THE BOOK--Playing The Percentages In Baseball

I’d make a couple notes regarding the hypothesized effects of steroids.  First, there’s the idea of the steroid “high” - the ‘roid-rage type of increased energy.  Your A/B analysis should catch that sort of impact since you look at before/after when he began taking the drugs.

There’s also the muscle-building aid.  If my college strength coach is correct in saying that it takes 3 weeks for the effect of a training session to improve strength in a muscle, then you should actually look at 3-4 weeks after he first took the drugs and use that as the line between before/after.

Last, there’s the idea of aiding in recovery (sort of similar to the previous paragraph).  I think your analysis would have caught this as well.  If he had weathered the stretch run in 1998 better than he had in 1997, that might be evidence of something aiding his muscle “recovery”.

But playing devil’s advocate… what evidence is there that the steroids didn’t simply mask a decline that he otherwise would have had at the end of 1998?

I thought I remembered Book of Shadows having an interesting analysis of Bonds’ performance as matched up against the drug calendar they found.  Essentially, he performed better during the 3 “on” weeks each month, and worse during the “off” week.  Could be a placebo effect, but it was interesting to read.  I always thought the purpose of steroids was to help you exercise muscles more, and thus make you stronger down the road.  I didn’t realize they potentially have near-term effects as well.

Your last paragraph was in Zumsteg’s book, and I debunked that in my blog here.  You can do a search for it.

I also was not arguing on whether there was an effect or not.  I was chiding the media for saying “6-6… then 14-0?  after he talked to him?”.

And clearly, the biggest beneficiaries of Roger meeting Brian were the Jays hitters!

***

Trying to find when/how/if steroids affects someone (especially when a sizeable portion of your opponents are also doing it) is as fruitful as looking for a clutch hitter.

If you look at the three factors to a player’s performance:
1. his genes
2. his work ethic
3. his chemical dependence

Is there any doubt that Clemens’ owes upwards of 90% of his success to the first two?

For most players, they owe 90% of their success to just the first one.  Guys on the juice is Clemens au naturel… when he’s got a 102 fever.

Like I said elsewhere, I really don’t care at all who is or is not juicing… as long as everyone is aware of, and accepts, the landscape.  Players have been mute, and still are, through all this. 

There should be a Chris Chelios revolt going on here, demanding of their union for weekly testing, and severe penalties.  Baseball players are p-ssies compared to Chelios.  They are complicit, as are GMs and owners.

Found it, thanks.  One suggestion I’d make to you or mgl or whoever is the webmaster among you guys, is to alter the code of your blog such that the title of a given page is the title of the post.  When one dos a site search, all of the page titles come up the same, and it’s tough to find what you’re looking for.  Plus, if you alter the page title to be equal to the post title, or something equally distinct/descriptive, you’ll get more traffic from the search engines.

I sort of care whether people are juicing or not.  I find it fascinating that Williams could hit the red seat HR about 525 feet, and yet Ortiz can’t get anywhere near that.  I find it fascinating that Gibson could hit a HR with no legs, and still hit the ball as far as a typical HR from someone like a healthy Johnny Damon (who is pretty bulked up).  I don’t understand how skinny (relative to today’s players) guys from the 70’s could hit a straightaway HR to center in Fenway - well over 400, and at least 15 feet high… when a monster like Ortiz hasn’t ever hit one more than a couple rows deep to that same spot.  I care because I like to see hard work, as opposed to hard work and pill popping, rewarded.

But honestly, I’m getting a little numb to it all, and I certainly don’t care as much as I used to.  As the BP rant post mentioned… it’s really just a drop in the bucket compared to the real problems on this planet.

As I’ve said in the past you don’t “care”.  You have a slight preference.  You don’t care enough to start a M.A.D.D (mothers against drunk driving) campaign.  You don’t care enough to stop watching games.  You don’t care enough to stop spending money.  Exactly what is the extent of your “caring”, on an emotional level? Do you really CARE? 

And what if you learned that all players at one point or another have taken amphetamines?

I’ll rate my caring at 35.612 cares right now, whereas it was at a peak of 107.230 last year at this time.

