THE BOOK--Playing The Percentages In Baseball

The AL is certainly superior, I think its far more significant than the DH thing.  Also, if the series goes 7 that’s 4 games where Ryan Spilborghs is in the lineup at DH - he’s probably weaker than what the Red Sox normally face at the position.  70% or so seems about right.

Everyone is picking the Red Sox.  What I guess BP is doing is being contrarian, if the Rockies somehow pull off an upset then you’ll see their call on the back cover of their book.

I’m less inclined to believe conspiracy theories in general, but I don’t think there’s anything fishy going on there.  I can’t imagine they decided to abandon their formula, make up a number, and post that on the off-chance that they’re right and can then brag about it.

If you’ve got a formula that you believe in, you stick with it… until it’s demonstrated (theoretically or in practice) to be wrong.  It would be nice for them to have a nice paragraph or two explaining why their math has come up with such a different prediction than the math of everyone else… but again, I can’t imagine they just pulled that number out of their ass to be different.

Oh, and ProTrade “wisdom of the crowds” has the Sox at 63%:

http://www.protrade.com/leagues/MLBLanding.html

In this article:
http://www.hardballtimes.com/main/article/is-the-al-really-superior-part-3/

MGL said:

The difference in offensive talent is somewhere around .4 to .5 runs per game, which would give the average AL team around a 55% advantage over the average NL team, assuming that the pitching were equal. If it is true that the AL has the better pitching overall as well (and there is some suggestion in the data that that is true), then the AL advantage may be as high as 56 or 57%.

If I counted right, the AL had a .544 record against the NL this year.  Over the last three years, it’s .571.  I think we can give a rule of thumb for this year to bump up all AL records by .025 and bump down all NL records by .025.  (This means that a true .525 AL team facing a true .475 NL team would result in a .550 win percentage for the AL team.)

Anyway, so you have a .050 win gain per game for the Redsox over the Rockies, just on the talent disparity of their leagues (or 8 wins in a full season).

Because of inter-league play, Ortiz was DH for only 140 games (the equivalent of 6.0 DH games per 7 games played).  With the World Series, if it goes to 7 games, he’ll be DH in 4 of them.  So, that’s 2 DH games where Ortiz will disappear.  In effect, one-third of his value goes away.

(Some of it comes back in the form of playing 1B, but his detriment in fielding compared to what Youk can do overall, is probably not going to change much.)

So, losing one-third of Ortiz as a DH probably costs you the equivalent of 2 wins in a full season.

Rally is right that the Ortiz/DH effect won’t add up anywhere close to the actual talent disparity between the two leagues.  You need an overall 6 win adjustment (8 minus 2) to their current 10 win gap to put everything on the same playing field.  That’s a 16 win gap for the season.  That’s the same as saying the Redsox are a true .550 and Rockies are a true .450 (or .600/.500, or .700/.600, etc, more or less). 

So, our one-game expectation is for the Redsox to win 60% of the time.  The binomial on that for a 7-game series is 71% Redsox win.

I’d like to know under what basis the Rockies have a 56% chance of winning.  I certainly don’t think BP is trying to be contrarian about it.

I heard about the BP thing and I thought that was a mistake.  It can’t possibly be correct.  It has to be a mistake.

I think the AL dominance is greater than .550.  Take the last 3 years’ AL wp in IL games at least.  More talent actually went to the AL this year as compared to last, so that the AL should be even a little stronger than last year, as compared to the NL.

Doing sims on individual games, I have Boston at around 73.5%.

If you did nothing else but used each team’s wp this year a proxy for their true strength (relative to their own league of course) and then adjusted for league strength, Boston would have to be in the 70’s.

If you did nothing else but assumed that their offenses and defenses (it is a myth that the Rockies have a great defense overall) are around even (which they are) again, relative to the rest of their leagues, and then look at their pitching, Boston has to be at least in the 70’s.

The Rockies still have a large home/road disparity because of the hangover effect and because visting teams are not used to Coors.  Therefore the extra game in Boston hurts the Rockies, I think (normally the extra home game being game 7 is not really an advantage for the team with 4 games at home).

So, no matter how you look at it, as long as that look is analytical, Boston is a large fave, somewhere in the 70’s.

BTW, the “Vegas (no-juice) line” right now is 68-69% depending on where you want to bet.

BTW, I just looked at BP and in their little “playoff odds report” box, they have Boston as a 59/41 fave, which is still way too low. My guess is that they did not do the proper league adjustments.  Almost no one does.

