“Does Bill James Even Like to Watch Baseball?”
Projecting players is a tough business. Because of the natural tendency to ignore injury risk and reversion to the mean, our instincts lead us to over project. We have seen this with the fans who are about half a win high (and a full win high projecting players on their favorite time) and who are generous with projected playing time.
So it is easy to look at projections and think they look low, and so we should give people a pass when they say so. But when they say so and in the process disparage the projector, I think it’s fair game to call them out. Here is a video of Bob Ryan and Joe Sullivan discussing Bill James’ projections for some Boston Red Sox players (h/t Repoz). Ryan and Sullivan were aghast that Bill James would project Jon Lester for just 14 wins in 2011, joking that their colleague Dan Shaughnessy would say “Bill James doesn’t even like to watch baseball!” and that if James actually watched Lester pitch he might think differently. Ryan claimed that Lester is a 19-game winner for the foreseeable future.
Lester is a great pitcher and won 19 games last year, but to win 19 or more games everything has to break just right. The pitcher has to pitch lots of innings, do so at a relatively high level (or be astronomically lucky), get good run support, and then have a bullpen hold the lead.
Can we expect that all out of Lester in 2011? First, let’s look at projecting innings pitched. He has pitched just over 200 innings in each of the past three years, which gives us a quick jumping-off point. I looked at all pitchers who averaged between 190 and 210 for three straight years (i.e., had a total of between 570 and 630 IPs during three consecutive years) in the 2000s and then saw how many innings they had in the next year. I came up with 136 pitchers, and they averaged 170 IPs in the next year — a good 30-ish innings fewer than they had averaged over the previous three years. That is aging and reversion to the mean. Lester is younger than this group on average so maybe you can expect him to have a few more than 170 IPs, but that number might be a safe guess for how many innings he will throw.
But maybe you are supremely confident in Lester’s talent and durability. Say you are sure he will have over 200 IPs and an ERA under four. How many games should we expect him to win? I found 188 pitchers who meet these criteria in a season since 2001. How many wins did they average in those seasons? Fifteen. Even pitchers who pitch well over a full season do not necessarily get near 19 wins (the lowest total was Brandon Webb’s 2004 season with just seven wins in spite of his 208 innings at a 3.59 ERA).
Sure, the Red Sox are an excellent team and will give Lester way more support than the 2004 Arizona Diamondbacks did Webb. But wins are fickle: Matt Garza won only eight games for the solid 2009 Tampa Bay Rays when he pitched 203 innings with a 3.95 ERA and last year Tommy Hanson won just ten games on a good Atlanta Braves team while pitching 202 innings with a 3.33 ERA.
The point is you cannot assume that Lester is going to have another 200-inning season, and even if he does, you cannot pencil him in for 19 wins (look at his 2009 season: 203 innings, a 3.41 ERA, and just 15 wins). That is why Bill James, who I am sure loves to watch baseball, projects 14 wins for Lester.
Dave Allen's other baseball work can be found at Baseball Analysts.
Interesting post. So why does Bill James project Chris Carpenter to win 18 games?
More IP, more GS, significantly lower ERA.
Yeah 18 seems high to me for Carpenter, and I think that I would take Lester to have more wins than Carpenter. The point of the post is not to say that I think James’ projections are gospel, but that 14 wins for Lester is not an unreasonable projection. And that I think 14 wins is a much better projection than 19 wins.
“But maybe you are supremely confident in Lester’s talent and durabilit. Say you are sure he will have over 200 IPs and an ERA under four. How many games should we expect him to win? I found 188 pitchers who meet these criteria in a season since 2001. How many wins did they average in those seasons? 15. Even pitchers who pitch well over a full season do not necessarily get near 19 wins (the lowest total was Brandon Webb‘s 2004 season with just seven wins in spite of his 208 innings at a 3.59 ERA).”
How many of those pitchers had the best offense in the league and a plus defense behind them, plus had their sub-4 ERA in a hitter’s park? That paragraph was extremely misleading… and borderline dishonest as a journalist.
14 wins seems solidly 2 wins off. There’s no way he’s as likely to end up with 10 wins as he is with 18 wins.
I would happily take Lester under 16 wins in 2011. I think he is more likely to win 15 or fewer games than 17 or more. Sorry that you found my logic ‘extremely misleading’ and ‘borderline dishonest’.