At this time last year, I already had a file of Marcels, ZiPS, Chones, and PECOTA all mashed together, with about 5 different metrics by which I tried to rank players for fantasy value.  I had done probably a dozen mock drafts.  This year, I have a couple .csv’s on my desktop that I’ve sorted descending by HR.

Believe it or not, I do care enough to stop spending money.  With the exception of social circumstances (dates, catching up with old friends), I haven’t purchased tickets to a game for the intent of watching baseball since perhaps 2000.  Part of that is due to prices of tickets, parking, etc. (I live right outside Boston so we’re talking Sox games).  And part of it is due to just getting depressed at how I sort of feel like I’m paying to watch people get rich, instead of paying to watch people play a game really well.

Amphetamines bother me about as much as Advil.  Which is nonzero, but not as much as steroids.  I think I’m more against steroids because I think they “help more”, and because I have this image in my head of juiced up guys smacking around their girlfriends in a “roid rage”, which may or may not even be a true side effect of the drugs.  I have a general distaste for drugs in general (advil, alcohol, etc, included)… mostly because when I think of the hardships that our ancestors went through without that stuff (like how we came down through Canada to get to the US, through the one channel of land in Canada that wasn’t actually completely covered in ice), it makes me feel like a complete piece of garbage for popping a couple pills cause my tooth hurts.

I’ll put it this way - I’m here for the numbers more than I’m here for the baseball.  If there was a website like yours that analyzed Wii Tennis statistics in some way that helped me learn about math, I’d be just as inclined to visit that site a few times a day and comment on it as I do here.

Mike, you are one of the few that really care!

And it is an interesting bias in being ok with amphetamines but against steroids.  It’s possible that we harbor resentment for the “chick dig the long ball” (and I suppose strikeouts).

The bias can be that we don’t want a drug to help in a player’s primary skill (power), but don’t mind if it’s to help a less important skill (speed).

There should be a Chris Chelios revolt going on here, demanding of their union for weekly testing, and severe penalties.  Baseball players are p-ssies compared to Chelios.  They are complicit, as are GMs and owners.

I totally agree in principle, but think the players need to go farther.  They are responsible for electing those that represent them, so they need to replace them for failing to do what should be the number one thing on the agenda (making working conditions as safe as possible).

As the union (as “the collective bargaining representative for all current Major League Baseball players” that “works closely with MLB in ensuring that the playing conditions for all games involving Major League players...meet proper safety guidelines") has fought drug testing at every opportunity, they have horribly failed.  MLB has failed to provide a safe workplace for its employees, and as such is guilty as well, but it is for the union to put workplace safety at the forefront.  The players need to step up, for their own safety.  Money means much less when you are not able to enjoy it.

And do I really care?  About the issue of workplace safety and unions I do, as I worked at a refinery with my brother, and he (and 5 others) were killed because of a lapse in workplace safety caused by greed and hubris.

Quoted segments above are from the mlbpa website.

Interesting - I know next to nothing about amphetamines, but I always assumed they were more of a mental aid as opposed to a physical aid.  I just envision them working like some concentrated caffeine - I picture players using them after a long night, or late in the season as sort of a mood upper.  It’s not against league rules to down a few Red Bulls before the game, and I equate amphetamines to that (though again, the little I know about them is from health class in high school, which can hardly be called objective).  I’ve never heard of Olympic runners getting caught using amphetamines, and I’d assume they’re tested for them and they’d try them if they helped one to run faster. 

If they do truly help improve physical speed, I think you’d hear a lot of backlash against them if someone like Ichiro broke .400 or .406 and was a “known” amphetamine user, who beat out a lot of infield ground balls.  .406 is no 755, and certainly no 61… but I think there’d be some public outcry there.

You’re assumption is that clemens started in 98.  He had McNamee first inject him in 98, but he already had the steroids on him.

I’m willing to bet that he actually started off in the off-season of 96.  Do I have proof? No, but I personally believe his 97 season was enhanced by steroids.

If I accept everything that McNamee says is true, then I’m 100% convinced that Clemens did not use before the summer of 1998, or used very little.

I base this on his statement that Clemens needed help injecting himself.  To me, that red-flags him as a beginner.

As a diabetic who was once deathly afraid of needles, trust me, you get used to it very quickly.

