Predictions and Projections

Projection Time! (Red Sox Edition)

Finally, PECOTA projections for 2009 are out. They join a whopping five other projection systems jockeying for assessing how well every player in MLB will do next season. Fantasy baseballers love this stuff, but I suck at fantasy baseball; I'm just curious for a semiscientific way of guessing at how my favorite players will do next season.

The projection systems are as follows:
  • PECOTA, run by Baseball Prospectus' Nate Silver, generally the most accurate.
  • Bill James, run by Bill James, not bad on offense, not that great on pitchers. Generally thought to overrate minor league numbers. 
  • Marcel, run by The Book's Tom Tango. Found to generally be the most accurate non-PECOTA system. 
  • CHONE, run by Sean Smith 
  • Oliver, run by Brian Cartwright of Statistically Speaking. This one's new and builds off the Marcel formula.
  • ZiPS, run by Dan Szymborski of Baseball Think Factory.  

That's just a ridiculous number. Why don't we stick to three? The most accurate seem to be PECOTA, Marcel and ZiPS, so let's limit ourselves to those.

What I've done in the past is look at the projections for the principal players on a team and compare them to their performances the previous seasons. This time around, I'll look at the Sox' and Yanks' 2008 performance by position, then compare that to the 2009 projections for the planned starter and principal backup at that position.

Also, I'll average the projections to compare the Sox' and the Yanks' starters in 2009. So here we gooo.

Red Sox 2008:
.218/.308/.342, 644 PA, 123 H, 14 HR, 69 BB, 172 K (Varitek 74.4%, Cash 23.4%)

Jason Varitek 2009:
P: .234/.323/.388, 258 PA, 7 HR, 28 BB, 63 K
M: .233/.325/..385, 493 PA, 14 HR, 55 BB, 116 K

Josh Bard 2009:
P: .257/.331/.387, 245 PA, 3 HR, 23 BB, 33 K
M: .266/.342/.395, 343 PA, 6 HR, 35 BB, 50 K

Everyone projects Varitek to improve upon his 2008 numbers. The most pessimistic line, PECOTA's, still has him gaining 10 points of OBP and 29 points of slugging, even while (apparently) splitting time with Josh Bard, on whom they are decidedly pessimistic — and even then, they seem him being slightly better than Tek09 and much better than Tek08. The idea here: No place to go but up.

Red Sox 2008:
.312/.382/.519, 705 PA, 194 H, 115 R, 25 HR, 64 BB, 115 K (Youkilis 68.4%, Casey 23.8%, Bailey 4.4%)

Kevin Youkilis 2009:
P: .275/.366/.475, 597 PA, 21 HR, 70 BB, 109 K
M: .287/.377/.477, 573 PA, 18 HR, 65 BB, 101 K
Z: .288/.383/.476, 608 PA, 19 HR, 74 BB, 111 K

Mark Kotsay 2009:
P: .269/.321/.380, 341 PA, 4 HR, 24 BB, 38 K
M: .258/.317/.373, 441 PA, 6 HR, 35 BB, 51 K
Z: .275/.332/.394, 390 PA, 5 HR, 30 BB, 39 K

No one thinks Youkilis will repeat his phenomenal 2008. No one has him hitting over .300 or reaching base at a .390 clip, like he did in 2008, or really coming within 100 points of his '08 slugging. That would be a pretty hefty decline. I'd take the over on most of these Youkilis projections. Kotsay, meanwhile, is projected all over the mediocre map — an OBP from .317 to .332, a SLG from .373 to .394 (and .397 in another model). 