“There’s no way he’s as likely to end up with 10 wins as he is with 18 wins.”
That’s not what the projection is saying. It may be saying that p(>14) = p(<14), but it's definitely not saying p(18) = p(10). Take it to its logical extreme: do you think that by projecting 14 wins, BJ is saying that Lester is just as likely to win 0 games as he is to win 28? I doubt it, because winning 0 (preseason injury) is much more likely than 28.
Hi guys, you are missing the point here – I will say it differently.
This paragraph is wrong. Either you completely missed the gaping hole in your logic, or presented your “information” knowingly, hence the bit about “misleading and journalistically dishonest”. I was giving you the intellectual benefit of the doubt, which was the wrong thing to do, it seems.
“But maybe you are supremely confident in Lester’s talent and durabilit. Say you are sure he will have over 200 IPs and an ERA under four. How many games should we expect him to win? I found 188 pitchers who meet these criteria in a season since 2001. How many wins did they average in those seasons? 15. Even pitchers who pitch well over a full season do not necessarily get near 19 wins (the lowest total was Brandon Webb‘s 2004 season with just seven wins in spite of his 208 innings at a 3.59 ERA).”
If this is true, then 14 wins is CLEARLY an incorrect forecast (if you assume 200IP and a sub 4 ERA, as you state), as the 2011 red sox team is far, far superior to LEAGUE AVERAGE, which is basically the sample you have pointed at.
Yes?
I’m not saying I disagree with the heart of your article, but like so many pieces on FG, they idea is a good one, but they are tragically ill explained and argued for, and stats are butchered in an attempt to make a point. If this were english class (or maybe stats class) the teacher would have a red X through half of your paragraphs.
The thesis is fine, but some of the evidence used to back up your point was shoddy.
The Duder,
Your responses miss the point. Your assumption is that because Boston has a better than average offense, Lester is highly likely to get better than average run support. That is not necessarily the case as run support for different pitchers on the same team often varies drastically from year to year. Furthermore, the article did cite two examples of good pitchers from good teams getting very low win totals. One was Tommy Hanson from last year, and he shows the luck of run support. While the Braves score the sixth most runs in the NL last year, Hanson’s run support was horrible; he had a RS/9 IP (run support per nine innings while he was pitching) of 3.4 runs last year. Hence, he was 10-11, with a 3.33 ERA and 202.2 IP. Teammate Derek Lowe on the other hand, was much worse, with an ERA of 4.00 and 193.2 IP. Yet, Lowe finished 16-12. Why? He had a RS/9 IP of 4.5, and 8 wins in non-quality starts.
I HEAR when you CAPITALIZE things it makes your argument STRONGER and more MEANINGFUL….but you forgot the explanation points!!!!!
14 wins is not an absurd forecast, especially when the win stat is not all that meaningful to begin with. Would folks have projected Beckett at 6 wins last year or 12 wins in ’08 (coming off of his 20win, 3.27 ERA 2007 season)?
An injury which causes him to miss a few starts, a blown lead or two by the bullpen and a tough loss or two where he doesn’t get run support and it’s difficult to get to 18-20 wins. With a full season of 33-34 starts, a pitcher is likely going to get 6-10 no decisions. That leaves somewhere around 23-28 decisions (thus winning 19 is not easy). If a pitcher misses a couple of starts then the 15 win range is a good outcome.
“That’s not what the projection is saying. It may be saying that p(>14) = p(<14), but it's definitely not saying p(18) = p(10)."
Just to expound on these two point a little further.
First, as Erik states, it only MAY be possible that P(14). It depends upon what the methodology James (or rather, the system now bearing his name) uses. A median is obviously a common measure of central tendency, but it’s hardly the only one.
Second, some distributions are symmetric, including the normal– but many are not. In fact, in my experience most non-Monte Carlo models based on count data do not feature symmetry.
“journalist.”
Someone’s a little confused about what a journalist is.
“14 wins is not an absurd forecast, especially when the win stat is not all that meaningful to begin with.”
Also not absurd because 14 wins is a lot. lol
AL East vs. NL central’s probably got at least something to do with it too.
As well as the bullpens? I’m not really sure which bullpen is projected stronger but I do know that Papelbon is liable to blow a few saves.
We have lots of projection systems out there besides Bill James. If you plug the assumption of 200 IP into these projections, what do they predict for wins? What does the probability distribution look like?