Clay now has the Red Sox at 59%:  http://www.baseballprospectus.com/statistics/postseasonodds.php.

Maybe he read this post.....

Here is a rundown of each starting lineup and each player’s current offensive and defensive projection in runs per 150 games, relative to their own league:

In Boston

COL

Matsui -6, 6
Troy 5, 11
Holliday 20,2
Helton 24, 3
Atkins 11, -4
Hawpe 9, -9
Yorvit -12, 0
Taveras 4, -2
Spillborgs +7, 0

Total of +62, +7

BOS

Pedroia +4, +2
Youk +11, +1
Ortiz +36, 0
Manny +22, -15
Lowell +2, +4
Drew +15, -2
Varitek 0, 0
Jacoby +3, +2
Lugo -3, -5

Total of +90, -13

The offensive numbers are relative to their own league.  The defemsive numbers are not.

In COL

BOS

Pedroia
Jacoby
Ortiz
Manny
Lowell
Drew
Varitek
Lugo

Total of 79, -20

COL

No Spilborghs

Total of +55, +7

So basically in Boston, you have +8 runs per 150 advantage for Boston and in COL, +3 runs for COL.  This does not include the HFA or the league adjustments.

Pitching-wise, it ain’t even close:

Here are the normalized park adjusted and defense adjusted component ERA projections for all the starters:  An average starter is 4.20.

Beckett 3.20 Francis 3.80
Schilling 3.50 Jimenez 4.20
Dice-K 3.70 Cook 3.90
Lester 4.40 Fogg 4.50

Advantage per 6.5 innings

game 1 Bos .43 runs
game 2 Bos .50
game 3 Bos .14
game 4 Bos .07
game 5 Bos .43
game 6 Bos .5
game 7 Bos .14

Let’s call the bullpens, benches, and managing a wash.

Let’s even call the offense and defense a wash.

Remember these are all with no league adjustments so far.

Here are the wp’s (with no league adjustments) for each game, based upon the difference in starting pitching.  I am also adding in a HFA of .55 for each team (it is usually .535).

Game 1 Bos 60%
game 2 Bos 61%
game 3 COL 54%
game 4 COL 55%
game 5 Bos 50%
game 6 Bos 61%
game 7 Bos 57%

I am not sure how to add in the league adjustments, so I will just add another 5% for Boston for each game.

Game 1 Bos 65%
game 2 Bos 66%
game 3 BOS 51%
game 4 BOS 50%
game 5 Bos 55%
game 6 Bos 66%
game 7 Bos 62%

Running those numbers through a mini 7 game series computer sim, we get:

Bos wins series 70% of the time.  Sweeps 11%.  Wins in 5 games 18%.  Wins in 6 games 23%.  Wins in 7, 18%.

I don’t know exactly what he (Clay) is using as “inputs,” but they don’t sound too good.  For example, if he is using team w/l results against RH and LH pitchers, that is laughably innacurate.

Tonight, JD Drew is starting against LH Francis, whereas against Sabathia, the Red Sox chose Bobby Kielty. Set aside for a moment who is the right choice and think about the thought process here. Francona made the change because 1) Drew is “hot” (over his two games) and 2) Kielty does not have a strong track record v. Francis as he did against Sabathia (which was 22 PA). That is mind-bogglingly bad managing in that Francona is allowing absurdly small sample sizes to dictate decisions. Also, Francis has larger platoon splits than Sabathia. So if Drew could only start against one pitcher, better it be Sabathia.

So who is the right choice? I have Kielty as a .329 wOBA hitter overall based on a weighted average of his last three years’ stats. Kielty has a career platoon split of .080 points. Then, I regressed the splits to get a “true” wOBA v. LHP of .361 and a split of .052 points. (A similar technique was used in The Book where Kielty’s figures were .373 and .032 using 2000-2004 stats. An overall drop in production and more sample size/less regression account for the changes, respectively).
Doing the same for Drew, and starting with MGL’s +15 number as a starting point, gets me a .369 overall wOBA and a .335 wOBA v. LHP.

NB: I use a spreadsheet where league average woBA is always .340.

32 PA, not 22 (not that it invalidates the argument). better placed in the strategy thread?

I actually meant to put it there. My mistake.

Clay Davenport updated his odds looking at EqA matchups (which was...interesting) here: http://www.baseballprospectus.com/article.php?articleid=6874

He has Boston at 73.6% to win the series.

Right, the lefty/righty thing Clay did was not good.

