Pettitte's use of performance-enhancing drugs is instructive as he searched for a quick fix to a painful injury.

In 2002, after a trip to the disabled list with elbow tendinitis, Pettitte was injected with human growth hormone. He got back on the mound and finished the season with the second best earned run average of his career up to that point.

More cherry-picking ########...

Pettitte's 2002 ERA before "injection"--3.09

Pettitte's 2002 ERA after "injection"--3.30

Also, Giambi being the "big winner" is kind of a farce in that his $4 million salary in 2001 was before he was eligible for free agency. Obviously he was going to get a huge raise as he was MVP in 2000 and finished 2nd in 2001. This very "analysis" says Giambi began in PED use in 2001.

I am not sure how anyone can pinpoint the time Giambi, or any player, "began using performance-enhancing substances" as this study tries to assert. As I mentioned on another thread, M Tejada won the MVP in 2002 and according to the Mitchell Report, the first known time he used PEDs was 2003.

At least if you were going to go to the trouble of a high-school caliber study, you would mention these two facts, even if they are rather meaningless in a real study. Of course there is an agenda and not all facts are helpful.

Not to mention that (all else being equal) among any random group 50% is exactly the number you'd expect to have a better season in year n+1 than in year n. Assuming "a better season than last given a stable skill set" is a random process.

Not to mention that (all else being equal) among any random group 50% is exactly the number you'd expect to have a better season in year n+1 than in year n. Assuming "a better season than last given a stable skill set" is a random process.

This is most certainly wrong. MLB talent is not distributed in a bell curve. Way more poor players (or available poor players) than average, and way more average than great. Plus n+1 is a year older which would also cause some decline. Maintaining any performance level is always a victory of sorts, so 50% improvement is indeed improvement and not random chance, regardless of the cause of the improvement.

I came across this picked cherry yesterday...

McNamee said he started injecting The Rocket – at the pitcher's insistence – in June of 1998, on four to six separate occasions. At the time, Clemens, a member of the Blue Jays, was 5-6, clearly in decline.

But soon after filling his body with Deca-Durabolin, Clemens was a new man. He went 15-0 the rest of the season, picking up his fifth Cy Young Award.

His ERA sat at 3.50 when he was 5-6...after his next 6 starts (the Jays bailed him out on some loses) his ERA jumped to 3.71.

After 8 starts his ERA was back down to 3.49.

Repoz, as you know, Rohan Baichu thinks McNamee has some serious issues. The "clearly in decline" comment, even if made by the reporter was influenced by McNamee, it certainly does seem to shade him that way.

The article states that 33% improved the next season, and that number appears artificially inflated.
The first few I checked appear to be misrepresented.
Paxton Crawford first year in ML was 2000, so I do not know how his ERA+ increased from the prior season.
Kevin Brown has 2001 listed as 1st year after alleged use, but his ERA+ decreased in 2001.
Ankiel had 1 at bat season after alleged use, but somehow is included as batting positive.

Nothing to see here but typical mainstream media BS.

A true analysis will be very difficult.
1) The information in the Mitchell report is not overly reliable, particularly from start date of possible usage.
2) Good luck trying to find a control group, since the Mitchell report came from so few sources. The control group players may have used more steroids than this list. You could use a different generation, but that introduces other bias.
3) Removing placebo response, which may be the entire cause of any increased production, will be very difficult, indeed.
4) Lumping drugs together randomly. Why not include marijuana and cocaine? We have about as much evidence that they increase production as HGH, or steroids for that matter.

Or we could just do the stupid thing:
"JSonline: 50% of Ballplayers in Mitchell Report Improved After Link to Drugs"
"JSonline: 50% of Ballplayers in Mitchell Report Declined After Link to Drugs"

Do steroids work quickly enough that Clemens would turn around his season by starting them half-way through?

Plus n+1 is a year older which would also cause some decline.

Only if these were all older players taking steroids. I'd expect a 24 year old to improve whether he takes them or not.

why is 2001 the magical year? does anyone really believe that it's only in MLB that baseball athletes use steroids? the incentive to use them in college and the minors must be great.

The headline reminds me of a Dilbert cartoon where the boss is scandalized upon finding out that 40% of employee sick days are taken on Mondays and Fridays.