On a related note–will there be/can you add a Bill James tab on the projections page?
Erik,
Not sure about your second question, but I am sure David Appelman will get to it soon. Your other question is interesting I will check that out.
I’m pretty sure that they can’t put a Bill James tab on the projections page.
Al–I assume you mean due to copyright/licensing issues. FG has the BJ projections on the individual player pages. It would have to be an odd license to allow that but not allow showing/sorting the projections for multiple players simultaneously.
That IS the case, though, Erik. He allows his projections to be viewed on the individual pages, but not on the projections pages, where they could be easily downloaded. I believe they are fully available at his own site if you pay. I’m not sure how they came to that agreement, but David Appelman has explained this in other comment threads in the past.
Thanks Newcomer, right you are: http://www.fangraphs.com/blogs/index.php/2011-bill-james-handbook-projections/#comment-331878
Thanks for the link and clearing this up, Erik.
With that offense 14 wins does seem pretty low considering he is expecting him to start 31 games with a 3.53 ERA.
vs mostly AL East teams.
And Okajima’s back
He does project Sabathia 18 W though, despite the Red Sox being projected superior in offense and defense.
Though he has Sabathia with 32 additional IP, 3 additional GS, and a 0.11 advantage in FIP.
ERA is more important for predicting wins than FIP.
3 additional games started, 4 additional wins? doesn’t jive. Like I said maybe 16 for Lester 17 for sabathia.
Sabathia has a 0.21 advantage in projected ERA. Also, Sabathia is projected for almost 0.5 additional IP per GS, which presumably would result in a greater yield of W/GS.
I’m not saying that these differences necessarily account for a 4 W difference, but that difference doesn’t just come out of thin air.
If 14 wins is projected that means roughly 19 is as likely as 9. I disagree. I could handle a 16 projection or maybe a 15. 19 being as likely as 13 and 12 respectively.
Obviously he can be below that, but if you project that many starts and that ERA, an appropriate win number is higher than 14.
I very strongly oppose projections in general though. And I do believe that playing time is underprojected. Playing time should be projected higher than it actually is, because injuries are very unpredictable.
They usually disclose they are not to be used for individual players though, only teams. So that makes it more reasonable. I.e if Youk doesn’t get hurt maybe someone else will to cancel it out.
“If 14 wins is projected that means roughly 19 is as likely as 9. I disagree. I could handle a 16 projection or maybe a 15. 19 being as likely as 13 and 12 respectively.”
What is “projection” exactly, the average value of some expected probability distribution? In that case the distribution would have to be normal for this to be true, right? I think the likelihood of Lester getting hurt and ending up with only a few wins is higher than that of him getting win totals above 25.
Some projectors could probably comment in more detail, but I suspect that they are aiming for either the most likely probability (meaning p(14 wins) is the largest p(wins)), or they are aiming for the median probability (meaning p(>14 wins) = p(<14 wins)). It definitely does not mean that the probability distribution around the projected value is symmetric (meaning that it does NOT mean that p(14 + n) = p(14 – n)).
“If 14 wins is projected that means roughly 19 is as likely as 9. I disagree. I could handle a 16 projection or maybe a 15. 19 being as likely as 13 and 12 respectively.”
You are assuming that Bill James’ win curves are equally distributed. I’m not saying they aren’t, because I am not sure. But do you know?
You have to worry about injury too. A Tommy John surgury in the middle of the season would definately make 14 optimistic. There is no similar, “up side” possibility where he, i.e. starts one in three games for his team and gets to 31 win.
“If 14 wins is projected that means roughly 19 is as likely as 9.”
Patently untrue – see Erik’s post above. Just because he may be saying p(>14) = p(<14) does NOT in any way imply that p(9) = p(19). Outcomes aren't uniformly or normally distributed around the mean result, in other words.
im no expert but it seems these predictions “underrate” wins for p and home runs for h, just the nature of the tool. it’s rare to see a guy projected over 30 hr – this goes for MARCEL too IIRC
Cole Hamels last year (anecdotal yes but informative):
IP W L K BB ERA WHIP
208.7 12 11 211 61 3.06 1.1789
Hamels more clearly:
IP: 208.7
W: 12
L: 11
K: 211
BB: 61
ERA: 3.06
WHIP: 1.18
Peripherals very much inline w/ Lester.
