PDA

View Full Version : BPro Pecota Projections


Dadawg_77
01-25-2005, 10:04 PM
Probally the best projection system out there that I have seen.
VORP is vaule over a replacement level player, who is some guy grabed from AAA to fill in for a player.

Avg/OBP/SLG/VORP
Joe Crede .266/.321/.459/11.0
Ben Davis .239/.299/.404/3.7
Jermaine Dye .256/.332/.449/11.2
Carl Everett .273/.338/.444/10.4
Ross Gload .270/.318/435/6.0
Willie Harris .262/.333/.368/9.1
Paul Konerko .276/.353/.494/25.9
Timo Perez .266/.307/.385/0.2
AJ Pierzyski .277/.327/.430/16.3
Scott Podsednik .278/.342/.416/17.7
Aaron Rowand .290/.341/.482/23.9
Frank Thomas .274/.399/.529/30.3
Juan Uribe .269/.316/.444/18.5
Wilson Valdez .254/.287/.330/-1.8

IP/ERA/VORP
Jon Adkins 57.0/4.79/9.5
Mark Buehrle 198.3/4.47/35.1
Jose Contreras 129.3/4.91/17.1
Neal Cotts 73.7/4.93/9.5
Freddy Garcia 178.7/4.55/30.5
Jon Garland 169.3/5.05/19.7
Dustin Hermanson 109.7/5.09/12.7
Orlando Hernandez 113.0/4.35/24.2
Damaso Marte 65.7/3.99/16.6
Cliff Politte 46.7/4.79/8.2
Scott Schoeneweis 103.0/5.24/10/0
Shingo Takatsu 54.7/4.54/11.1
Luis Vizcaino 60.7/4.91/10.0

MRKARNO
01-25-2005, 11:23 PM
The following players were on the White Sox roster the whole season (or got injured, because PECOTA takes that into account to) last year with their predicted VORP vs Actual VORP:

Sandy Alomar -4.3/-3
Joe Crede 26.5/1.2
Willie Harris 5.2/5.2
Ross Gload 7.6/17
Paul Konerko 15.7/48.1
Carlos Lee 19/46.8
Magglio Ordonez 40.8/11.8
Aaron Rowand 4.5/50
Frank Thomas 29.3/34.2
Jose Valentin 23/14.9
Jon Adkins 2.5/8.3
Mark Buehrle 28.6/50.8
Neal Cotts 3.9/.6
Jon Garland 15.9/25.2
Damaso Marte 22.9/23.4

Total of selected players: 241.1/333.3

Now it might just have to do with the selection of players I picked, but I find PECOTA to be a pessimistic system, but maybe A. my selection was a misleading one or B. It's just an average and we were relatively healthy last year.

It also might be that the system assigns value to some players who wont realistically get a lot of playing time to be in the 50th percentile in case one of those backups gets to play the whole year.

Interestingly enough, Thomas beat his PECOTA forcast in less than half of a season.

I still think it's a fairly pessimistic system on the whole and that while it's interesting to see where they think players are going (up or down) and their essays.

stillz
01-25-2005, 11:41 PM
Fascinating stuff. I think the system is somewhat negative, but it does make me believe that if we can just get a healthy year from all our starting hitters, this lineup is bound to click and produce runs. Getting Frank back along with a productive Everett and Rowand would give the Sox a real edge. I wonder what Tadahito's VORP would be?

The starting rotation looks truly enviable if healthy. :smile:

CWSGuy406
01-25-2005, 11:54 PM
Forgive my ignorance, but why is the IP for our starters so low?

I can understand El Duque's, he's obviously injury prone, and 113 seems like a reasonable number for him... But, 198 for Buehrle? 178 for Garcia? And, IMHO, the most outrageous one of all -- 169 for Garland?

Dadawg_77
01-26-2005, 10:45 AM
The following players were on the White Sox roster the whole season (or got injured, because PECOTA takes that into account to) last year with their predicted VORP vs Actual VORP:

Sandy Alomar -4.3/-3
Joe Crede 26.5/1.2
Willie Harris 5.2/5.2
Ross Gload 7.6/17
Paul Konerko 15.7/48.1
Carlos Lee 19/46.8
Magglio Ordonez 40.8/11.8
Aaron Rowand 4.5/50
Frank Thomas 29.3/34.2
Jose Valentin 23/14.9
Jon Adkins 2.5/8.3
Mark Buehrle 28.6/50.8
Neal Cotts 3.9/.6
Jon Garland 15.9/25.2
Damaso Marte 22.9/23.4

Total of selected players: 241.1/333.3

Now it might just have to do with the selection of players I picked, but I find PECOTA to be a pessimistic system, but maybe A. my selection was a misleading one or B. It's just an average and we were relatively healthy last year.