Red Sox 2008:
.321/.371/.484, 757 PA, 218 H, 88 R, 17 HR, 53 BB, 55 K (Pedroia 95.6%)

Dustin Pedroia 2009:
P: .303/.364/.447, 649 PA, 12 HR, 51 BB, 53 K
M: .312/.370/.465, 621 PA, 13 HR, 48 BB, 54 K
Z: .307/.367/.467, 635 PA, 15 HR, 51 BB, 47 K

No backup for Pedroia because he hasn't shown a propensity for injury, and if he were injured, the backup for shortstop would simply slide over anyway. Everyone else sees Pedroia reaching base at a basically the same rate as 2008, with slightly less power. Sounds reasonable. But the projection systems have been underrating Pedroia since he came into the league, so maybe we should expect another big season from the reigning MVP.

Red Sox 2008:
.268/.358/.360, 655 PA, 149 H, 63 R, 2 HR, 70 BB, 107 K (Lugo 46.7%, Lowrie 28.5%, Cora 24.7%)

Jed Lowrie 2009:
P: .260/.341/.432, 414 PA, 9 HR, 44 BB, 77 K, 29 2B
M: .268/.345/.425, 353 PA, 6 HR, 37 BB, 69 K, 24 2B
Z: .257/.337/.395, 498 PA, 7 HR, 55 BB, 91 K, 34 2B

Julio Lugo 2009:
P: .255/.325/.347, 286 PA, 3 HR, 25 BB, 44 K, 13 SB
M: .252/.320/.365, 416 PA, 6 HR, 36 BB, 65 K, 17 SB
Z: .266/.338/.391, 440 PA, 8 HR, 41 BB, 79 K, 21 SB

No one's particularly bullish on Lowrie, which I think is a mistake, but projection systems are largely skewed to statistical results and don't know he had a broken hand most of 2008, and as bleh as Lowrie's numbers look (slight improvement from two, no improvement from ZiPS), each system projects Lugo as a worse player: PECOTA has a .773 OPS for Lowrie, to .672 for Lugo; Marcel says .770 versus .685. Only ZiPS says it's at all close, with a .732 for Lowrie and a .729 for Lugo. PECOTA probably has this right — even down to the ultimate breakdown in plate appearances.

Red Sox 2008:
.281/.351/.473, 710 PA, 176 H, 119 R, 24 HR, 66 BB, 99 K (Lowell 65.1%, Youkilis 19.0%, Lowrie 15.2%)

Mike Lowell 2009:
P: .272/.332/.442, 469 PA, 14 HR, 37 BB, 62 K, 27 2B
M: .283/.343/.452, 499 PA, 15 HR, 40 BB, 63 K, 29 2B
Z: .283/.345/.449, 554 PA, 16 HR, 46 BB, 64 K, 34 2B

Everyone sees a gentle slop downward for Lowell — same on-base skills and a drop of 10 points in slugging. Nothing crazy or controversial from any of these projections. Youkilis and Lowrie are obviously the backups here.

Red Sox 2008:
.284/.357/.458, 737 PA, 185 H, 101 R, 24 HR, 69 BB, 153 K (Ramirez 39.1%, Bay 28.6%, Ellsbury 23.3%)

Jason Bay 2009:
P: .271/.364/.493, 626 PA, 26 HR, 76 BB, 137 K
M: .269/.359/.474, 596 PA, 24 HR, 70 BB, 129 K
Z: .278/.375/.510, 679 PA, 31 HR, 89 BB, 154 K

Except for his injury-plagued 2007 season, Jason Bay has never had a batting average below .280, only once reached base below a .370 clip, never slugged below .520 and only once had fewer than 30 home runs (both those "onces" came in his first full season, 2004). So the projection systems all come out with what amount to career lows in every category? I see that as unlikely, though given how horrible 2007 was, it's hard for a computer-based system to simply dismiss that — particularly a system like Marcel, which gives 2007 two-thirds the weight as his 2008 and double the weight of his 2006. The ZiPS line looks the most realistic to me.

