A's Grand Unified Projection: Batters

by Ryan Armbrust on Jan 16, 2007 2:57 AM PST

There are a few major projection systems out there that all claim to be the best. They all present plausible theories and evidence that they're more accurate than the others. PECOTA seems to have risen to the top, but it's still just one system.

I decided to stop trying to figure out which one to use, and just use four of them at once.

After importing all of the data into a spreadsheet, I unified it to produce one projection for a player. I figure it's better to have four experts reach a compromised consensus on how something will turn out, than to rely on just one or another. We'll see how it turns out at the end of October.

I tried to post the data in a table in here, but the "scoop" software AN runs doesn't like my table formatting at all... I swear, I'm not trying to lead you all away from AN, it just wouldn't format correctly.

So... here's a link to the table, in a nice pretty, readable setting.

http://thepastime.net/2007/01/16/the-grand-unified-projection-batters/

Now, I'm not claiming that I've created a system that's better than the sum of its parts. It may in fact be less than the sum of its parts. We'll have to see how it looks at the end of October.

If you'd like to see the data I'm working with, or are curious as to my process, let me know. I can share the spreadsheet on Google docs if there's some interest. I don't want to be too open about it, though, since I'm using BP's brand new PECOTA data, and they really don't like it being distributed gratis...

Anyway, what do you think? Even though I'm probably his biggest supporter, I'm surprised at how highly all four systems regard Dan Johnson. It also suggests not to get too excited about Mike Piazza and his 15 projected homers.

I'll post the pitcher projections either later today or tomorrow.

Update [2007-1-16 21:9:57 by Nebraska]: Weighted projections are now up.

0 recs | Comment 46 comments

Story-email Email Printer Print

Comments

Interesting stuff!

Keep up the great work, Nebraska-- this and your 'umpire graphs' were great!!

Me and your kid probably have the same IQ!! -Swisher to Chavvy

My Blog

by rungood on Jan 16, 2007 5:51 AM PST 0 recs

Great job! But the team can't hit...

Basically no one's great, Chavez and Swisher are decent, and everyone else is pretty bad.

by WaddellCanseco on Jan 16, 2007 7:42 AM PST 0 recs

Neat entry

thanks for the work, I guess the systems predict we will platoon our DHs of Piazza (294 ABs) and Durazo (341 ABs) or something to that effect. The combined short armed mullet monster looks like it will hit 25 HRs for us this year.

"Choosing between Milledge and Gomez is like choosing between Mozart and Beethoven" --NY Mets Message Boards

by apilgrim on Jan 16, 2007 8:38 AM PST 0 recs

PECOTA doesn't

and I'd guess the others work in a similar way.

They make no assumptions on roster use. They ABs they are assigned are based on recent history, adjusted based on various characteristics of the player.

Pizza had about 300 ABs because he has recently averaged about 300 ABs and he is old.

Read my latest articles

by devo on Jan 16, 2007 1:40 PM PST up 0 recs

Thanks for sharing that work...

Does the system allow Vs RHP and VS LHP splits for platooning information?

We know players like DJ, Durazo, Keilty, & Bocachica excel in platoons and I suspect Barton and Robnett are two more...

...Any evidence?

VORS = Value Over Replacement Suckitude

by Billy Ball 2005 on Jan 16, 2007 10:14 AM PST 0 recs

Wow

We suck more than I thought

by Colorado Fan on Jan 16, 2007 10:14 AM PST 0 recs

Ha

Not a single batter with a slug over .500 lol.
With an offense like that i smell sub .500 season too.
I am much less worried about the pitching than the hitting now.
Very interesting projections :D

BRING BACK MIGGY!

by sublimeguyjohn on Jan 16, 2007 10:25 AM PST 0 recs

Great stuff Nebraska. Very interesting....

But I am shocked anyone is surprised by this.....Was there any indication that our offense has suddenly gotten good in the off-season? They lost the guy who literally carried them the second half and have added one guy, who might be good, but wont be the lightning in a bottle Hurt was. The only way I see this offense being competitive is if Crosby suddenly finds the All Star level many expected out of him last year and Bradley manages to stay healthy and play 150 games, a REAL longshot.

