Major league equivalencies
Sean Williams
more about major league equivalencies. Several people have said that
minor league stats, if properly interpreted, are reasonably good
predictors of major league stats. Well, I'd like to see some of them.
Could you give examples, such as what is/was predicted for Kendall,
Guillen, Giles, Morris, Hermansen, etc.?
Also, only hitters have been mentioned. Are major league equivalencies
accurate predictors for pitchers?
Thanks for any info.
Sean
Grativo
predicted the following using their projection system for 98.
I don't think pitchers can be reliably predicted.
Kendall
predicted for 98: .308/.378/.468
actual 98: .326/.388/.478
predicted for 99: .312/.374/.475
Guillen
predicted for 98: .291/.313/.431
actual 98: .267/.297/.418
predicted for 99: .282/.309/.432
Giles
predicted for 98: .314/.420/.580
actual 98 .253/.383/.438
predicted for 99: .267/.378/.465
Morris
didn't make a prediction for 98.
predicted for 99: .280/.337/.389
Hermanson
predicted for 98: .261/.346/.419
actual 98: .243/.315/.460
predicted for 99: .268/.337/.465
Womack
predicted for 98: .255/.319/.365
actual 98: .281/.324/.359
predicted for 99: .276/.318/.356
Kevin Young
predicted for 98: .281/.332/.444
actual 98: .268/.323/.487
predicted for 99: .280/.327/.482
The only projection above they completely blew was Giles, and I thought that
projecting Giles to hit for 1.000 OBP+SLG was completely looney.
Grativo
Eric Gaumann
If you want to feel free to ost some names. Martin Benjamin, Brown,
Brown and Brown, Sprague, Garcia.....
Cheers,
ERIC
Mark Jareb
OK, the numbers I show you for various batters are MLE's (Major League Equivalencies, a system developed by Bill James and now published by STATS) and DT's (Davenport Translations, a system developed by Clay Davenport and published in the Baseball Prospectus). MLE's are translations of minor league stats to the players' MLB organization's home park in that season i.e. if a batter plays in the Rockies organization in a year with very high offense his MLEs will be higher than someone who played in the Dodgers organization or if he put up those numbers in a year where offense was low in the majors. DT's are translations of minor league stats to a neutral park in a league where offense is set at a BA/OBP/SLG of 260/330/420 (prior to '97 the league baseline was typically set at the AL level of offense that season). DT's in the last 2 years have also been age-adjusted -- on a philosophical level of the purpose of DTs I don't like the age-adjusting and I also disagree with the reasoning behind using that methodology. So, the DT's I list are either those prior to ears in which the age adjustment was used and if recent ones de-age adjusted by me. Also, important with regard to Kendall -- DT's don't take into account HBP hence the seemingly larger drop in OBP for him.
Kendall:
raw# DTs
MLEs
93 A 276/325/352 227/260/280
n/a
94 A+ 318/406/437 245/317/321
n/a
95 AA 326/414/448 293/369/403
300/361/402
96MLB 300/372/401 292/352/392
n/a
97MLB 294/391/434 288/357/429
n/a
98MLB 327/411/473 326/388/478
n/a
Guillen:
raw# DTs
MLEs
95 A- 314/367/527 256/277/378
n/a
96 A+ 322/357/498 284/307/403
n/a
97MLB 267/300/412 259/289/403
n/a
98MLB 267/298/414 267/297/418
n/a
Giles:
raw# DTs
MLEs
93 AA 327/409/449 311/384/407
n/a
94AAA 313/390/479 301/374/432
296/366/444
95AAA 310/395/501 315/394/488
303/380/479
96AAA 314/395/594 319/397/550
299/375/550
96MLB 355/434/612 345/433/586
n/a
97MLB 268/368/459 273/378/475
n/a
98MLB 269/396/460 253/383/438
n/a
(note that the DTs for '98 for Giles look very screwy to me -- I think a mistake was made in that translation)
Morris:
raw# DTs
MLEs
97 A+ 306/390/470 284/347/413
n/a
98(T)AA 331/401/521 286/341/415
n/a
98(P)AA 331/419/523 289/354/444 305/359/464
Hermansen:
raw# DTs
MLEs
96 A 252/377/513 240/320/425
n/a
96 A+ 275/352/462 238/304/382
n/a
97 AA 275/373/478 238/310/393
249/315/417
98AAA 258/334/520 231/302/452
226/283/421
Mark
Grativo
>Brown and Brown, Sprague, Garcia.....