One question about Win Probability after Game 1:

Has it ever been mentioned before that WP we used now should be fine tuned in post-season (according to post-season stats, maybe) since the game strategy is more or less different from regulars? Or, it’s a illusion of mine?

Thanks.

Ty, please explain more what you mean?  The only “strategy” that is really different in the post-season is the bullpen - often using closers for more than one inning, using 5th (and sometimes 4th) starters in relief.

Other than that, you want to use current projections for players and that is about it.  Using season stats for each player or season team w/l totals is not a very accurate way to figure post-season odds for a number of different reasons.

And again, the biggest mistake that people are going to make is to fail to take into consideration the large difference between the leagues.  We only get a sense of each team’s strength in relation to the rest of their league.  For example, if the Rockies were a true 90 win team in the NL and played like a 90 win team during the season, they would look like an excellent team, period.  But if you put them in the AL, they would win around 82 games, be a .500 team, and look quite average.  That is a hard concept for most people to get their arms around.  For example, an NL team facing Beckett is like them facing Peavy on steroids.  And their pitchers facing the Red Sox lineup is like them facing themselves on steroids.  The BoSox lineup is WAY better, and their pitching is ridiculously better, after accounting for league differences.

It is (very) unlikely that the way players have played in the post-season, this year or in previous years, has any (significant) predictive value to how they will play in any post-season game.  Just another one of 1,238,946 things that everyone thinks is true in baseball but is not.  If that is what you mean.

Thanks for you reply, MGL. Sorry if I didn’t (and can’t) express myself very well cause my English isn’t too good.

My original thought is: since teams in post-season have to achieve goal and handle situations (much bigger leverage, facing better teams, facing teams from different leagues, more off-days, etc) very different from in regular season, will these factors change Win Probability and Run Expectancy (in any in-game situation)? IOW, Is it right to use WP and RE based on regular season games in playoffs? (like Fangraphs did)

And if we use all post-season games as a sample to calculate WP and RE, will we find out any significant difference? (but I guess the sample is too small)

Thanks.

Ty, you are right that you will have a different win probability set of numbers.  That’s not necessarily a post-season issue however.  Since 1969, the average number of runs scored in the playoffs is 4.0, which is over 10% lower than the regular season.  This is small-ball territory.

I handicap baseball for a living and came up with 72.2% Red Sox before Game 1 (82.2% now) using a 54% AL advantage.  Click my name for the link.

Here’s the custom RE charts:
http://www.insidethebook.com/ee/index.php/site/comments/run_expectancy_by_run_environment/

You can see how things change by looking at some simple breakeven points.  For example, look for the 4.0 RPG environment, and note the RE numbers of man on 1B, no outs, man on 2B, no outs, and bases empty 1 out: .810, 1.048, .233.  That’s a SB gain of +.238 runs and a CS loss of .577 runs.

Compare that to the equivalent 5.0 RPG environment: .953, 1.190, .297, or +.237, -.656.

As you can see, the gain from a SB is virtually identical in either run environment!  If you’ve been following this blog, you should have expected this.  The LWTS run value of the double, relative to the single, is extremely stable in the 3.0 to 6.0 runs per game environment.  The same can be said for the SB run value.

The killer is the out.  It is alot less costly when fewer runs are expected.  The above breakeven points are: 70.8%, 73.5%.  (In The Book, I give you a better way to figure this out, which really lowers these numbers a bit.)

This is a fairly substantial difference, since you’ve got many runners that are in the gray area between these breakeven points.

James/18 has more explanation here:
http://playoffodds.blogspot.com/2007/10/how-this-works.html

However, it’s fairly light in the explanation.  As he notes, it all depends on the inputs.  Using an up-to-date Marcel, with handedness splits, would be ideal.

And of course with this WS being played in two of the best hitters’ parks in baseball AND between two of the best hitting teams in their respective leagues, you can expect to see at least 5 rpg per team.

The reason that prior WS games have been low scoring is 5 fold:

1) The weather.
2) Teams that are in the WS tend to play in pitcher’s parks. 
3) Teams that are in the WS tend to be a little better in pitching and defense than in hitting (I think).
4) Teams can use their 3 or 4 best starters.
5) Teams use their bullpens more efficiently.

Since 1969, the average playoff team has scored 5.0 and allow 4.0 runs per game in the regular season.  The average team scored right around the mid-point of that.  I don’t think #2 and #3 apply.

I think it’s almost certainly the case of disproportionate innings going to your best pitchers (basically, #4 in MGL’s list, plus a bit of #5).