The personal belief of #9 is based on… what exactly?  After the fact results of fitting the 1997 data to the data prior to that, in order to more corroborate the summer of 1998 data?

I just wouldn’t be shocked if he used in 97.  Clemens had the steroids on him already, someone else could have injected him.

We’ll never know the whole truth, but I doubt his steroid use was limited to merely three years.  If it is, I would be shocked.

I believe 98 was the year he averaged his highest strikeouts/9 total.  It’s also interesting to me that he had 14 games with ten or more strikeouts in a single game in 97, two more than his previous high back in 88.  He then again almost equaled it in 98 with 11.

Randy Hendricks released a statistical analysis on Clemens, according to the AP.  However, there was no citation for it.  I even bothered to write to Hendricks (no response yet).  Of course, I should have just checked-in with David Pinto:
http://www.baseballmusings.com/archives/024612.php

He’s got a quick take on it, plus the link to the report here:
http://www.rogerclemensreport.com/

Without reading the report, we already know what it’s going to say (otherwise, they wouldn’t have released the report!).  Hendricks is a lawyer, not a scientist.  The only question is if they did the report in a well-balanced manner.

http://sonsofsamhorn.net/wiki/index.php/Twilight_of_his_career

His Redsox “twilight” years.  The quote by Duquette makes it very clear that Clemens, Redsox, and twilight were meant post-1996, similar to the way that the Giants would have signed Bonds for his twilight years post-2004: still a great player, and wanting to capture the remaining glory that was expected to come. 

Basically, a guy’s “twilight years” is any year where he’s 35 and older.

I’m reading the report, only looking for bias in reporting the results.  It seems that the Adobe file is protected from Copy, which is ridiculous.  (Makes my life harder in quoting.)

I remember reading that Adobe has a weak security system, so let me try to work around that.  In the meantime, it seems that this report when through twenty versions, and was written (at least in part) by Aaron Harrell.

I’ll be back in a few…

Well, that was easy.  Print the file to something called “Microsoft XPS” (which you can then open with IE).  The driver was already installed on my machine (an XP).  Anyway, anyone looking to convert the PDF to something extractable, that’s how you do it. 

Note to you Hendricks people: don’t think you can prevent people with your “security” system.  All you do is annoy people.  Just republish the document with cut/pasting allowed. 

***

For this reason, the won-lost record is a product of all three of the basic elements of baseball: pitching, hitting, and fielding. While a win or a loss in a given game is typically more dependent upon the performance of the starting pitcher than that of any other player, due to the fact that he usually pitches the most innings, it is better understood as a measure of overall team performance. By contrast, a pitcher’s ERA is based much more on the pitcher’s individual performance.

Now this is a silly statement to make.  The author is absolutely right that W/L record is based on pitching, hitting, and fielding (because it is based on runs scored and runs allowed).

But, ERA, by the same token is simply the W/L record without the hitting.  That means, that fielding is also part of ERA.

I wouldn’t make the point, but since the author’s went out of their way to show pitching is one of three things that influence the W/L record, then they should show ERA has having two influences.

I agree with their general point though.

ERA Margin is a sound and reliable measure of the quality of a starting pitcher’s performance.  ERA Margin is the difference by which a pitcher’s ERA has exceeded or fallen short of his league’s ERA. This enables one to compare pitchers across the two major leagues on essentially the same scale. A more pure statistical approach to calculating ERA Margin would be to apply percentage differences and to then recalculate the figures on that basis.  But the raw differences are more accessible, easier to calculate, and more readily understood; and, the results are not substantially different.

Well said, except for the “more accessible”.  The “easier to calculate” is true (insofar that ANY two numbers is easier to subtract than divide, in anything).  Anyway, good job in putting this out there.

An analysis of two distinguished contemporaries of Clemens, Randy Johnson and Curt Schilling, shows that year-to-year variations in ERA Margin are to be expected throughout the career of a starting pitcher. 

Selective sampling.  If you are going to be fair, choose his contemporaries on the basis of what they did through their age33 season.  That probably includes David Cone, maybe Bret Saberhagen, Dwight Gooden, Orel Hershiser, Frank Viola, Dave Stieb, among others.