I am very worried about this O and have been all off-season. I think everything needs to break right for the O as is currently sits to be succesful. That's scary, cuz rarely does everything break right

Bring back Hammer.

by OaktownPower on Jan 16, 2007 10:42 AM PST 0 recs

undo, undo

those numbers just made me get under my desk and curl into a ball.

by Eric in Atlanta on Jan 16, 2007 10:56 AM PST 0 recs

the parts

"I'm not claiming that I've created a system that's better than the sum of its parts. It may in fact be less than the sum of its parts."

perhaps the four could be weighed based on how accurate they have been in the past. i don't know if this information exists for previous years, but the 2006 results are here:
http://www.baseballthinkfactory.org/files/newsstand/discussion/2006_projection_results/

Hitters
PECOTA .736
BIS/James .685
ZiPS .684
Chone .677

Pitchers
ZiPS .459
PECOTA .451
BIS/James .445
Chone .424

here's something for 2003, but 4 of the 6 are different:
http://www.baseballprospectus.com/article.php?articleid=2515

A's v Giants "is kind of like the difference between going to see the Ramones and going to see the Bee Gees. A's fans will go see the Ramones." -BB 07/27/05

by xbhaskarx on Jan 16, 2007 11:26 AM PST 0 recs

I've been considering weighting them

but I was unsure of how much to give each system. Thanks fr the heads up on those numbers, I might plug them in a and see what happens.

--Nebraska--

ThePastime

by Ryan Armbrust on Jan 16, 2007 11:57 AM PST up 0 recs

two more links

http://www.insidethebook.com/ee/index.php/site/comments/forecasters_how_accurate_can_they_possibly_b e/

the diamond mind projection system was close to pecota:
http://www.diamond-mind.com/articles/tmpred06.htm

A's v Giants "is kind of like the difference between going to see the Ramones and going to see the Bee Gees. A's fans will go see the Ramones." -BB 07/27/05

by xbhaskarx on Jan 16, 2007 12:10 PM PST up 0 recs

i don't know why everyone is so upset

how different are these from the 2006 offensive projections?

A's v Giants "is kind of like the difference between going to see the Ramones and going to see the Bee Gees. A's fans will go see the Ramones." -BB 07/27/05

by xbhaskarx on Jan 16, 2007 11:57 AM PST 0 recs

Not very...

But the offense was far from special last year too.

Bring back Hammer.

by OaktownPower on Jan 16, 2007 12:03 PM PST up 0 recs

Last year's projections

I saved last year's projections (PECOTA, ZIPS, Bill James, and PrOPS) for the A's, and have a comparison of the predicted OPS along with the actual results. It's on my home computer, but I'll try to remember to post it tonight.

My take on this year's projections is that once again the A's offense is dependent on a lot of players with iffy health. Projection systems are all necessarily "dumb" - they have the information of how well players hit in the past, but no sense of why they did so. Crosby, Chavez, and Bradley are all very good bets to beat their projections if they stay healthy, but, of course, they are also poor bets to stay healthy. So the numbers listed here make sense as averages, but there is a lot of unpredictable variance there, even beyond the luck/sample size issues detailed in one of the articles Bhaskar linked.

The main difference from last year's team is that we don't have any hitters like Frank Thomas, who was a much more extreme version of the same thing - very high risk (health) but very high reward (production) when he was able to stay in the lineup. Even in the best case, I think it's unlikely that Piazza, Chavez, or anyone else will match Thomas' production from last year. We do have Rich Harden, though, who's sort of the pitching equivalent.

In the stands the home crowd scatters For the turnstiles

by andeux on Jan 16, 2007 12:52 PM PST up 0 recs

Could you post entries

for Erstad and Milledge?
Thanks

"Choosing between Milledge and Gomez is like choosing between Mozart and Beethoven" --NY Mets Message Boards

by apilgrim on Jan 16, 2007 12:03 PM PST 0 recs

sure, give me a bit

I'll get that together

--Nebraska--

ThePastime

by Ryan Armbrust on Jan 16, 2007 12:06 PM PST up 0 recs

and here they are

All of the systems are down on Erstad, especially PECOTA. Milledge looks like a lock, though he does strike out a lot.

http://thepastime.net/2007/01/16/the-grand-unified-projection-additions/

--Nebraska--

ThePastime

by Ryan Armbrust on Jan 16, 2007 12:30 PM PST up 0 recs

i would have thought

the CHONE one would have ersty at like 52 game winning home runs in 48 at bats.

A's v Giants "is kind of like the difference between going to see the Ramones and going to see the Bee Gees. A's fans will go see the Ramones." -BB 07/27/05

by xbhaskarx on Jan 16, 2007 12:40 PM PST up 0 recs

thanks for the post

those Milledge numbers look nice.