Al Martin
1998 predicted: .295/.363/.485
1998 actual: .235/.292/.365
1999 predicted: .275/.331/.419
This is a clear miss by the system. They predicted Al Martin to have a career
year at age 29, but instead he went into the toilet. The 1999 prediction
sounds about right to me...a bounce back from his miserable 98, but not enough
to make him a useful player again.
Benjamin
no prediction made for 1998
Prediction for 1999: .254/.293/.377
I'll bet anyone here $50 that if Benjamin gets more than 200 ABs he doesn't
slug .377.
Adrian Brown
no predictom made for 1998
Prediction for 1999: .260/.308/.323
"would the the perfect center field solution for the Bucs if he could just hit"
Emil Brown
Prediction for 1998: .227/.293/.333
Actual 1998: .282/.330/.397
Prediction for 1999: .279/.330/.404
I wouldn't have put much stock in the 1998 prediction because I don't think it
was based on enough ABs. Hard to predict how a young player would react to
getting picked in rule V draft and getting a year of his development
interrupted.
Sprague
1998 Prediction: .248/.315/.397
1998 Actual: .216/.262/.396
1999 Prediction: .224/.286/.415
Sprague hit 30 points lower than predicted, but other than that, dead on. The
system predicted Sprague to suck, and he was WORSE. This is what Cam Bonifay
thought was worth $1 million.
Garcia
1998 Prediction: .271/.312/.451
1998 Actual: .249/.304/.471
1999 Prediction: .249/.300/.454
Dead on. Anyone who thinks Sprague is a better option than Freddy Garcia at
five times the cost is either an idot or insane.
Grativo
ron.j...@ccrs.nrcan.gc.ca
sean...@discover.earthlink.net wrote:
> At the risk of starting another war of words, I am interested in hearing
> more about major league equivalencies. Several people have said that
> minor league stats, if properly interpreted, are reasonably good
> predictors of major league stats. Well, I'd like to see some of them.
> Could you give examples, such as what is/was predicted for Kendall,
> Guillen, Giles, Morris, Hermansen, etc.?
Please remember that an MLE is not a prediction. It's a translation
of what they did. As such any given MLE tells you as much as any
given season. The more years you have, the more accurate the estimate
you can make of a player's ability.
Having said that:
Kendall 1995 MLE .300 .361 .402 (Under-rates Kendall's OBP because MLEs
don't include HBP. For most players, no
big deal. For Kendall ...)
Kendall 1996 NL .300 .372 .401 Bingo. Unusally close.
MLEs don't work below AA ball, so nothing available for Guillen.
There is another method called DTs. This evaluates his 1996
at Lynchburg as being extremely similar to his 1997.
Giles 1995 MLE .303 .380 .479
Giles 1996 MLE .299 .375 .550
Giles 1997 AL .268 .368 .459
His career numbers are .284/.391/.485. Seems like the MLEs and
his results to date match up reasonably well. Yes, there's a
difference of 84 points of OPS between his 1996 and 1997. 1996
was his best year, 1997 his worst. You won't have to look very
far to find an established major league hitter with greater
variations. (This by the way is all that's claimed of MLEs. Not
that they perfectly nail a player - simply that they tell you
as much as major league stats do.)
Morris 1998 MLE .305 .359 .464
Ability level or fluke? Can't tell. It's substantially better than his
1997. Not that his 1997 was bad. It's quite clear to me that he's already
better than Mike Benjamin and Tony Womack and that he should have been
promoted to AAA last year by the Pirates. All that his 1998 numbers
really show me is that he was way too good for AA.