If I look at all pitchers born within 5 years of Clemens, through their age33 season, here’s the guys with the most IP:
http://www.bb-ref.com/pi/shareit/Y5XB

Schilling and RJ are at #34 and #50.

The report itself is well-done, so I’m just going to be picky:

Official run support figures are not available for the first four years of his career, through 1987. The box below the chart displays the year-by-year run support averages after 1987 for the relevant leagues. 

What does “Official run support” mean?  Run support is not an official category (i.e., it’s not covered in Rule 10), therefore, there is no official run support.  In any case, the league keeps track of all their games ever.  To state that you can’t get a run support number for Clemens is being silly again.

Run support can be calculated in a few ways, be it as total runs scored per game in which Clemens was the starter, or total runs score per 9 IP, or total runs scored per 9IP while Clemens was the pitcher of decision, or while he was still in the game, or whatnot.

Whatever the non-official official position of MLB (i.e., Elias) is, that could be calculated throughout history.  If you are going to calculate park factors as they did in this report, which is decidedly unofficial, I don’t see why they couldn’t do it here for the 3 missing Clemens season.

I’m being picky, I know.  That’s why we have blogs.

As noted, however, a starter’s won-lost record can be distorted in any particular year by the amount of run support that his offensive teammates give him and by the quality of the relief pitchers who pitch behind him.

At the beginning of the paper, of which I quoted, they said it was fielding.  Now fielding went away, and the bullpen makes its way here.

They all impact of course.  But, a bit sloppy.

Clemens was far from being in the “twilight of his career” or “washed up” in 1996, as some have speculated.

As noted, the term twilight here, an obvious reference to the Duquette quote, is inappropriate here.

The four sections in the above chart contain the game-by-game statistics for Roger Clemens for the period that comprises the apex of his career. The period begins on August 11, 1996 and extends through April 10, 1999.

Uh, is this a damning statement for those who want to extrapolate that if he did drugs in 98 then he must have done it earlier?  Or is this a supporting statement that if the earliest he could have done it is 98, then it’s really no different than 97?  The answer to that is predicated on the bias of the reader.

***

By the way, in the earlier link where I showed Clemens’ peers, you can further filter that list based on the best K/BFP if you like.  The key though is not to do selective sampling.

***

All-in-all, a pretty good sabermetric primer.  An enjoyable read.

If this is the kind of example of what we’d see in an arbitration hearing, then we can see that we are limited to what we can present an arbitrator.  DIPS for example seems to be off-limits (and would have been a great thing to show, as per my blog posting on top).

Regarding the Adobe issue, Foxit Reader is light years better--even letting you copy text directly. For anyone with Firefox, get Foxit Reader and the PDF Download extension.

Actually, just downloaded the report and the copy protection is there in Foxit, too. That is annoying.

I think if you keep your Windows XP up-to-date, XPS comes with it. 
http://windowsupdate.microsoft.com/

Otherwise, you can do a separate download here:
http://www.microsoft.com/whdc/xps/viewxps.mspx

If you want to download some free software:
http://www.verypdf.com/pwdremover/index.htm

This removed the security from the Clemens PDF document.

"If this is the kind of example of what we’d see in an arbitration hearing, then we can see that we are limited to what we can present an arbitrator.  DIPS for example seems to be off-limits (and would have been a great thing to show, as per my blog posting on top).”

One of the attorneys in my firm is one of the salary arbitrators.  I can’t mention specifics, but broadly speaking, this is in much greater detail for the individual player (here Clemens) than is in the usual arbitration submission.

Most arbitration cases present the particular player somewhat simplisticly based on fairly common stats (G, IP, W/L, H, BB, K, ERA, Sv for pitchers and their parallels for hitters). Then they use something akin to a similarity score for players who were in the same service class one or two years previous and then compare the requested / offerd salary to the average of the comparator class.  The fight is largely over whether the player should be compared to group A or to group B.

(As an aside, each of the panel members used to get the annual Baseball Register and are now given copies of the Bill James Handbook for the particular season.  They are instructed to review that if a player comparison comes up where the two sides have not already provided a statistical summary, but that is pretty rare.  The panel is not supposed to try to find other comparable players beyond those listed by one or the other party.)