I hadn’t considered the weather, but that certainly would apply.  In 2007, the ERA was 4.12 in April, and 4.60 in August.  (I won’t quote September, because of the callups.  We want to at least get a group of pitchers that are the same in both months.)

On the other hand, in 2006, the ERAs in April and August were 4.63, 4.38 (reverse weather splits!; (*) ).  And in 2005, they were both 4.27 (and July was 4.29). I’m pretty sure I’ve seen month-based long-term totals somewhere.  Easy enough to do if I had the data at the office.

(*) Yup, August played cold in 2006.  Make sure to adjust all the stats based on that.  I’m being facetious.  I can’t stand all those “the park played like a hitter’s park this year” talk.  What the heck does that mean?  The only way that makes any sense is if the climate impacted the park.  Otherwise, it’s a b-s statement to make.  Every month, there’s about 30,000 PA.  So, 1 SD in OBP = .003 points.  At 2 SD, 39 PA per game, and converting OBP to runs, that gives us a 95% interval of +/-0.20 runs per game.  So, the reverse splits in 2006 is probably real, but not necessarily climate-related.  The 2006 World Cup might have had an effect, as Nate Silver detailed at the time.  If April and August have the same runs per game, this could be totally by luck.  30,000 PA is, hard to believe, not alot of data.

Anyway, getting back to MGL’s point, the weather would have something to do with it, but I would guess just a little bit.

Here’s that excellent Silver article:
http://www.baseballprospectus.com/article.php?articleid=5060

(There’s a broken link in that article.  It references an earlier article, which you can find here:
http://www.baseballprospectus.com/article.php?articleid=5054
)

Excellent insight about the 1995 return of the lockout.

Clearly, some people have forgotten about May 2, 1987, the greatest “first game played with no spring training or minor league play at all” ever:
http://www.baseball-reference.com/boxes/NYN/NYN198705020.shtml

MGL/5: you talked about the Rox fielding.  What do you have for them?

http://www.tangotiger.net/scouting/scoutResults2007_COL.html

Fans see Tulo as +20, Kaz at +10, Helton at +10, and Atkins at -20.  Overall IF at +10.  In the OF, they come out at even.  So, it looks like the Fans agree with you, that the Rox fielders are a bit above average, but nothing more than that.  Atkins seems to be a real killer according to the fans.  Basically, as good as Tulo is, Atkins is that bad.

It’s easy for people to add up all the pluses, but forget about the minuses.  Kinda like going to Vegas and saying you “won” big, without counting all the other times you actually lost bigger.

Yup, I agree about adding up the pluses and forgetting about the minuses.  That IS the rationale for people thinking that The Rox are “a great fielding team” (that and the fact that they had the fewest errors in reg season all-time I think).  Same thing with Manny: “Yup, his fielding is pretty bad, but who cares, look how good of a hitter he is?”

I gave some approximate defensive projections for all of the Rox players above.

I listen to the XM radio baseball channel about half the time I am in the car.  I usually listen to the commentaries until I get so red-eared than I have to change the channel. 

Out of all the commentary (Dibble, Kennedy, Steiner, Wilson, et al.), I would say that easily 95% of it is complete and total B.S.  It NEVER ceases to amaze me how phony the mainstream sports analysis industry is.

Here are some typical questions asked of commentators, players, managers, etc., and the answers I would LIKE to hear:

“How good was (insert name of any manager on a succesfull team) this year in leading his team to (insert any end-point)?”

Answer:  Not very good at all, actually, Chuck.  I think that the (insert team) won DESPITE the manager or at least any other manager would have accomplished essentially the same thing.  After all, the team was talented to begin with and were EXPECTED to win 89 games!  If you want to give a manager credit for his team winning exactly what they were expected to win, that’s fine by me.  But personally, I find that stupid.  Any other questions, Chuck?”

“So, what do you think the keys to the (insert team) winning the next game are?”

Answer: “Well, Rob, I’ve thought about this a lot, and I think that if they just stay within themsleves and score more runs than the other team, they’ll do just fine.  What do you think?”

“Do you think the long layoff will hurt the (insert team)?”

Answer: “Well Charile, I think it might.  But then again, I think it might not.  Let’s just wait and see who wins.  If they lose, then I guess it hurt them.  If they win, it didn’t.  We’ll just have to see, won’t we?”

“How much does (insert favorite player) mean to this team?”

Answer: “Above average, bench, or replacement?  In win shares, VORP, WARP, or Superlwts?”