Red Sox 2008:
.291/.356/.432, 700 PA, 181 H, 70 R, 13 HR, 60 BB, 93 K (Coco Crisp 58.1%, Ellsbury 41.0%)

Jacoby Ellsbury 2009:
P: .291/.348/.409, 569 PA, 7 HR, 42 BB, 75 K, 42 SB
M: .289/.348/.422, 517 PA, 10 HR, 38 BB, 71 K, 34 SB
Z: .287/.345/.403, 584 PA, 9 HR, 43 BB, 75 K, 45 SB

Rocco Baldelli 2009:
P: .263/.320/.443, 262 PA, 8 HR, 17 BB, 61 K
M: .266/.324/.456, 260 PA, 9 HR, 17 BB, 53 K
Z: .277/.338/.482, 150 PA, 6 HR, 9 BB, 32 K 

Everyone projects Ellsbury to be better in 2009 than he was in '08 — about 10 points better each in OBP and slugging. They project another fine season on the basepaths, and they also think he'll cut down on the strikeouts a bit. Basically, they see him doing as well as he and Coco Crisp did combined, when Terry Francona was playing each one whenever the hot streaks came. Baldelli, meanwhile, is a bit of a mystery. ZiPS alone thinks he'll repeat his .344/.475 numbers from 2008, but only in 150 appearances, while the other two don't seem as impressed. 

Red Sox 2008:
.272/.369/.465, 722 PA, 167 H, 99 R, 23 HR, 93 BB, 123 K (Drew 61.5%, Ellsbury 20.6%, Kotsay 10.2%)

J.D. Drew 2009:
P: .269/.374/.451, 464 PA, 13 HR, 65 BB, 85 K
M: .270/.378/.455, 483 PA, 14 HR, 70 BB, 88 K
Z: .277/.388/.474, 408 PA, 13 HR, 62 BB, 72 K

Somehow people don't expect a 33-year-old Drew to again be the best right fielder in baseball (.280/.408/.519) or have one of the best Junes in team history (.337/.462/.848). Still, this is another situation where a terrible season for explainable reasons, also in 2007 as it turns out, is driving down projections that have nowhere to look but at the past two or three seasons. Other than 2007, Drew has slugged .498 or better every season since 2002, so it's hard to see (given the predicted playing time) Drew slugging much less than .490. Especially odd is that the system projecting Drew to miss the most time, presumably because of injury, is also forecasting the highest slugging percentage, which was the most adversely affected by Drew's injury last season. 

Red Sox 2008:
.273/.379/.490, 697 PA, 161 H, 122 R, 29 HR, 98 BB, 115 K (Ortiz 70.4%, Ramirez 19.2%)

David Ortiz 2009:
P: .269/.375/.504, 569 PA, 27 HR, 81 BB, 96 K
M: .285/.395/.542, 512 PA, 26 HR, 77 BB, 84 K
Z: .286/.395/.562, 598 PA, 34 HR, 91 B, 98 K

Those numbers may not look Ortizian, but it's important to note that projection systems are inherently conservative; you can't predict seasons like Ortiz had from 2004-07 because those seasons are historically rare. But Marcel projects Ortiz with the fourth-best OPS in baseball — one point behind Matt Holliday and six above Alex Rodriguez. PECOTA, which is definitely not as optimistic, still puts Ortiz just behind Mark Teixeira, fifth in the American League, in slugging. CHONE puts Ortiz third in baseball in OPS, between A-Rod and Howard. Oliver places him fourth, also behind A-Rod. And none of them projects Ortiz to slug as high as ZiPS projects. All in all, the systems see a nice rebound indeed from the big man. The Hall of Fame push continues! ;-)

For starting pitchers, I ordered the 2008 slots by ERA, though obviously this is a bit flawed (the Sox "third starter" was clearly their most effective — and least lucky, for example). But I think this might give us a better idea of what production each projected starter is truly replacing.

