goes for stolen bases. A runner has to succeed something like 85% of the time in order for a stolen base to have a positive effect on runs scored.

So in effect, NL pitchers are being gifted outs because there are more managers that still use these suspect strategies, yet they are not counting in their favour. Not that these gifts should be counted in their favour, but that the number of at bats that count towards a pitchers opposing average is skewed to the more difficult batters to get out. The guys who are good batters and whom a manager is never going to ask to sac bunt.


T

Big T is offline  
Old
08-20-2004, 01:21 PM
  #15
oilswell
Registered User
 
Join Date: Mar 2002
Location: Wandering the globe
Posts: 1,760
vCash: 500
Nothing much to contribute here, but I thought I'd say something to reinforce how I think this thread is interesting, even if that something isn't useful or interesting or even correct.

Quote:
Originally Posted by lowetide
He would then take hundreds of minor league seasons, say the two years before a player's rookie major league season, and start nicking the offense 40 percent. Then he'd fiddle with ballpark effects and see what was happening. He would take maybe 300 examples and ask himself if maybe the number should be 35 per cent instead of 40, that kind of thing. Eventually he landed with a system that docks a specific number from each category (takes away three doubles in 10, that sort of thing).
Hmm, yes I suppose this is just another classic method of building predictive models from prior data. You take your historical data, split it into training and test data, choose a model type (eg linear, polynomial, etc.) with trainable parameters, and then tweak to make sure the trained model predicts well against the test data. You can see an example for predicting Olympic medal performances here.

This basic recipe is also used in machine learning/neural networks and AI, although the forms of trainable models can be radically different. From what you say, James' technique for creating these models is completely manual: he fools around with numbers and makes guesses until one of the models predicts well. He has to guess that stadium effects exist, which is reasonable but still a step. Various learning models are able to sometimes discard the noisy channels and just pick out the important inputs.

oilswell is offline  
Old
08-20-2004, 01:46 PM
  #16
igor*
 
Join Date: Mar 2002
Posts: 3,276
vCash: 500
Quote:
Originally Posted by oilswell
...
Hmm, yes I suppose this is just another classic method of building predictive models from prior data. You take your historical data, split it into training and test data, choose a model type (eg linear, polynomial, etc.) with trainable parameters, and then tweak to make sure the trained model predicts well against the test data.
...
Ya, I kind of suspected that this was the way he did it. But if it works it works ... that's the bottom line.

Explains why James' reasoning translates to hockey really well ... but his equations don't. And while he may hit the average number really well ... he's not going to get the distribution well. "Hockey guy's" roulette analogy not only pegs the average for a hitter vs pitcher matchup well ... but the how many guys will be 10 points higher than the expected against quality pitchers, or twenty points lower etc. ... for any given year ... if all of us roll a pair of dice 36 times you expect us to average rolling a "seven" 6 times each. But how many of us will roll 5 of them? or 8?

A lot of the baseball stuff is well done, and there are a lot of great ideas on that link that mudcrutch posted ... the only thing that makes me shake my head a bit is the way they are always digging through the minutae to rationalize the outliers. When really for the most part [and I haven't dug into it, just a feeling] they are probably there because they are inevitable, shyte happens.

igor* is offline  
Old
08-20-2004, 06:45 PM
  #17
Cloned
Sexy Genesis
 
Cloned's Avatar
 
Join Date: Aug 2003
Posts: 21,852
vCash: 500
Quote:
Originally Posted by Big T
I have to agree that a pitcher in the AL or the NL has his skewed. The sac bunt and the stolen base and other 'small ball' strategies affect this not because they aren't counted in batting average. (though you're right they aren't, and your overall point here is a very good one.) But these stategies actually hurt NL pitcher's that see much more of them.

The reason being is they are getting outs and yet they are not being counted in their favour. A sac bunt is so inneffective it's ridiculous that MLB managers still use the strategy. They use it because it's 'thought' to be good, not because it's a good way to produce runs. The same goes for stolen bases. A runner has to succeed something like 85% of the time in order for a stolen base to have a positive effect on runs scored.

So in effect, NL pitchers are being gifted outs because there are more managers that still use these suspect strategies, yet they are not counting in their favour. Not that these gifts should be counted in their favour, but that the number of at bats that count towards a pitchers opposing average is skewed to the more difficult batters to get out. The guys who are good batters and whom a manager is never going to ask to sac bunt.


T
Yeah, thanks for elaborating on it. I knew it had something to do with the stats, I just didn't know what exactly, or more specifically, how exactly.

But this is just one of the reasons they call baseball a numbers game.

Cloned is offline  
Closed Thread

Forum Jump


Bookmarks

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off



All times are GMT -5. The time now is 12:18 PM.

monitoring_string = "e4251c93e2ba248d29da988d93bf5144"
Contact Us - HFBoards - Archive - Privacy Statement - Terms of Use - Advertise - Top - AdChoices

vBulletin Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
HFBoards.com is a property of CraveOnline Media, LLC, an Evolve Media, LLC company. 2014 All Rights Reserved.