Granted, Hamels is a good example for how a great pitcher with a good offense can still get the shaft in terms of wins, and further granted, most of us here don’t really care much for wins, either projected or actual.
But it’s worth noting that while Hamels played for a good offense (.745 OPS, 4.77 R/G, both top-five NL) that projects to probably get worse given the departure of Werth, Lester played for an excellent offense (.790 OPS, 5.05 R/G, both top-two in MLB) that projects to improve (or at least stay the same) with better health and the additions of Crawford and Gonzalez. It seems like a 16-17-win projection makes far more sense than a 14-win projection in that case, and obviously a 12-win scenario, while not impossible, is highly improbable, assuming Lester performs up to his projected peripherals.
But really, who cares? Wins are always “too low” for elite starters in these systems. I don’t even bother looking at them, not least because wins are fairly worthless as a stat anyway. It just shows that Bob Ryan hasn’t been paying attention when he complains about it now.
Dan Shaughnessy is a douchebag.
To be clear, Dan Shaughnessy never said “Bill James doesn’t even like to watch baseball” (to the best of my knowledge), but Sullivan said that is something Shaughnessy would say.
Yeah, but he’s still a douchebag.
There are no 19 win pitchers. Only six instances of a pitcher averaging 19 wins over so much as three consecutive seasons have even occurred since the strike:
2008-10 Roy Halladay (58)
2008-10 CC Sabathia (57)
2000-02 Randy Johnson (65)
1999-01 Randy Johnson (57)
1998-00 Pedro Martínez (60)
1997-99 Pedro Martínez (59)
So when you say “no 19-win pitchers,” you actually mean “two 19-win pitchers.”
I mean that there are no 19 win caliber pitchers currently in the major leagues. The “19 game winner for the foreseeable future” tag is reserved for those of the Johnson/Martínez/Maddux mold, if even then.
Wait, what?
So your points are:
1. There are no 19 win caliber pitcher currently in the major leagues.
2. Roy Halladay has averaged over 19 wins from 2008-2010.
Makes perfect sense to me.
Roy Halladay AND CC Sabathia.
I understand that it’s rare and done only by the best of the best starting pitchers. But it’s weird to say there are none, then post a list that includes two who are currently in the midst of a string of 19-win seasons (or at least averaging 19 wins per season for the past three years and counting).
No 19 Game winners in the Major Leagues.
Roy Halladay has won 19 games in three straight years.
So, Roy Halladay does not pitch in the Major Leagues.
Roy Halladay pitches in the National League.
So, the National League is not the Major Leagues.
Maybe AAAA?
I refer to the caliber of the pitcher, although it’s still just one man’s opinion. No pitcher in the fan balloting currently has more than 18 wins projected for 2011, let alone the years beyond it, and these projections tend to be optimistic.
The pitchers that are of 19 win caliber are highly uncommon, the ones that actually win 19 much less so.
Um, reply FAIL!
Why not just combine Johnson’s and Pedro’s two (overlapping) streaks into four year streaks? It looks, at first glance, misleading, like it is. Minor nitpicking, though.
Still, because pitchers are only responsible for part (half?) of their win, in that the impetus always lies on the offense to score some runs just as much as it does on the pitcher to prevent them, it’s impossible to really say that a pitcher has a “true talent” level for wins when they don’t have full control over whether they get them or not. A pitcher must either be historically great (Halladay, Johnson, Pedro), pitch for an offensive powerhouse (Sabathia, Halladay ’10), or some combination of the both, to reliably (I use the word loosely) win that many games.
Lester isn’t quite good enough to be projectable as a 19 game winner, even considering the great offense he’ll pitch in front of.
There are at least two correct projections for players.
One looks like the season of a healthy player. Guys hit 45 homers, win 19 games, pitch 225 innings and get 650 AB. The problem is that it overprojects hitters by about 10 percent and pitchers by 15 percent, which is the amount of playing time any random group of regular players will lose from one year to the next because of all the reasons players lose playing time.
The other regresses projected playing time to the mean and 200 IP starters end up with 170 IP in their projection, with all the havoc that creates for their counting stats. The problem is that nobody is projected for 40 HR or 18 wins, but the projected stats more accurately describe what any random group of regular players will do than the first set. They just don’t look like the stats of the players we know.