Red Sox 2008: 
2.90/1.32/.211, 167.2 IP, 94 BB, 154 K [Edited – corrected from 1.90]

John Smoltz 2009:
P: 3.57/1.21/.241, 108 IP, 29 BB, 91 K
M: 3.73/1.28/.260, 94 IP, 27 BB, 86 K
Z: 3.70/1.24/.259, 116.2 IP, 29 BB, 102 K

It seems highly unlikely any Sox starter is going to match Dice's 2.90 ERA [Edited – corrected from 1.90]. Certainly no one can project such a microscopic number. The lowest ERA for a starting pitcher projected by Marcel is Rich Harden's 3.02. CHONE gives the honor to Lim Lincecum (3.21), and PECOTA pegs Harden at 3.04. Smoltz ranks fourth in the AL under PECOTA's projection. So keep all that in mind when looking at those numbers. Most striking is how bullish all the projection systems are about Smoltz — they see him pitching at least half a season and being the best pitcher on a club that has Josh Beckett, Daisuke Matsuzaka and Jon Lester. 

Red Sox 2008:
3.21/1.27/.256, 210.1 IP, 66 BB, 152 K

Josh Beckett 2009:
P: 3.72/1.21/.239, 190.2 IP, 51 BB, 167 K
M: 3.85/1.22/.255, 167 IP, 43 BB, 152 K
Z: 3.97/1.21/.244, 183.2 IP, 45 BB, 171 K

The projection systems are unanimous in thinking Josh Beckett will continue his dominance for a third straight season, and that he'll be luckier when it comes to runs allowed this time, as well. They essentially project for him a season on par with Lester's 2008, but with better control. I'll take it.

Red Sox 2008: 
4.03/1.19/.256, 174.1 IP, 34 BB, 172 K

Daisuke Matsuzaka 2009:
P: 4.32/1.37/.238, 155 IP, 67 BB, 135 K
M: 3.79/1.34/.237, 164 IP, 75 BB, 148 K
Z: 3.79/1.37/.227, 178 IP, 87 BB, 176 K

First, that 2008 line is from Beckett. Everything to the right of his ERA is very impressive. Now, as for 2009, the projection systems all see Daisuke as the Sox' third-best starter — generally agreeing that he will trade some of his walks for hits, and as a result see his ERA rise to merely "very good" levels. At this point, the Sox' depth seems to be showing through, as this is the first spot where a definite improvement, at least in ERA, is projected.

SP 4
Red Sox 2008:
4.13/1.18/.228, 181 IP, 60 BB, 117 K

Jon Lester 2009:
P: 4.45/1.44/.260, 162 IP, 62 BB, 114 K
M: 3.76/1.32/.254, 171 IP, 61 BB, 129 K
Z: 4.09/1.42/.253, 156.1 IP, 63 BB, 109 K

Brad Penny 2009:
P: 4.47/1.44/.262, 112.2 IP, 42 BB, 80 K
M: 4.25/1.42/.269, 127 IP, 48 BB, 87 K
Z: 4.81/1.50/.272, 144 IP, 55 BB, 81 K

It'll be hard to top Tim Wakefield's 2008 from the No. 4 spot, but most of the systems seem to feel Lester can do it — and don't forget they only have one fully healthy season (last year) from which to project future years' production. Now that could mean that Lester, having thrown more than 200 innings for the first time ever, could feel those effects and regress, but I don't buy that. More likely, he'll probably take a bit of a larger step forward than what is projected here. Meanwhile, no one knows what to think of Penny. ZiPS thinks he'll pitch a whole season but be below average doing it. PECOTA and Marcel think he'll be decent but not healthy. I personally think he'll be bad and unhealthy, but I don't have the fancy computers.