It also might be that the system assigns value to some players who wont realistically get a lot of playing time to be in the 50th percentile in case one of those backups gets to play the whole year.

Interestingly enough, Thomas beat his PECOTA forcast in less than half of a season.

I still think it's a fairly pessimistic system on the whole and that while it's interesting to see where they think players are going (up or down) and their essays.

Remember that PECOTA gives you a range and what you and I reported is the weighted mean of the range. One of the biggest examples was Konerko, who was coming off of a wretched year. I think 2003 might be noise which is confusing PECOTA as I think the projected VORP of 25.4 is too low. As for Frank, PECOTA similar hitter are way off, which is probally caused by his 2001 and 2002, which I think are also noise.

Randar68
01-26-2005, 12:11 PM
IP/ERA/VORP
Mark Buehrle 198.3/4.47/35.1
Freddy Garcia 178.7/4.55/30.5
Jon Garland 169.3/5.05/19.7
Orlando Hernandez 113.0/4.35/24.2
Shingo Takatsu 54.7/4.54/11.1


Given their recent years production, how can anyone justify the system spitting out the numbers you see above?

I'll take the under on all those ERA's and run all the way to the bank.

Not one of MB, FG, or JG pitching 200 innings? I'll bet anyone that 2 of the three will pitch over 220 innings and that 3/5 starters will have over 200...

nlentz88
01-26-2005, 12:31 PM
Not one of MB, FG, or JG pitching 200 innings? I'll bet anyone that 2 of the three will pitch over 220 innings and that 3/5 starters will have over 200...

I think I read somewhere that Ozzie wants to try and keep all his starters under 200 innings in 2005 so as to be rested for playoff run. He thinks that the bullpen is strong enough to take some of the pressure off the starters who in the past have been called upon to work quite late into games.

jabrch
01-26-2005, 12:45 PM
Given their recent years production, how can anyone justify the system spitting out the numbers you see above?

I'll take the under on all those ERA's and run all the way to the bank.

Not one of MB, FG, or JG pitching 200 innings? I'll bet anyone that 2 of the three will pitch over 220 innings and that 3/5 starters will have over 200...

No kidding... Those numbers look like more black magic to me.

SoxxoS
01-26-2005, 12:47 PM
I think I read somewhere that Ozzie wants to try and keep all his starters under 200 innings in 2005 so as to be rested for playoff run. He thinks that the bullpen is strong enough to take some of the pressure off the starters who in the past have been called upon to work quite late into games.

We haven't been to the playoffs since 2000. Don't jump before they put the bungee cord on. We need to get there first. I am all for saving arms and using the bullpen, but make that decision as the game unfolds, no with that in the back of your mind every game (which I don't think will happen).

But I agree with Randar, 2 of 3 will get 200 innings.

kempsted
01-26-2005, 01:04 PM
No kidding... Those numbers look like more black magic to me.
Just to let you know if you don't PECOTA is completely a computer program. The BP guys are constently disagreeing with it but it does a better job then most. It works by looking for similar players in the history of baseball and figures out how those players did the next season. It gives a range of possible value and is worth looking at. It is not meant to be taken as gospell, just another data point so you don't base everything on - well Maggs is awsome did you see the home run he hit on July ...

Randar68
01-26-2005, 01:27 PM
I think I read somewhere that Ozzie wants to try and keep all his starters under 200 innings in 2005 so as to be rested for playoff run. He thinks that the bullpen is strong enough to take some of the pressure off the starters who in the past have been called upon to work quite late into games.

What's he going to do, pull Buehrle and Garland after 80 pitches? Hernandez and Garcia I can see, but if they are both healthy, there's no way they pitch less than 200 innings. In addition, this is a statistical model, meaning anything Ozzie does or doesn't say means nothing.