"Choosing between Milledge and Gomez is like choosing between Mozart and Beethoven" --NY Mets Message Boards

by apilgrim on Jan 16, 2007 6:52 PM PST up 0 recs

Projections

Using the ZIPS projections and my own playing time projections, I have the A's scoring 781 runs and allowing 717.

A Yankees Blog ran the CHONE projections through 100 Diamond Mind Baseball seasons and had the A's scoring 836 runs and allowing 722 (this assumes Crosby and Harden stay healthy): http://yankeefan.blogspot.com/2006/12/extremely-early-projections.html

by Danny on Jan 16, 2007 12:08 PM PST 0 recs

but chone was the least reliable of the four

at least last year, and that's according to the guy who did them.

A's v Giants "is kind of like the difference between going to see the Ramones and going to see the Bee Gees. A's fans will go see the Ramones." -BB 07/27/05

by xbhaskarx on Jan 16, 2007 12:11 PM PST up 0 recs

For sure

just throwing it out there...

by Danny on Jan 16, 2007 12:16 PM PST up 0 recs

Good stuff ...

Could I get your data set? dlavelle at alum d0t pomona d0t edu

Read my latest articles

by devo on Jan 16, 2007 1:43 PM PST 0 recs

done

I sent you an email.

--Nebraska--

ThePastime

by Ryan Armbrust on Jan 16, 2007 2:05 PM PST up 0 recs

Go Stags!

Bring back Hammer.

by OaktownPower on Jan 16, 2007 2:29 PM PST up 0 recs

What happened?

What happened to the A's ability to draft and develop qualtiy hitters? Giambi, Tejada, Chavez, (even Grieve!)? Swisher seems like the only legitimate hitter we've drafted for a while. Hopefully Buck and Barton will develop into quality big league hitters.

by Flash G on Jan 16, 2007 3:06 PM PST 0 recs

Last year's results

On the off chance that anyone else is interesting, here are last year's projections for OPS along with the actual numbers:
JAMES ZIPS PrOPS PECOTA Actual Chavez 849 854 810 833 786 Thomas 873 804* 839 926 Bradley 820* 854 799* 802 818 Johnson 828 826 840 815 704 Crosby 800 801 777 799 636 Swisher 776 782 785 802 864 Ellis 804 771 744 777 704 Kotsay 781 786 762 746 719 Payton 783 739 714 753 743 Kendall 745 713 715 671 709

Ginter 775 738 758 721 -- Kielty 769 751 754 756 770 Melhuse 756 664 716 648 Perez 773 672* 722 389 Scutaro 738 718 720 703 747

Everyone missed low on Thomas (coming off two injury-plagued seasons) and Swisher (coming off a mediocre rookie year with lots of strikeouts), and to a much lesser extent Scutaro (career year). Everyone missed high on Chavez (injuries), Johnson (AAAA), Crosby (injuries), Ellis (coming off a career year), Kotsay (injuries), Perez (just sucks). They were pretty accurate on Bradley, Kendall, Payton, and Kielty.

On the whole, PECOTA was a little better than the others, but not spectacularly so. And it kind of shows the limits of these systems, as they really can't say anything useful about who will suddenly get hurt, or unexpectedly stay healthy. All they can do is give a rough talent level of the hitters.

This year's team might be even harder to predict than last year's. The biggest question mark is obviously health, again. In addition there is Piazza (once a great hitter, but PECOTA hates him) and Durazo (hasn't played in the majors in a year and a half, so who knows), along with some prospects (Barton, Goleski) who probably aren't ready yet, but could surprise us.

In the stands the home crowd scatters For the turnstiles

by andeux on Jan 16, 2007 7:35 PM PST 0 recs

about individuals

And it kind of shows the limits of these systems, as they really can't say anything useful about who will suddenly get hurt, or unexpectedly stay healthy. All they can do is give a rough talent level of the hitters.

That's why I don't put much faith in the individual player's projection, but when you look at the whole team, it's more accurate. For every player that exceeds expectations, there's one that doesn't meet them. I think they're more useful for projecting team results than individual. That's not to say that they can't get close for one player, but don't go betting the farm on it.

--Nebraska--

ThePastime

by Ryan Armbrust on Jan 16, 2007 10:56 PM PST up 0 recs

And how did the pool game go?

Nevermind, that's your own business.

What I don't understand is how PECOTA and everybody came up with those projections for Lastings Milledge. Let me rephrase that, I know how they came up with those numbers I'm just shocked that no one had the common sense to realize how screwed up their data was.