All internally consistent projection systems choose a spot on the spectrum between these two approaches, and each can be useful, depending on what you’re looking for.
I think this article points out how the “Win” statistic for pitchers has so many variables involed and is really not a good metric for evaluating the effectiveness of pitchers. I thought with Felix winning the Cy Young despite “only 12” wins was a step in the right direction, but I guess the boobs at ESPN want to keep looking at old antiquated metrics as gospel. I think since the advent of the modern bullpen (i.e. 9th inning closers, setup men, and specialists) “Wins” for starters are so much more difficult to achieve and I largely think this is a reason why it is entirely probable that we won’t see another 300 game winner. There is so much luck involved with winning 15 or more games it’s insane. It’s not as if we’re in the pre-1980s when starters were giving you 7,8 IP on a consistent basis.
^^ This. Who the hell cares how many wins some player is projected for when it’s such a miserable metric to measure individual pitchers’ performances to begin with.
I generally take the rate projections more seriously than the playing time, win-loss, and counting stats projections. The rate stats, like FIP, ERA, OPS, etc., are based on individual performance projections. The other stats depend on so many uncertainties (ranging from team performance to injuries to managerial decisions) that they should be considered almost a “just for fun” projection.
As for the source of this discussion, we are now at the level of evaluating a talk-radio critique of Bill James?
I initially read it that way too. Certainly if you say “duh, if you take injury risk into consideration, of COURSE he won’t project 19 W,” it doesn’t make for a very interesting article. But what Dave wanted to show was that even if you take a best-case-scenario for IP/GS, as a talk-radio type person might, you STILL probably aren’t going to get 19 W even if you are Jon Lester. The real meat of it I think is this:
“But maybe you are supremely confident in Lester’s talent and durabilit. Say you are sure he will have over 200 IPs and an ERA under four. How many games should we expect him to win? I found 188 pitchers who meet these criteria in a season since 2001. How many wins did they average in those seasons? 15. Even pitchers who pitch well over a full season do not necessarily get near 19 wins (the lowest total was Brandon Webb‘s 2004 season with just seven wins in spite of his 208 innings at a 3.59 ERA).”
Maybe Dave buried the lede, but it’s still an interesting perspective IMO.
These guys are dinosaurs. What they don’t realize is that no one cares about how many wins John Lester is going to have, and there are very few front offices that will evaluate him in that manner.
These guys think they are so smart, but the truth is that the guys they are covering and commenting on, are laughing at them.
14 Wins projected because the Sox are going to match him up with Varitek every time out.. haha
Sorry to interrupt serious debates.
Carry on
wins are hard to specualte
i’m more confused by things like why guys like Will Venable, a nifty sleeper, get jipped in the projected plate appearances
http://ducksonthepondkid.wordpress.com/2011/01/12/searching-for-stubbs-will-venable/
i must agree with the title too- sometimes it seems like his projections really sell players short, even for modest expectations
-w
This is another example of why projecting stats is so difficult: too many variables, too much noise.
Much easier to examine underlying skills and determine worth that way.
1,000,000 simulations will take care of the noise. For pitchers’ wins, though, there are tons more variables to work with than just the pitchers’ stats. I thought most folks reading a site like this understand the relative uselessness of pitchers’ wins/losses anyway (I see most do; the Boston talking heads/writing hands don’t) . Anyone projecting them should be suspect: they’re probably providing false info for their fantasy opponents to read!
If someone cannot tell me what the +/- is for a projection, then I’ll assume they don’t know what they’re doing. Actually, I’ll assume that they don’t know that they don’t know what they’re doing. A +/- tells me that they know they don’t know.
FTR, I don’t like watching baseball. At least, not live in front of a TV. Going to game > watching on TV delayed so I can speed through 80% of it > watching on TV live.
I’m not sure how BJ projects wins (since they are so correlated to factors out of the pitcher’s control). I am surprised that he would come up with only 14 though.
If you take Lester’s defense independant stats from 2010 and project them forward, then combine those with the Red Sox’ projected defense and offensive run production, I’d be surprised to see 14 wins as a result.
Yeah if you did that you would get more than 14 wins. But you are missing a critical step: regress Lester’s defense independent stats and innings pitched to the mean.
Actually, when I said project them forward I meant with regression, I didn’t mean carry the 2010 numbers forward. I should have been more precise with my words. I think even with reasonable regression, he would likely come out with more than 14 wins.