SP 5 
Red Sox 2008:
5.48/1.53/.275, 233.1 IP, 93 BB, 172 K (Buchholz 32.1%, Masterson 23.1%, Byrd 21.0%, Colon 16.7%)

Clay Buchholz 2009:
P: 4.56/1.43/.248, 164.2 IP, 73 BB, 146 K
M: 4.82/1.47/.270, 97 IP, 42 BB, 84 K
Z: 4.48/1.36/.255, 136.2 IP, 50 BB, 111 K

Tim Wakefield 2009:
P: 4.98/1.43/.264, 97.2 IP, 34 BB, 61 K
M: 4.45/1.34/.257, 169 IP, 61 BB, 105 K
Z: 5.04/1.41/.256, 153.2 IP, 57 BB, 100 K

Another area in the "nowhere to go but up" category. While the systems generally (though not Marcel) look dimly at the likelihood of Wakefield continuing his impressive streak of league-average pitching for 180 innings, even numbers like that would be light years ahead of Buchholz's performance in 2008. For that matter, everyone thinks Buchholz in 2009 will be light years ahead of Buchholz in 2008. Some of this undoubtedly is the same reason why no pitcher is projected to post an ERA under 3: It's simply not practical to project any given player to be historically good or bad (even if "historic" in this sense means one of two or three in the league this season). Marcel does project 10 starters with a 5.10 or worse ERA, though, so it looks as if they truly believe Buchholz will show signs of improvement next season.

Red Sox 2008 (top six in games played, ranked by ERA): 
1. 2.43/1.35/.245, 59.1 IP, 27 BB, 38 K (Lopez)
2. 2.61/1.16/.212, 61 IP, 23 BB, 60 K (Okajima)
3. 3.27/1.12/.205, 74.1 IP, 28 BB, 72 K (Delcarmen)
4. 5.55/1.73/.268, 48.2 IP, 35 BB, 49 K (Aardsma)
5. 5.58/1.70/.240, 30.2 IP, 23 BB, 25 K (Hansen)
6. 5.66/1.62/.302, 49.1 IP, 20 BB, 32 K (Timlin)

Takashi Saito, 2009
P: 2.93/1.14/.211, 46 IP, 15 BB, 48 K
M: 3.11/1.18/.229, 55 IP, 19 BB, 57 K
Z: 2.63/1.12/.212, 54.2 IP, 12 BB, 68 K

Hideki Okajima, 2009
P: 3.72/1.31/.241, 54.2 IP, 20 BB, 48 K
M: 3.43/1.24/.240, 63 IP, 22 BB, 53 K
Z: 3.25/1.21/.225, 61 IP, 21 BB, 59 K

Manny Delcarmen, 2009
P: 3.59/1.33/.235, 58.2 IP, 24 BB, 54 K
M: 3.56/1.22/.235, 67 IP, 24 BB, 58 K
Z: 3.43/1.25/.224, 65.2 IP, 25 BB, 62 K

Ramon Ramirez, 2009
P: 3.51/1.33/.231, 57.2 IP, 25 BB, 53 K
M: 3.79/1.32/.246, 63 IP, 25 BB, 54 K
Z: 3.95/1.41/.241, 54.2 IP, 25 BB, 45 K

Javier Lopez, 2009
P: 4.33/1.52/.263, 45.2 IP, 21 BB, 29 K
M: 3.74/1.37/.252, 59 IP, 25 BB, 40 K
Z: 3.81/1.44/.250, 59 IP, 26 BB, 37 K

Justin Masterson, 2009
P: 4.20/1.43/.253, 107.1 IP, 44 BB, 80 K
M: 3.69/1.27/.236, 78 IP, 31 BB, 61 K
Z: 4.15/1.39/.249, 138.2 IP, 55 BB, 85 K

It's easy to forget how horrible the Red Sox' bullpen was before the advent of Justin Masterson allowed the Sox to dispose of using Mike Timlin and/or David Aardsma in key situations. Half the six most-used relievers posted horrific ERAs in 2008, while the worst projection of the 18 for the likely top six relievers in 2009 is 4.33. I wonder about the Saito projections, as they have the opposite problem as Bay and Drew: Saito was felled quickly by a major injury in 2008 and pitched poorly on his return — but not for long enough to seriously affect his stellar numbers for the season. So I wonder if he's not being overprojected here. The rest seem pretty reasonable (Ramon Ramirez apparently has no choice but to walk 25 batters next season), with Masterson's projections varying, depending on whether he's seen as a starter or reliever. 