Hermanson 1997 MLE .249 .315 .417
1998 MLE .226 .283 .421
Garcia 1997 MLE .245 .277 .444
1998 MLE .237 .277 .468
Geez, who'd have though they already had Ed Sprague?
> Also, only hitters have been mentioned. Are major league equivalencies
> accurate predictors for pitchers?
Nothing predicts pitchers.
Nobody's actually done any work on translating minor league pitching -
the year to year variation even among established pitchers is huge.
It's my belief that you could do the equivalent to an MLE for pitchers
using a similar methodology. IE raise the ERA by 22% and adjust for
league and park context and you'd have the same range of accuracy
as for major league pitchers. IE not much.
--
RNJ
-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own
Bemulli
ron.j...@ccrs.nrcan.gc.ca w rote:
>In article <36E80EF7...@discover.earthlink.net>,
> sean...@discover.earthlink.net wrote:
>> At the risk of starting another war of words, I am interested in hearing
>> more about major league equivalencies. Several people have said that
>> minor league stats, if properly interpreted, are reasonably good
>> predictors of major league stats. Well, I'd like to see some of them.
>> Could you give examples, such as what is/was predicted for Kendall,
>> Guillen, Giles, Morris, Hermansen, etc.?
>
>Please remember that an MLE is not a prediction. It's a translation
>of what they did. As such any given MLE tells you as much as any
>given season. The more years you have, the more accurate the estimate
>you can make of a player's ability.
>
A couple of questions. First, is the formula to come up with MLE's widely
available? In other words, can any stathead come up with1999 MLE's or is he
dependent on the numerous Rotisserie League guides that plague us every year?
Secondly, you state above that MLE is not a prediction but a translation of
what a player did previously. If so how in the world can this system spew out
that Al Martin was due for a career year in 1998? Iirc, none of the resident
statheads around here were predicting anything of the sort for Martin last
Spring.
FWIW, my stance to the whole stathead phenomena is very similar to Paul
Rodi's. Since I started posting to this NG I've definitely been enlightened. I
doubt I'll ever be a slave to the calculator, but I'm learning.
Skeptically,
Brett
ron.j...@ccrs.nrcan.gc.ca
bem...@aol.com (Bemulli) wrote:
>
> ron.j...@ccrs.nrcan.gc.ca w rote:
>
> >In article <36E80EF7...@discover.earthlink.net>,
> > sean...@discover.earthlink.net wrote:
> >
>
> A couple of questions. First, is the formula to come up with MLE's
> widely available?
Yes. A basic version was published in the 1985 Bill James Baseball Abstract.
I prefer to calculate park adjustments slightly differently than the way
James advocates, but the method in there still works pretty well.
Conceptually it's pretty easy. Adjust for level of competition (AAA is
about 82% as strong as the majors, AA only slightly weaker. MLEs don't work
below AA, but frankly that's no biggy. If a guy is skipping drom A ball
to the majors, he's either a monster prospect or a rule V suspect. In the
former case, don't worry - he'll almost always be a star. In the latter,
don't worry. More than likely he'll be sent back to the minors and you
can get some feel for his ability.).
Then adjust for the offensive context (league and park), then adjust
for the major league park.
Actually this year any MLEs have to be done the old way. Stats didn't
publish the minor league park effects this year.
> In other words, can any stathead come up with1999 MLE's or is he
> dependent on the numerous Rotisserie League guides that plague us
> every year?
You can do them. Personally, I buy the Stats products anyhow and they
have the MLEs in them.
> Secondly, you state above that MLE is not a prediction but a
> translation ofwhat a player did previously. If so how in the world can
> this system spew out that Al Martin was due for a career year in 1998?
You're confusing Vlad ( Baseball Prospectus' old projection system )
with MLEs. What Martin did in the minors has very little bearing on
how well he'd been expected to play in 1998. (Though Martin's minor
league career is an interesting one. He was a complete stiff in the
Atlanta organization but played really well in 1992. Roughly 9 times
out of 10, that's a stone fluke. Like Billy Ashley, Doug Frobel and
Dave Hostetler for instance. Every now and then though a guy figures
it out fairly late in his career.)