This guy is dead! We'll list him as day-to-day for possible reincarnation.

A's Medical Staff, 2006

by grover on Jan 16, 2007 11:09 PM PST up 0 recs

actually

there was no pool game. we got too caught up in conversation at the bar, and before we knew it, it was last call. that's not a bad thing, though.

--Nebraska--

ThePastime

by Ryan Armbrust on Jan 16, 2007 11:58 PM PST up 0 recs

BP explained

Sort of. I forget who it was, but someone said that translated stats from Norfolk are actually better than actual stats because it's such an extreme pitchers' park that the adjustment for offensive context is more strongly positive than the adjustment for the league context is negative.

When you consider that he was 21, and he put up a major league equivalent .489 (I think it was) slugging, it's not surprising that he would have a good projection. As the scouts are generally in agreement with the stats guys on him, I'm inclined to agree. If we do get him, he'll probably be our best player within a couple of years and he has MVP potential down the road.

by yarky on Jan 17, 2007 4:08 AM PST up 0 recs

The bulk of his success came in April

Milledge hit 357/500/524 his first time through the IL.

He hit 247/341/408 for the rest of his AAA stint.

He hit 241/310/380 in 166 big league at bats.

Look at those last two sets of numbers, consider that they represent a shade under 400 consecutive AB, and tell me how you can project 281/358/443 next year in the Show.

This guy is dead! We'll list him as day-to-day for possible reincarnation.

A's Medical Staff, 2006

by grover on Jan 17, 2007 5:50 AM PST up 0 recs

This is how

PECOTA is based on this basic premise: take a player, put his age, basic physical characteristics, defensive position, stats, and places played into the program. PECOTA then uses that information and compares it to how other players that have played at that age, with those basic phycial characteristics, put up those stats, etc. Here is what I think makes Milledge special in PECOTA's eyes, but you should send an e-mail to Nate Silver and ask him if he agrees: players that have a history of quality minor league numbers that hold their own in their first taste of the major leagues at an early age have an incredible history of success in the major leagues. Because those players have had great success, and they are Milledge's most comparable players, PECOTA will project that Milledge will put up those numbers.

You raise legitimate concerns about Milledge's stats from last year, his drop off could indicate that pitchers have figured out weaknesses to exploit. It could also mean that Milledge had a minor injury after April, or that his momma stopped loving him, or he failed to emotionally handle his callup to New York and demotion back to Norfolk. These concerns would need to be addressed by scouting, because PECOTA (or any other statistically based system) is not going to have the answers to those questions.

Don't be surprised if that PECOTA projection ends up being pretty accurate, grover, and if it is close to true Milledge will be untouchable come July. Once a young potential star has success at the major league level, they almost never get traded. If we are going to trade for Milledge, this offseason will likely be the last chance to do so. Having never seen Milledge play, I have no particular scouting opinion of him, but the stats guys and scouts both think that Milledge has stardom in his future. They may be wrong, but that's a pretty broad spectrum of experts to disagree with.

I'd like to eat my lunch, but Billy just kicked me out of my office.

by BlameChannel53 on Jan 17, 2007 6:56 AM PST up 0 recs

pray for the health ...

of Beltran, Alou and Green ... if they stay healthy, Milledge has little chance to succeed at the big league level before 7/31.

Read my latest articles

by devo on Jan 17, 2007 10:06 AM PST up 0 recs

yes, but . . .

If he hits something like 320/410/500 in Norfolk, I would imagine the Mets would be mighty impressed and would be reluctant to trade him. My guess is that March 2007 will be the last chance anyone has to trade for Milledge. He is likely to explode once play starts and will become untouchable.

I'd like to eat my lunch, but Billy just kicked me out of my office.

by BlameChannel53 on Jan 17, 2007 10:12 AM PST up 0 recs

I'm not sure his momma ever loved him

She named him "Lastings" afterall.

I think the estimates need to be a touch more conservative in this case. I'm not saying Milledge will never reach those numbers (or beyond) but I think it a strecth to expect them in 2007.

Maybe an email to BP is in order...

This guy is dead! We'll list him as day-to-day for possible reincarnation.

A's Medical Staff, 2006

by grover on Jan 17, 2007 11:18 AM PST up 0 recs

I think you're missing something

I haven't seen research indicating that April stats are less predictive than September stats, but if it's out there, BP should definitely include that info in their system.