Answer:  Not very good at all, actually, Chuck.  I think that the (insert team) won DESPITE the manager or at least any other manager would have accomplished essentially the same thing.  After all, the team was talented to begin with and were EXPECTED to win 89 games!  If you want to give a manager credit for his team winning exactly what they were expected to win, that’s fine by me.  But personally, I find that stupid.  Any other questions, Chuck?”

No offense… but I’ll insert “Yankees” into the parentheses for Joe Torre’s cry out loud rooters.

Of course we can give a manager some credit for their players’ pre-season projections if that manager was with the team in prior seasons.  But then again, we can give a manager credit for just about anything the team does, good or bad, right?

As part of an article I wrote for THT Annual (I hope they don’t mind me spilling a little of the beans), here are the pre-season w/l projections for all teams, based on my own pre-season projection for all the players on that team AND based on their actual roles and playing times during the season.  It also uses their actual schedules to account for “strength or schedule” issues.  The algorithm just adds up every player’s (batters and pitchers) projection in linear weights and creates a “one-time” win percentage for each team.  It then plays out the actual schedule of each team several thousand times and averages the season w/l totals.

For example, the Yankees, based on their players’ pre-season projections might be a .580 team, using lwts projections for all their batters, fielders, and pitchers (giving extra weight to closers and setup men) and their actual playing time.  If they play Boston in the first game of the season and Boston is a .570 team, it does a log5 and uses a random number to come up with a winner.  The next game for the Yankees, they are a still a .580 team versus whomever they are playing and their “one-time” win percentage. Etc.

Anyway here are the team w/l projections for every team based on this model and their actual pythag w/l records at the end of the season. 

Note two things:

One, how accurate (IMO) player projections are in predicting season records (once you know everyone’s playing time of course, which is “cheating"), even considering that there is around a 6.5 SD in wins for every team based on chance alone. 

Two, the presumption is that a manager could not possibly have “helped” his team if its record is equal to or less than that expected from the player projections.

The first column is the projected w/l based on the player projections only (and their actual playing time).

The second column is their 2007 pythag record.

Of course there are a million factors unrelated to a manager as to why a team’s pythag record would differ significantly from that expected from its player projections, not the least of which is plain old luck, but if we continue with the presumption that a manager should not get ANY credit unless his team wins more games than if that team had NOT MANAGED at all (or at least had an average manager), then:

The only managers who MIGHT deserve some credit are:

Francona
Gibbons
Leyland
Sciosca
Washington
Manuel
Piniella
Narron/Mackanin (I had to look that one up)
Hurdle

The managers whose teams far underperformed (in pythag wins) their player projections were:

Perlozzo/Trembley
Maddon
Guillen
Gardenhire
Gonzalez
LaRussa
Garner/Cooper
Tracy

I don’t know about you, but that looks like a pretty random dichotomy of managers to me.  Garednhire, Guillen, Maddon, and LaRussa are generally considered good managers I think.

Gibbons, Manuel, Narron, and Hurdle before this year were not considered particularly good from my recollection.

Why aren’t Torre, Acta, Wedge, Randolph, Black, Melvin, and Cox on the “good list?” They are all considered excellent managers.  To their credit, they collectively achieved 5 more wins (less than one per) than expected from their players’ projected performance.

Although this data is not really indicative of anything (it is a toy and even that is overstating the issue), anyone who tells you which manager they think is good or bad is full of it!

NYY 94 97
BOS 90 101
TOR 82 86
BAL 75 71
TB 81 67

AL CENTRAL
CLE 91 92
MIN 83 79
DET 80 90
CWS 75 67
KC 73 75

AL WEST
OAK 81 79
ANA 83 90
TEX 76 80
SEA 81 78

NL EAST
ATL 87 89
PHI 83 87
NYM 84 87
FLO 78 72
WAS 72 71

NL CENTRAL
STL 81 72
MIL 85 83
CHC 80 86
HOU 78 73
PIT 78 70
CIN 70 74

NL WEST
SD 89 88
ARI 81 79
LA 82 81
COL 80 90
SF 77 78

Your fieldign numbers shows +7 per 150, or +8 for the full season, which is right where the Fans have the Rox fielders (though not a match for each player).

The big gap is Atkins.  2007 and 2006:
http://www.tangotiger.net/scouting/scoutResults2007_COL.html
http://www.tangotiger.net/scouting/scoutResults2006_COL.html

In 2006, he was below average, but in 2007 he was far below average, the 2nd-worst in the league. I saw him last night, and it’s hard to believe he can be anything close to that bad.  He seems capable.  But, I suppose that’s why it’s better to have the day-in day-out fans telling you what they see.