The thing to understand about Vlad is that it takes the most optimistic
view plausable for a player. If you only use one projections system,
I'd have to suggest using Stats (which is based on Brock2 - speaking of
which. For those who want to do their own projections, you can get
Brock2 from Sean Lahman's site http://www.baseball1.com). Basicly 3
year averages with a regression to the mean and some simple age adjustments.
> Iirc, none of the resident statheads around here were predicting anything
> of the sort for Martin last Spring.
Well I'd have expected something on the order of .290/.355/.455
> FWIW, my stance to the whole stathead phenomena is very similar to
> Paul Rodi's. Since I started posting to this NG I've definitely been
> enlightened. I doubt I'll ever be a slave to the calculator, but I'm
> learning.
It may not seem this way to others, but statheads do not speak with
a single voice. Especially when it comes to player evaluation.
(Dejanews doesn't have any of the Craig Wright/Chris Karhl flame
fests. But Clay Davenport and Don Malcolm can't exchange two civil
words and they're two of the more influentant statheads. I have
tremedous respect for both.)
However there is total agreement that the Pirates would be better off
setting fire to (say) a 1/2 million dollars and playing Freddy Garcia
than signing Ed Sprague. That's regardless of our opinion of Garcia.
At least he has an up side.
Bob Rich's first rule of marginal players - Better a young bum than
an old bum. They work cheap and every now and then they improve.
Grativo
>that Al Martin was due for a career year in 1998?
BP's Vlad projections are different from MLE. MLE doesn't try to predict the
future; it just says how a minor leaguer would have performed in the majors.
Vlad uses substantially same info as MLE to predict how the player will perform
in the upcoming season.
>Iirc, none of the resident
>statheads around here were predicting anything of the sort for Martin last
>Spring.
The projections weren't that Al Martin would have a year substantially greater
than his 97 or 96. A few more singles than his peak years, that's
all...instead Al Martin looks like one of those players whose performance fell
off a cliff relatively early. Another thing about Vlads is that they tend to
miss on the high side. There are some players that Vlad is too optimistic on,
but there are far fewer players that the projection says will suck and they
end up putting up good years.
Grativo
ron.j...@ccrs.nrcan.gc.ca
In article <19990313015253...@ng137.aol.com>,
bem...@aol.com (Bemulli) wrote:
> A couple of questions. First, is the formula to come up with MLE's
> widely available?
I though I'd demonstrate the method the Bill James published back in
1985. There are some obvious refinements, but the basic method works
pretty well and the data required isn't always available.
Understand that the absolute most useful thing that MLEs do is
identify the guys who can't play. For this you don't need super
precision.
Also understand what is being claimed by people using MLEs. The
standard deviation between two player seasons in which he plays
full time in the majors in around 15 runs created. MLEs are equally
useful. They establish a general range of ability. Nothing more,
but nothing less.
Anyhow, let's look at Alex Rodriguez's 1995 (It's worth noting that
Stats took a look at this MLE, said - Nah, don't buy it, and fudged
it for their projections in the 1996 Handbook. Have some faith in
your own methods!)
The first thing to do is establish the offensive context he played in.
The sinple way is to take the runs scored and allowed by the team and
divide by the games played. Since this information is readily available,
that's what I'll use. It's actually slightly better to take the league
context and adjust for the park. But if you have the park factors, you
probably have the Minor League Handbook - and this has the MLEs in them.
(I'll be doing some rounding, but normally you carry the fractions.)
Tacoma: Games Played - 144, Runs Scored - 682, Allowed 734
Runs per Game 9.8
AL 1995 10.1
Interesting. The offensive level in Tacoma was lower than the AL.
When calculating runs per game in the NL, multiply by the result
by 5% to adjust for the fact that a player is moving from a DH
legue to a non DH league.
Next, divide the league average by the park average (ie 10.1/9.8)
and multiply this by .82
The result ( .84 in this case) is called the m factor. It's
how much the result stats are going to be adjusted (in broad terms),
That is to say that ARod's stats would be about 84% of what they were
in the minors.