“Actually, when I said project them forward I meant with regression, I didn’t mean carry the 2010 numbers forward. I should have been more precise with my words. I think even with reasonable regression, he would likely come out with more than 14 wins.”
In terms of FIP, Lester’s 2009 was nearly identical to his 2010 for a team that scored 5.38 R/G (i.e., comparable run support to what he can expect in 2011, if not a bit better). Yet he won only 15 games that year rather than the 19 he did in 2010.
Factor in the chances that he won’t make 32 starts or will be less effective in 2011 than 2009/10 (which is greater than the chance that he’ll be more effective)–i.e., regress toward the mean. It’s pretty clear to me that his expected win total would be much closer to 14 than 19 or even 16.
Walt, If he won 15 in ’09 and 19 in ’10, I can’t see how you would expect 14 in ’11. Small sample, but regression to mean seems closer to 17 than 14 in ’11.
The point of bringing up 2009 was to illustrate that he can be as effective as he was last year and still win “only” 15 games even on a team with a great offense. In other words, there’s a tremendous amount of “luck” associated with a pitcher’s win total in any given year (especially considering that the 2009 Sox scored considerably more runs). So this stochastic element introduces a good deal of variance in itself.
The regression to the mean occurs not entail some arbitrary discounting against some convex combination of his 2009 and 2010 win totals, but rather what one should regress is the games started or innings pitched. He’s managed to pitch a little over 200 IP over the past three years by staying healthy and being effective. Neither is certain for 2011 (for Lester or anyone). Missing just a handful of starts would easily compromise his chances at accumulating wins. Just using some very rudimentary methods, it’s easy to show how this might happen with a simple example:
In 2008-10, Lester accumulated 73 decisions (wins or losses) in 97 starts (~75%). Of these games, he won .667. Note: both of these rates are among the best in the game; also, it’s just a nice coincidence that both these rates work out nicely).
There is some non-zero probability that Lester will start fewer than 32 games next year that is greater than the probability that he’ll start more. I don’t know what it is, you don’t know what it is, Bill James doesn’t–no one does. But we know that it exists. For illustrative purposes only, let’s say that he starts 4 fewer games but maintains his same level of effectiveness in 2008-10 (75% decision rate, .667 winning percentage). 28*.75*.667=14
Of course, that’s just an artificial illustration, but it shows the effect of what missing just a handful of starts can have on a pitcher’s win total even if he remains as effective in his remaining starts.
Now consider that there is some density function that assigns probability to him missing 32 starts, 29 starts, 28 starts, etc. While individually these probabilities are all quite small, cumulatively they will drag down the expected outcome.
Finally, there is some non-zero probability that he will lose not just starts, but effectiveness. While obviously number of starts and the two measures we’re using for effectiveness (decision rate and winning percentage) exhibit some multicollinearity, this further depresses his expected win total.
In summary, there are three main, non-mutually exclusive sources of variance to consider:
– the stochastic element (i.e., luck)
– potential loss of health (i.e., fewer starts)
– potential loss of effectiveness (i.e., lower decision rate and/or winning percentage)
If you build an objective, well-designed model (or have some experience working with one) based not just on Lester’s performance but across a large sample of pitchers across time, an expected win total of 14 is perfectly reasonable.
I forgot to add…
A natural response to my last post might be to ask, well what happens if his effectiveness (again, as measured by either decision rate or winning percentage) improves?
While I acknowledge that there is some chance of that happening, the likelihood is relatively remote. A 75% decision rate and .667 winning percentage are essentially close to what we can consider the upper-bound of reasonable expectation to (in any given season, there’s a small chance that a pitcher can exceed those rates, but only because of the stochastic element). As such, while there is some potential scenarios wherein his decision rate or winning percentage might improve, it’s relatively minuscule compared to either decreasing.
Similarly, his chances of starting more than 32 or 33 games are slim (no pitcher has started more than 33 games in Francona’s tenure as Red Sox skipper).
So setting aside all the technical reasons for regressing to the mean, it’s pretty clear why the uncertainty that does exist is overwhelmingly negative for all the aforementioned reasons.
FFFFan,
that is the wrong mean to regress to. You need to regress to the league average mean, not his personal mean.
“This year we’re hoping to get everyone’s win-loss record projected right. Last year, we were exactly right on 8 of 387; that’s a start. It’s a cornerstone.”– Bill James Handbook 2011.