Jonathan Papelbon 2008:
2.34/0.95/.229, 69.1 IP, 8 BB, 77 K

Jonathan Papelbon 2009:
P: 2.49/1.01/.203, 61.1 IP, 15 BB, 67 K
M: 2.91/1.08/.228, 65 IP, 16 BB, 67 K
Z: 2.51/0.99/.211, 64.2 IP, 12 BB, 68 K

Eight. Jonathan Papelbon walked eight batters last year. He faced 273 batters and walked eight. That's insane. Remarkably, no one since 1997 had walked so few batters while facing at least 250 — and then Papelbon walked eight of 273 and Mariano Rivera walked six of 259 in the same season. Crazy. (Though no one beats Timlin for the Sox in 2003 — nine walks out of 340 batters faced.) No reason to expect anything different next year, and the projection systems agree.


The projection systems generally agree the Red Sox should improve at catcher, shortstop, left field, designated hitter, the back of the starting rotation and the bullpen. They should stay about the same at second base, third base, center field, right field, closer and the front end of their rotation. They should get worse at first base. 

The big improvement clearly is in the pitching staff, where Wakefield, Smoltz, Penny and Buchholz all could capably fill either of the last two slots (and the systems project much more than capable pitching from Smoltz), and the additions of Saito and Ramirez, combined with Masterson, lead to a much-improved bullpen — improved (by a lot) even if Saito turns out to be injured or ineffective.

Obviously, some of these players will do worse. Others will do better. That's why these are only projections. But should the Sox live up to these numbers, particularly on the pitching side, this will be a season to remember.

Yanks projections forthcoming.

25 replies on “Projection Time! (Red Sox Edition)”

If you have the time and interest, I wonder how these systems did with 2008 relative to actual performance. Do they publish “error” terms so we know a range of how far they’re usually off? I think they’re going to be far off on Youkilis. He seems to be a late bloomer rather than a one-year wonder.
To nitpick, if Drew has the year they’re projecting, that’s a clear step down.
If the Sox get either the PECOTA or MARCEL projection for Lowrie, you’ll be pleased. That’s a top five shortstop in the AL.
I don’t see the optimism for the pitching. I don’t see how Smoltz puts up 100 innings. Either way, that means more Wakefield and Penny. Their projections aren’t that good, especially the K:BB numbers (where Penny is really bad) in the AL Beast. All combined, that’s a lot of innings to a lot of unknowns. Even if the top 3 don’t regress (and hard to see how at least Dice-K doesn’t and perhaps Lester too), it’s tough to be very confident about the staff. I know I’d think more highly if they were giving a slot to a kid with more upside.

Great work Paul, you always seem to outdo yourself. I’m really looking forward to some goddamn baseball!

To nitpick, if Drew has the year they’re projecting, that’s a clear step down.
Absolutely, but Baldelli under these projections would be a clear step up from the production the Sox got out of Ellsbury and Kotsay when they were in right field last year.

As for last year’s projections, the Hardball Times looked at some of this, though it didn’t include Marcel. (ZiPS turns out pretty poorly, actually; maybe I should have used CHONE.)
Tango had this to say:
… [W]e can stop with the nonsense that one system is better than the other. As I’ve said in the past, it is a very dubious claim, akin to saying a team that wins 85 games is better than a team that wins 84 games. Marcel will win 83 or 84 games, the good forecasting systems will win 84 or 85, and the not so good will win 81 or 82.

Cool, thanks. Rather than a ranking, I’m interested in what kind of variability we can expect like +/- AVG, +/- SLG, etc. If they’re saying that that Drew’s SLG will be .460 +/- .50, that’s not saying much.