Or we could just do the stupid thing:
"JSonline: 50% of Ballplayers in Mitchell Report Improved After Link to Drugs"
"JSonline: 50% of Ballplayers in Mitchell Report Declined After Link to Drugs"

Exactly. Where are all the outraged columns complaining about the inconsistent quality of PEDs?

Steroids: 50% of the time, it works every time

3) Removing placebo response, which may be the entire cause of any increased production, will be very difficult, indeed.

Not to mention, that the whole point of steroids is to allow more intense workouts. I would assume that many players who start using steroids are also increasing their workout intensity, whether steroids are helping them or not. And presumably working out harder would help players improve their performance - again, whether steroids help them or not.

This is most certainly wrong. MLB talent is not distributed in a bell curve. Way more poor players (or available poor players) than average, and way more average than great. Plus n+1 is a year older which would also cause some decline. Maintaining any performance level is always a victory of sorts, so 50% improvement is indeed improvement and not random chance, regardless of the cause of the improvement.

This doesn't really matter as we're not talking about the overall distribution of talent. We are talking about individual players and their level of "true talent" (or "established performance") and whether any particular season fell above or below that level. All else equal, top players are just as likely to under/over-perform their level of talent as poor players are. (OK, that may not quite be true but it's surely close)

That's not to say this is necessarily a random process nor completely unpredictable (age, injury history, etc.) but yeah a reasonable null hypothesis is that 50% would "outperform" their true talent in any given season -- depending on how strictly we define "outperform" (e.g. statistically significant? in which case we'd expect very few).

Walt, I did a study of every player who got 300+ PAs in consecutive seasons between 1955 and 1998.

I'm planning on redoing it to see if there's a big change in the number of players who show "significant" improvement.

The reason for the quote is that it's not close to statistically significant. It's players who moved more than the standard error. And I called it "important" if they moved half that. (Using what Steve Mann called OBS, OPS with simple support for base stealing bolted on. No park adjustments, but also no Colorado players. Probably going to start with OPS+ next time, but I think this serves as a baseline)


Age # BA OBP SLG BA OBP SLG SI II SD ID
18 1 .250 .276 .346 .267 .307 .367 0 1 0 0
19 6 .269 .325 .411 .259 .311 .402 0 1 1 2
20 36 .269 .324 .407 .280 .332 .427 8 14 3 6
21 109 .269 .327 .401 .272 .332 .414 20 39 10 26
22 217 .267 .328 .406 .271 .335 .412 38 76 23 47
23 379 .268 .333 .409 .267 .333 .408 46 100 48 108
24 514 .267 .332 .409 .269 .336 .414 74 159 52 142
25 650 .272 .337 .415 .271 .337 .416 87 185 81 189
26 723 .272 .339 .418 .270 .337 .413 85 194 105 214
27 699 .272 .339 .420 .270 .338 .413 81 175 101 219
28 658 .272 .340 .416 .268 .338 .410 83 167 92 203
29 599 .270 .340 .414 .268 .338 .411 66 155 78 187
30 503 .271 .341 .419 .270 .340 .415 50 125 58 137
31 413 .274 .345 .426 .271 .344 .418 51 104 68 140
32 348 .275 .347 .423 .269 .340 .413 30 75 59 128
33 255 .276 .348 .427 .272 .345 .418 27 66 40 84
34 185 .278 .350 .430 .269 .340 .411 11 32 36 71
35 130 .276 .346 .428 .273 .346 .425 13 33 15 40
36 90 .281 .358 .443 .274 .346 .422 5 12 17 32
37 65 .281 .354 .428 .267 .342 .404 4 14 17 31
38 41 .272 .353 .425 .267 .349 .418 5 8 6 11
39 24 .274 .356 .442 .272 .355 .429 3 8 5 9
40 14 .274 .366 .445 .260 .354 .410 1 2 6 10
41 6 .260 .349 .396 .244 .323 .378 0 1 1 3
42 3 .271 .343 .397 .279 .365 .399 0 1 0 0
43 2 .286 .368 .394 .253 .347 .366 0 0 1 1
</pre>

Age = Age Jul 1st of 1st year (IE there were 723 players who had 300+ PAs at both 23 and 24)

This is followed by the average BA, OBP and SLG of the first year then the second year

SI = number of players who improved a significant amount. (See disclaimer. I know significant isn't the best choice of words)

II = number of players who improved by half that. Important

SD/SI opposite of SI/II.