Sounds a lot like self-deprecation. Almost to the point of, “Don’t bother with these.”
I’d like to see the standard error.
You guys all really think that Bill James himself sits down and runs these projections personally? And that he doesn’t just have a team run out projections and slap his name on it for book sales?
Do you also still leave cookies out for Santa each Christmas Eve?
Stop saying “he” projects…
Well Santa does deliver 2.7 more presents above replacement when compared with other holiday folk heroes, so why not leave cookies?
Love the end quote, gold, pure gold.
James (he puts his name to them… he gets the credit / blame) are run through Baseball Info Solutions. They tend to appear conservative because fans look at their team’s projections through their hope glasses. Fans over-rate their own teams and object when others don’t. Ironically, James works for the Red Sox…
It’s clear where the Bill James projection system is wrong. After the Red Sox spend-a-thon this offseason, Red Sox fans and media members have declared the Sox the “greatest team evuh,” a lock to win 120, and in order for that to happen, Lester needs to win 30 games. Lester at 14 doesn’t fit their meme.
If one of the best pitchers in the league, a durable one with a great ERA, supported by perhaps the league’s best offense, is projected to win 14 games because of the fickle nature of wins, what’s the point in predicting wins at all?
I’m not saying the 14 figure is unjustified (though I’d probably give him a couple more wins, if pressed to pick a number). It IS justified, but it’s a low number because if we want to be more ‘correct’ more often we basically have to jam every pitcher’s win projection into the middle of the curve. Projecting 14 wins for Lester tells us far, far more about wins than it tells us about Lester. I guess this could be a tool for fantasy baseball players to use. Then again, maybe the fantasy players that can’t figure out what peripherals are more likely to yield more or fewer wins deserve to lose in their leagues.
All this talk of Lester’s wins is kind of a waste of time, but I guess that’s the point.
Sabathia goes a complete 7 or more about 65% of the time, Lester, 09 Garza, and 2010 Hanson went 45% and under. Intuitively I would say Sabathia, who goes deeper into games more frequently, would be more likely get wins (because he is good).
Lester goes less than 6 complete innings under 20% of the time, 09 Garza and 2010 Hanson went less than 6 more than 30% of the time.
Re: The title comment …Just a breeze from the past … One of my favorite Bill James pieces from the Analyst days was “Why I Love Watching Danny Jackson Pitch”, which included a beautiful description of Jackson’s physical mechanics … Don’t get that stuff from the spreadsheets …
I think 18 wins for Carpenter is off, but 14 wins for Lester is reasonable. If 14 win projection means P(w>14)=P(w<14), then maybe it is still a tad too high because pitcher missing time with injury is common, being healthy for 33 starts with high quality is not.
I gave 3 minuses to the ratings. Two were to Duder and one on Erik himself. Shame on you Duder!!! To all, we’re all in the business (very loosely applied word) of projecting wins. We choose Lester over other pitchers with similar era and k projections as he pitches for the Red Sox. He’ll get more wins. And wouldn’t we reduce our projection if he were to be traded to the Royals? Shoot, we’d raise our wins projections on the Royals. Erik, (I’m writing with little knowledge of statistics here), don’t you statisticians/sabermaticians work on averages/means? Why would you lower the win projection on Lester based on the possibility that he’ll get less run support than other pitchers on his team. Big minus on that post. Overall I like the article a lot. I’ll take the over on 16. One big plus to Sox27 on Santa thoughts! Like FG and visit daily.
Typical Boston sports media.
Hung up on something that pisses them off, not realizing that James is still projecting out Lester like he is an All Star.
But zOMG wins!!!!
Of course if Lester does win closer to 19 games, they will use this to deride James, and if he comes closer to 14, they’ll either pretend it didn’t happen, or cite things like injuries or bad luck (which is why James would low ball projections to begin with).
The point of this article isn’t “is Jon Lester going to win 14 games or not?”. Who cares how many wins he finishes with anyway? The point is that these 2 talking heads were irresponsible in bashing Bill James’ projections when they didn’t even bother to do any research to determine whether 14 wins was a reasonable projection or not. Not only did they bash his projections but they also bashed him personally without regard for any factual evidence whatsoever.
Now, carry on with the great “Lester’s definitely going to win 14 games” argument.