You also need to take the square root of the m factor. .92 in this
case. Called the M factor.
James calculates runs and RBI. I don't bother since both are lineup
dependent.
Anyhow, Here's ARod's raw stat line:
G AB H 2B 3B HR BB K
54 214 77 12 3 15 18 44
H = ML H * .98 * M = (77 * .98 * .92) = 69
2B = ML 2B * M = (12 * .92) = 11
3B = ML 3B * .85 * m = ( 3 * .85 * .84) = 2
HR = ML HR * m = (15 * .84) = 13
BB = ML BB * m = (18 * .84) = 15
K = ML K * 1.05 = (44 * 1.05) = 46
After which you probably want to park adjust these numbers.
He was moving into a park that could be expected to deflate his BA
by about 1% and his HR by about 2%, absolutely kill his 3B, raise
his 2B by around 10%, his Ks by around 7% and his BBs by around 6%.
But in general, you wont have this info and in truth you don't need it.
Park effects change the shape of the stats but not the value.
Anyhow, the resulting stat line is: (Note that the AB is adjusted
downward by the difference between the hits at the minor league
level and the MLE.)
G AB H 2B 3B HR BB BA OBP SLG
54 206 69 11 2 13 15 .336 .382 .594
Stats cut this to .271/.314/.434 in their projections. As Bill James
said, it wasn't so much that they doubted that he was that good. They
simply expected him to struggle for a while at the start of the year.)
Arod was actually better than his MLE. A none too shabby .358/.414/.631
Bemulli
ron.j...@ccrs.nrcan.gc.ca wrote:
>> Secondly, you state above that MLE is not a prediction but a
>> translation ofwhat a player did previously. If so how in the world can
>> this system spew out that Al Martin was due for a career year in 1998?
>
>You're confusing Vlad ( Baseball Prospectus' old projection system )
>with MLEs. What Martin did in the minors has very little bearing on
>how well he'd been expected to play in 1998. (Though Martin's minor
>league career is an interesting one. He was a complete stiff in the
>Atlanta organization but played really well in 1992. Roughly 9 times
>out of 10, that's a stone fluke. Like Billy Ashley, Doug Frobel and
>Dave Hostetler for instance. Every now and then though a guy figures
>it out fairly late in his career.)
>
>The thing to understand about Vlad is that it takes the most optimistic
>view plausable for a player. If you only use one projections system,
>I'd have to suggest using Stats (which is based on Brock2 - speaking of
>which. For those who want to do their own projections, you can get
>Brock2 from Sean Lahman's site http://www.baseball1.com). Basicly 3
>year averages with a regression to the mean and some simple age adjustments.
>
Thanks. I knew MLE (Major League Equivalent) wasn't logical for somebody
already in the Majors, but I didn't know the name of the other system.
>It may not seem this way to others, but statheads do not speak with
>a single voice. Especially when it comes to player evaluation.
>(Dejanews doesn't have any of the Craig Wright/Chris Karhl flame
>fests. But Clay Davenport and Don Malcolm can't exchange two civil
>words and they're two of the more influentant statheads. I have
>tremedous respect for both.)
>
You're right. It does seem like the statheads band together. At least around
here. It would appear that there's a stathead NG? Maybe if I work up enough
courage, I'll drop by to lurk sometime. Honestly, the stuff does intrigue me,
but the formulas really frighten me. I ask this not to be snide, but because
I'm really curious. Would you say that you and statheads in general have a
strong Mathematics background?
>However there is total agreement that the Pirates would be better off
>setting fire to (say) a 1/2 million dollars and playing Freddy Garcia
>than signing Ed Sprague. That's regardless of our opinion of Garcia.
>At least he has an up side.
>
>Bob Rich's first rule of marginal players - Better a young bum than
>an old bum. They work cheap and every now and then they improve.