That being said, there have been some presentations that have delved into more advanced metrix.  I don’t recall DIPS expressly being used, but plenty of times you hear about defensive support and I’ve even seen DER used. I recall one a couple of years ago where the team argued that despite the pitcher’s excellent ERA (and peripherals) he should not be compared to others with similar numbers because his usage pattern - demonstrated with a quasi Leverage or Wins Added method - reflected that he was not highly valued by his manager.

Hmmmm… this is interesting.  So, if I can convince Bill James and John Dewan to publish Leverage Index in their book, then it becomes part of the defacto standard?  And same for including BABIP (i.e., 1-DER) and wOBA?  Sweet.

If its on the main player data page rather than in the errata lists at the end of the book (player projections, pitcher batting, managerial tendencies, etc.) then yes.  The former pages are copied and included in the handouts like condoms at a hooker’s convention.

NY Times statisticians (including sometime reader of this blog Shane Jensen):
http://www.nytimes.com/2008/02/10/sports/baseball/10score.htm?_r=1&oref=slogin

Phil responds to the NY Times statisticians:
http://sabermetricresearch.blogspot.com/2008/02/clemens-report-criticism-misses-point.html

JC also responds:
http://www.sabernomics.com/sabernomics/index.php/2008/02/a-critique-of-the-clemens-report/

And the Hendricks brothers respond (through JC’s blog):
http://www.sabernomics.com/sabernomics/index.php/2008/02/official-clemens-response-to-the-ny-times-article/

I think Phil does some great critical thinking:
http://sabermetricresearch.blogspot.com/2008/02/at-freakonomics-today-justin-wolfers.html

Nice job by Nate to show what we should have expected from Clemens in 1998-2001:
http://www.baseballprospectus.com/article.php?articleid=7139

I will disagree with his opinion here:

There is no doubt that, as a piece of science, the Hendricks report is pretty poor.

Insofar that Nate and others are echoing my post 18 (selective sampling of pitchers), that’s about as bad as the report is.  Au contraire, the rest of the report is pretty good.  Furthermore, since Nate says, and I agree:

his longevity past the age of 40. But even that is not entirely without precedent. Although pitchers like Ryan, Schilling, and Johnson are not exactly typical examples, neither are they atypical, and they all showed up prominently on Clemens’s comparables list

And, as the Hendricks stated, that is exactly the point.  That what Roger did during the McNamee years was not unusual (and Nate here pretty much confirms that it was expected), and that what Roger did in his 40s was not unprecedented (i.e., he’s not such an outlier that it makes it impossible to believe), which Nate here is also confirming.

In any case, Nate’s last sentence is dead-on, and I wish all statisticians would just keep repeating it:

But where his statistical record is concerned, there is no smoking gun.

Waitaminute.  I just thought of something.  Nate’s 1998-2001 forecast was done on a year-by-year basis I will presume.

As I’ve shown, a pitcher in his late 20s will lose 10% of his innings per year (Rule of 10).  And yet, Nate is not showing such a drop.  If in 1998 he’s expected to pitch 219 IP, then the next 3 years would be, at most (if you apply Rule of 10): 197, 177, 160, for a total of 753 IP.  Nate is showing 795, which is 42 more innings that we can reasonably expect.  And he actually threw 847, which is 100 more than we expect.

The 753 IP would include chances for injury, sometimes catastrophic.

I’ll ask Nate whether he did a year-by-year forecast, or whether he did a 4-year forecast.

Clemens did have a very impressive group of comparables in Nate’s column. It certainly looked like a very durable bunch that PECOTA is regressing towards.

I don’t think Nate’s article disproves anything, or make clemens look any less suspicious.  It’s a biased study anyway, given that nate doesn’t seem to apparently care that In the Mitchell Report it specifically stated that Mcnamee did not inject clemens with steroids in 1999.  He was with the BlueJays that season.  In short, we have no evidence clemens was on steroids that season, and it’s arguable that clemens drop in performance was perhaps a result of not having drugs injected into his body during the season.

Nate, however, using PECOTA, didn’t take that into account.  If we believe his analysis without actually reading what’s in the report, you get the idea that steroids didn’t work for clemens at all since in 1999 he did not meet his expected projection, or even come close to it.  Perhaps the reason for that is a lack of steroids.