But anyway, this isn't a guy the stats like, but scouts have doubts about. He's raved about by scouts and purely statistics-based systems project him to be minor star at 22. I don't think a disappointing cup of coffee or a bad month at triple A are enough to counter all the weight on the positive side of the ledger here.

by yarky on Jan 17, 2007 2:51 PM PST up 0 recs

There's a red flag here

I'm not arguing about Milledge's long-term talent, I just think his 2007 projections are inflated.

Milledge's best month, by far, came in April. He accomplished a 357/500/524 line while debuting in a new, more advanced league than he had played before. Normally a player will struggle when debuting in a more advanced level as he has to adjust to better talent, Milledge's early success does nothing but validate all the scouting reports that praise his natural talent.

However, being new to the league is a double edged sword. The opposing pitchers didn't know how to attack Milledge, they didn't have a scouting report that pointed out his tendencies, strengths and weaknesses. Some pitchers paid the price early in the year to devlope those scouting reports, but they learned and adapted and Milledge's production dropped off dramatically. He lost over 100 points in every offensive category and saw his K's go up.

The league had adjusted to Milledge and he showed little if any ability to adjust himself. Some of that was undoubtedly due to bouncing between Norfolk and New York but the cold hard truth is the opposition wrote a book on how to attack Milledge at the plate and they had a lot of success with it from May - September of 2006.

Baseball is not just about talent. It's also about learning and adjusting and that is something Milledge has yet to learn, or at least display. In fairness, the Mets have rushed him through 4 levels in the last two years and he has never had the time to learn from his mistakes and correct them. But it should not be ignored that for all his talent, Lasting Milledge struggled for the bulk of 2006 and it is his ability to learn from those struggles that will ultimately determine how good he becomes.

This guy is dead! We'll list him as day-to-day for possible reincarnation.

A's Medical Staff, 2006

by grover on Jan 17, 2007 5:30 PM PST up 0 recs

Whenver you're dealing with an extreme ...

it is difficult to project, because, given that it is extreme, by definition, it has fewer comperables.

That said, it cuts both ways.

Read my latest articles

by devo on Jan 17, 2007 5:46 PM PST up 0 recs

That was a little vague

I can think of 3 ways that could be applicable to my comment. Therefore I shall choose the most pro-me view and say "Thank you".

This guy is dead! We'll list him as day-to-day for possible reincarnation.

A's Medical Staff, 2006

by grover on Jan 17, 2007 5:49 PM PST up 0 recs

You're welcome ...

what I meant is that his 2007 MLB projection is because of his incredibly pro-pitcher home park. He plays in a very extreme environment and because of that, his performance numbers are not at all comperable to other players at the same level in more neutral parks.

Read my latest articles

by devo on Jan 17, 2007 10:48 PM PST up 0 recs

OK, there are 4 ways

I didn't consider that youwere talking about park factors. And I assume you're referring to Milledge spending the bulk of his 2006 in Norfolk and thus he gets a bit of a mulligan interms of his under-production after April.

Well, it's not like Shea Stadium is a RH hitters Mecca! Why wouldn't park effect come into play on both sides of the equation?

This guy is dead! We'll list him as day-to-day for possible reincarnation.

A's Medical Staff, 2006

by grover on Jan 18, 2007 5:43 AM PST up 0 recs

It does come into play ...

but from what I understand, Norfolk is more extreme (in the opposite direction) than Coors was at its peak. Most "hitters and "pitchers" park inflate or deflate offense by something like 2-5% ... Coors on the other hand (pre humidor) inflated offense by ~25%.

Among players with at least 200 ABs, he did lead the Tide in OPS.

Read my latest articles

by devo on Jan 18, 2007 9:59 AM PST up 0 recs

OK

All these projections (I assume) are based on some kind of average of what similar players have done. So if you're saying that a more granular analysis leads you to conclude that he'll be in the low end of the group, fair enough. Your explanation for his breakdown is reasonable, but there are many other possible ones (including just luck). I don't think the evidence is very strong to support any of the possible explanations that I can think of.

by yarky on Jan 18, 2007 1:54 AM PST up 0 recs

Each player is unique

I'm not comparing Milledge to his "similiars". (Is it just me or does that sound slightly ominous, like something they would have used in an old episode of Star Trek or The Twilight Zone?) I don't know, maybe I've too many "He's the next Pedro!" and yet there's still only one. What I'm saying is that I think Milledge's 2007 MLB projection needs to lose about 20 points across the board.

This guy is dead! We'll list him as day-to-day for possible reincarnation.

A's Medical Staff, 2006

by grover on Jan 18, 2007 5:48 AM PST up 0 recs