>
>--
>RNJ
>
I know it's a dirty word to all statheads, but the only explanation I would
have would be: intangibles. Oddly enough, the Bucs best streak of the year (9
in a row, I think) coincided with Garcia's tear. I think it's obvious that he's
got the most raw power on the team. It really would have been interesting to
see what he would've done if they wouldn't have given up on him so early last
year.
After the pitiful September (in which Ramirez got the majority of the starts)
obviously Bonifay and Lamont felt some "veteran leadership" was needed. Enter
Ed Sprague.
Brett
Ronald Johnson
Bemulli <bem...@aol.com> wrote:
>
> ron.j...@ccrs.nrcan.gc.ca wrote:
Sorry for the untimely response. Never saw this.
>
>>It may not seem this way to others, but statheads do not speak with
>>a single voice. Especially when it comes to player evaluation.
>>(Dejanews doesn't have any of the Craig Wright/Chris Karhl flame
>>fests. But Clay Davenport and Don Malcolm can't exchange two civil
>>words and they're two of the more influentant statheads. I have
>>tremedous respect for both.)
>>
>
> You're right. It does seem like the statheads band together. At
>least around here.
Trust me, it only seems that way. Ask Paul G. about some of the old
hands at RSB. I doubt he has fond memories.
> It would appear that there's a stathead NG? Maybe if I work up enough
>courage, I'll drop by to lurk sometime.
rec.sport.baseball
C'mon in. Though rlm is back. Be prepared for a certain amount of
extreme nastiness.
Not really a stathead group. Though most statheads in the team NG
hang out there too.
> Honestly, the stuff does intrigue me,
>but the formulas really frighten me. I ask this not to be snide, but
>because I'm really curious. Would you say that you and statheads in
>general have a strong Mathematics background?
Nope. I have zero formal training in math or stats and frankly it
shows sometimes. I've taught myself the basics of stats and I
work in a place loaded with physicists.
What makes RSB work is that it's very much a "show your work" NG.
And we've got a few math pros and statisticians there for when
the math gets really hairy.
In general though the math is presented in a way that if you've got
a spreadsheet you're in business.
>
> >However there is total agreement that the Pirates would be better off
>>setting fire to (say) a 1/2 million dollars and playing Freddy Garcia
>>than signing Ed Sprague. That's regardless of our opinion of Garcia.
>>At least he has an up side.
>>
>>Bob Rich's first rule of marginal players - Better a young bum than
>>an old bum. They work cheap and every now and then they improve.
>
> I know it's a dirty word to all statheads, but the only explanation
>I would have would be: intangibles.
Here's the way I look at that. Whatever you choose to believe about
intangibles, there's ZERO evidence for believing that Sprague has
positive intangibles.
Think about it. For it to matter the intangibles have to manifest
themselves in a tangible way.
Sprague's teams have been excellent, they've been terrible and they've
been in between.
Further, the Jays played much better after he was traded. When working
Craig Grebeck into the lineup improves the team, there's pretty fair
evidence that he's not much of a player.
Do Sprague's teams win more games than you'd expect given the
runs scored and allowed? No.
Do Sprague's teams score more runs than you'd expect given the
raw offensive stats? No.
Is there a pattern of players on Sprague's teams playing unexpectedly
well (and if so, do you have any reason to attribute this to Sprague.)?
No.
If you can't answer any of these questions with a positive, frankly
you got no business (and this is a general statement - not directed
at you personally) making decisions based on what amounts to figments
of somebody's imagination.
Do I believe in intangibles? Well it sure can't hurt to have your
best player with a work ethic like (say) Stan Musial. Curt Flood
commented that if a player as good as Musial felt it neccessary
to work as hard as he did, then he (Flood) had better work too.
And if you look at the Cardinals of Musial's day you'll see a lot
of young players who showed a lot of improvement.
Musial? Dunno. But at least there's something tangible to point to.
> After the pitiful September (in which Ramirez got the majority of
>the starts) obviously Bonifay and Lamont felt some "veteran leadership"
>was needed. Enter Ed Sprague.
I don't dispute the "logic". Merely question the sanity of choosing
this clown as a leader.
--
RNJ