And Matt Holliday also took a step up.  I would guess that his great hitting season may have biased his results.

Here’s the thing (or one of the things) about luck, fluctuation, etc., with regard to fielding.

Some of the “luck” associated with a defensvie metric (or lots of other evaluation methods and metrics) is the exact location and “temperament” of the balls and things like that.  For example, not all hard hit balls to a particular UZR section of the field are created equal.  Some are probably much easier than others, which you would see if you watched each play.  In addition, UZR does NOT take into consideration the position of the fielder when the ball was hit, other than the adjustments it uses for outs and baserunners.  So, for example, with certain batters, the third baseman is playing way up guarding against the bunt.  UZR does not know that (although perhaps it should) so that maybe a routine ball, according to UZR, is actually not catchable.  Again, those are the types of “discrepancies” that observation/the fans will pick up.

The other type of luck/fluctuation (and this is most evident with pitchers) is that on some days (and even seasons) a fielder looks good and gets to lots of tough balls and on other days he doesn’t.  That will NOT be picked up by the fans. I say that because maybe this year Atkins just looked real bad (even though he isn’t) and last year he looked good.  Atkins UZR was -14 this year and -3 last year, so it matches up with the fans I guess.

Or a player could be playing with an injury for a good part of a year or gain or lose some weight or actually improve his defense (in true talent) from one year to the next, so not only are we trying to smooth out the fluctuations and reduce the noise when using multi-year data for projections and estimate of true talent, but we are also coming up with a number that is our best estimate of the player’s true talent averaged accorss those years and NOT necessarily his CURRENT true talent, although we have to assume that it is.

Interestingly, Holliday’s UZR also paralelled the fan ratings - +1 laat year and +11 this year.

That is interesting.  I wouldn’t be surprised that if I gave you the 30 guys who most “improved” according to the Fans, and the 30 who most declined, that their composite UZR would also have tracked that, to some extent.

That is, a group of players that Fans see as gaining 30 points (+20 runs) would likely show a gain in UZR of something like +8 runs.

In effect, we can differentiate between two players that UZR both saw as a +8 gain: one was real, and one was due to sampling errors.

I think it is the oppisite.  If the fans saw a +10 improvement, then UZR should see about the same (a little less, probably).

If UZR sees +10, then fans should see around +4, or whatever.

According to the 01 to 07 UZR database, players who play both left and center are 9.4 runs different.  If they play right and center, they are 12.6 runs different, suggesting that RF is a weaker field than left, as you would expect (because they need better arms and have to put some slower players there that might otherwise play first base).

I think you were saying that otherwise CF’ers might play RF if they had a strong arm (like Ichiro) but not LF, which would make RF a stronger position.  I see that argument too.

Players that played RF and LF lost 3.4 runs when going from RF to LF, also suggesting that LF has the better fielders.

3.4 + 9.4 = 12.8 which is close to 12.6, which is nice, and suggests that there is not much difference in the skills necessary to play each of the outfield positions.

I think it’s easy to see that the average LF would have a -1.5 arm and the average RF would have a +1.5 arm, thereby knocking out any difference, which is what I was saying (speed and arm are inversely related in the corner positions).

Your numbers therefore average to 11 runs difference between CF and the corners.

Going back to my equation:

(corner - 9)*4/3=CF

If you set corner to zero, that equation gives you -12 in CF. 

If you set CF to 0, that equation gives you +9 in the corners.

All this is entirely consistent with the numbers you are providing.

It’s for this reason, that when I do the positional adjustments, I simply give CF +0.5 wins and the corner OF -0.5 wins.

Of course I am not including arms in the numbers I presented.  I agree that -1.5, +1.5 is good, which makes up for the difference in UZR.

Tango,

Basically, I estimate the probability of each game going to the Rockies or Red Sox (or whoever is playing in the series).  Then I use these to calculate the chances of each possible series outcome.

Simple example: Boston is a 70-30 favorite at Fenway and 60-40 at Coors, regardless of starting pitchers.  Then their probability of winning the series in the order WWLWW = .7 * .7 * .4 * .6 * .6 = 7.1%.  Do this for every winning combination, and you have the team’s total chances to take down the series.  (81.3% in this example)

Obviously I can’t give away the exact method I use to determine the individual game odds, because there are valuable secrets I need to protect.  But they involve a lot of your research and MGL’s, so I owe you each a drink.

That’s it, just a drink?