Ol' No. 2
01-26-2005, 02:14 PM
Just to let you know if you don't PECOTA is completely a computer program. The BP guys are constently disagreeing with it but it does a better job then most. It works by looking for similar players in the history of baseball and figures out how those players did the next season. It gives a range of possible value and is worth looking at. It is not meant to be taken as gospell, just another data point so you don't base everything on - well Maggs is awsome did you see the home run he hit on July ...Unfortunately, when you do a reality check, you find that PECOTA and FLIPS or whatever program you use predicts just about as well as using a players 3-year average. Much ado about nothing.

maurice
01-26-2005, 02:16 PM
I agree that some of the numbers look fairly pessimistic, particularly . . .
AJ Pierzyski .277 AVE
Frank Thomas .399 OBP
Mark Buehrle 4.47 ERA
Damaso Marte 3.99 ERA

However, I'd be pretty happy to be ensured no less than . . .
Joe Crede .266 AVE
Scott Podsednik .278 AVE (especially if it comes with 60+ SBs)
ERAs near or under 5.00 from 7 different bullpen guys

SOXintheBURGH
01-26-2005, 02:33 PM
Unfortunately, when you do a reality check, you find that PECOTA and FLIPS or whatever program you use predicts just about as well as using a players 3-year average. Much ado about nothing.

One of the big reasons I never buy into these projections and expertise. I go with gut feelings based on what I'VE seen, and I get similiar results.

ma-gaga
01-26-2005, 06:07 PM
One of the big reasons I never buy into these projections and expertise. I go with gut feelings based on what I'VE seen, and I get similiar results.

Well, you can do that with one team, maybe a division, but they do it for all teams and all potential players including a ton of prospects that won't see one at bat this year.

The 'Prospect' projections are too optimistic, and the current players seem too pessimistic (especially looking at this on January 26th, where every team has a chance). But they work hard at this and these numbers aren't just pulled out of thin air. They look at just about everything...

(now if they'd just update their "fantasy projections" so I can get my RBI's...)

Dadawg_77
01-26-2005, 08:56 PM
(now if they'd just update their "fantasy projections" so I can get my RBI's...)

Ask them for holds.

SABRSox
01-26-2005, 11:17 PM
The problem with this system, and any other system for that matter, is that it's near impossible to project accurate playing time, IP or AB's. I know that when Bill James does his projections, he tends to give prospects too many AB's, just to show readers what they might do given that amount of playing time. He projects Calvin Pickering to hit 34 home runs next season. He's stuck behind Mike Sweeney and Ken Harvey in the lineup. I doubt he'll get anywhere close to enough AB's to hit 34 home runs.

I assume Baseball Prospectus is the same way with the PECOTA system. I think it's more for conversation and entertainment purposes than being an end-all projection system. I wonder what the accuracy rate on PECOTA is?

jeremyb1
01-27-2005, 06:38 PM
Given their recent years production, how can anyone justify the system spitting out the numbers you see above?

I'll take the under on all those ERA's and run all the way to the bank.

Not one of MB, FG, or JG pitching 200 innings? I'll bet anyone that 2 of the three will pitch over 220 innings and that 3/5 starters will have over 200...

Randar, the reason the innings pitched are low is because PECOTA generates its projections based on comperable players. So if a player highly similar to Buehrle over the last three seasons at the same age was injured after pitching 20 innings that figure is included and averaged together with the pitchers that pitched 200 innings and therefore drags the figure down. You might think it unreasonable to decrease the innings pitched based on what is not a large chance Buehrle will go down. But the weighted mean is the averave of the probability of various outcomes one of which is that Buehrle will miss time with an injury.

Personally I find Buehrle's ERA projection to be somewhat confounding. In fact I've been thinking about e-mailing Nate Silver (the PECOTA creator) for his take on the predictions. A 26 year old pitcher with the best full season K/9 and BB/9 totals of his career with only a minimal slip in HR/9 should lead to a strong season and Buehrle does have a high breakout and improvement rate in the projections. My best guess is that due to his workload the last few seasons, PECOTA is picking up players whose workload caught up to them. The other ERAs don't seem particularly high to me especially when considering the fact that they're park adjusted for the Cell which allowed a ton of home runs last season. Garland's ERA was close to 5 last season and the system is no doubt pessimistic about his declining K/9 the past few seasons. When you think about it a 25 year old who is consistently taking a step backwards as far as fooling hitters is a little scary. El Duque's age is no doubt going to catch up to him at some point. Garcia posted an ERAs of 4.4 and 4.5 the last two seasons and then posted an ERA of 4.4 with the Sox last season so his projection appears dead on.