> I wonder how these systems did with 2008 relative to actual performance. Do they publish “error” terms so we know a range of how far they’re usually off?
Averaged across all players in the league, the deviation from the expected mean will be minuscule. On a case-basis, of course there will be varying degrees of drift but I don’t see the point in player-by-player analysis of errors in predictors other than how the true statistical gurus may find whether those drifts were the fault of some specific issue with the predictive analysis tool rather than “why we play the game.”

Last year’s error on OPS (not adjusted):
THT, .051
CHONE, .053
PECOTA, .055
ZiPS, .061
So within 50 and 60 points for all hitters. That’s hardly small, but it is pretty good.
Last year’s error on ERA (not adjusted):
PECOTA, .620
THT, .629
CHONE, .665
ZiPS, .676
So almost 2/3’s of a run for all pitchers. Again, not insignificant, but pretty good.
Overall, it’s a good not great picture. For a projected .850 OPS hitter, we could be looking at a .800 or .900 OPS hitter. For a projected 3.60 ERA pitcher, we could see a 3.00 or 4.20 ERA pitcher.

For a projected .850 OPS hitter, we could be looking at a .800 or .900 OPS hitter. For a projected 3.60 ERA pitcher, we could see a 3.00 or 4.20 ERA pitcher.
Which I ghope we’d all be kind of adjusting for in our heads anyway, even if we didn’t have the exact numbers in mind. It’s why I’m pretty careful to use the word “projection” and not “prediction.”

ex-sox/ex-yankee ramiro mendoza singed on with the brewers today. who knew he was still out there?

“A good friend of mine used to say, ‘This is a very simple game. You throw the ball, you catch the ball, you hit the ball. Sometimes you win, sometimes you lose, sometimes it rains.’ Think about that for a while.”

Seriously? Someone takes the time to put together free and interesting content, during the most boring time of the baseball year, and folks are complaining that it’s boring and irrelevant?
Show some basic manners or don’t comment.

ag, as someone that actually understands volatility, error bars, standard deviations, and stuff, I disagree with you.
I would be interested though, to see if performance is normally distributed (or some variation like lognorm) since otherwise, the errors won’t mean much..

Actually, without going into a huge deal about it, one of the things that that irks me is that, generally, articles usually just give averages, which paints a picture, but perhaps a fuzzy one.
Without giving a math lesson (maybe that’ll be an article!) volatility/variance cleans that picture up a little bit – it allows us to visualize the probability buckets, and maybe with systematic contributions (basically, error functions to the model) or player contributions (overperform/underperform).
I think the problem is that there’s no real easy way to do it, other than maybe analyzing previous data/predictions. I’ve thought about this problem a bit. (say, someone who hits .300 every month, vs someone hitting .400 in October, both with the same average)
On a somewhat related note (but perhaps not), one observation is that “Game Score” by pitchers actually easily fits into this – you can easily compute the “Game Score” average, and thus the stdev/vol, and all the good stuff.
It turns out (not surprising to game theorists) that you can work out that you actually want a higher game score volatility if your average game score is not particularly high – I can elaborate, but this was actually referring to Josh Beckett’s awful 5 ERA year. Go figure.
Maybe I should code this up, hahaha.

That lack of recognizing variance also really bothers me, esp. in the projections. But if they say that they’re 95% confident that Wang will have an ERA between 3.34 and 4.66 (or Jeter will OPS between .750 and .850), you soon realize how much folly is involved. Still, for clarity, they absolutely should. It reminds me of the Netflix competition. The projection systems are incremental changes but the data is so inherently noisy we’re just stuck happily watching the games.

rob- pretty sure dc’s “zzzzzz…” was in response to the mendoza signing. besides, that’s his standard operating procedure. rest assured he wasn’t denigrating pauls fine effort.

I don’t mean to denegrate Paul, either. Clearly his research is solid and relevant and worthy of our respect.
However, I do have moments when I’m painfully old fashioned, such as when I have to do some new-fangled cipherin’.

Comments are closed.