Not necessarily my own beliefs, or as accurate predictions as one could come up with when adding analysis to the figures, but hopefully some insight into where the numbers are coming from.

santo=dorf
01-27-2005, 06:43 PM
Jeremy, wouldn't it just be easier to just ignore these stastical projections instead of trying to get in touch with the statheads behind it? :?:

Mohoney
01-27-2005, 07:02 PM
Jermaine Dye .256/.332/.449/11.2

I think Dye's going to be WAY better than this.

SoxxoS
01-27-2005, 07:34 PM
I think Dye's going to be WAY better than this.

There is no "thinking" here. These are stats. You can't argue with stats and whatever they say shall be taken like they were wrote from Jesus himself.

Jeremy-haven't seen you post in a while...you don't like us anymore?

Dadawg_77
01-28-2005, 03:02 PM
Jeremy, wouldn't it just be easier to just ignore these stastical projections instead of trying to get in touch with the statheads behind it? :?:

Ignoring data just because it disagrees with you assumptions doesn't lead to learning but to idioticy.

Ol' No. 2
01-28-2005, 03:05 PM
Ignoring data just because it disagrees with you assumptions doesn't lead to learning but to idioticy.No. Data is something that is real. If you computed his 3-year average, that's data. This is someone's PROJECTION of what he will do in 2005. Big difference.

Dadawg_77
01-28-2005, 03:14 PM
I think Dye's going to be WAY better than this.

Why? The VORP is low since PECOTA is only predicting 425 PA but the rate stats are in range of his healthy years (2002,2004). He may have a great year, PECOTA gives him a 31% chance of a breakout season this year.

Breakout Rate is the percent chance that a hitter's EQR/PA or a pitcher's PERA will improve by at least 20% relative to the weighted average of his EQR/PA in his three previous seasons of performance. High breakout rates are indicative of upside risk. Breakout rates measure change relative to a player's previously-established level of performance. For this reason, a high Breakout score can create a falsely optimistic picture for a player who has a very poor performance record. It is far easier for a player with a baseline of 40 EQR per season to improve upon that figure by 20% than it is for a player with a baseline of 100 EQR per season; as a result, his Breakout score is likely to be higher.

Dadawg_77
01-28-2005, 03:17 PM
No. Data is something that is real. If you computed his 3-year average, that's data. This is someone's PROJECTION of what he will do in 2005. Big difference.

Well it depends on which definition we are using.
http://www.webster.com/cgi-bin/dictionary?book=Dictionary&va=Data&x=0&y=0
I am using two and you are using one.

Ol' No. 2
01-28-2005, 03:27 PM
Well it depends on which definition we are using.
http://www.webster.com/cgi-bin/dictionary?book=Dictionary&va=Data&x=0&y=0
I am using two and you are using one.A PECOTA (or any other model) projection is a model prediction of what will happen in the future. While it may be based on data, it reflects the biases of the model's author and the limitations of the sample set used in its construction. A projection is not data. It is, at best, an extrapolation from previous data. It becomes data when, and if, it happens. And given the accuracy of these projections in the past, they're little more than fortune-telling.

Dadawg_77
01-28-2005, 03:36 PM
A PECOTA (or any other model) projection is a model prediction of what will happen in the future. While it may be based on data, it reflects the biases of the model's author and the limitations of the sample set used in its construction. A projection is not data. It is, at best, an extrapolation from previous data. It becomes data when, and if, it happens. And given the accuracy of these projections in the past, they're little more than fortune-telling.

I wasn't using Data in context of statical model but out knowledge out there. Everything is data. Like I said we are using two different definition of the word, neither one being more or less correct.

People can put different weight data or if you like information but to ignore something because you don't like what you said, doesn't help you.

maurice
01-28-2005, 03:52 PM
Ignoring a conclusion because you don't agree with the conclusion = illogical.
Ignoring a conclusion because the methodology is unreliable = logical.

Fortunately, the chicken entrails and the tea leaves both say that the Iguchi signing will cause the Sox to win the 2005 WS.
:tongue: