PECOTA takes on standings

[byrdz]

If I were to quibble, I'd probably take 2 wins away from New York and give them to Toronto.

Then you would have to give them back plus some after they sign Clemens.

[Tony-OH]

Maybe, or you can take his season-by-season OPSes and you'll see an age-27 peak in '05, and a .729 fits in nicely for 2007.

You and I say "he was good two years ago, got hurt, then started to get back to his peak last year." PECOTA says throughout history guys like Roberts often get hurt and rarely rebound.

Or it could be a fluke.

That's pretty interesting. I didn't realize that PECOTA believes that a Roberts type of player is on the downside. What other players like Roberts are they referring to? Who's his PECOTA twin brother from a past?

[davearm]

99% is admittingly high because people will predict with their hearts, not minds.

But if most people project them to finish 4 or 5, chances are they will be within 1 spot of where they finish.

And you know what, those people don't need a project system to do that. Its not hard to predict standings and come within one spot.

It is easy to identify the best and worst teams in each division. After that, you just need to have a few things go right and boom, you are a walking, talking PECOTA system.

Well as baltimoron has explained, you're pretty strongly mischaracterizing what PECOTA is, by implying that it is a standings prediction tool. That's one application of it, but it's pretty far removed from the core app.

At its core, PECOTA projects individual player production.

Using some playing time assumptions, individual player stats can be extrapolated to team RS and RA numbers.

Team RS and RA numbers can be plugged into the pythagorean formula to get expected wins and losses.

Then expected wins and losses can be compared across the league to generate projected standings.

The bottom line is that PECOTA doesn't exist to project standings, even though it can be used for that purpose.

[Sapper]

BP article from yesterday. Nate Silver piece on projected standings, and where he thinks the system underrated/overrated/nailed individual players.

Exerpts from O's relevant bits:

                       W         L         RS     RA     AVG     OBP    SLGBoston Red Sox         93        69        892    762    .280   .361   .464New York Yankees       93        69        897    771    .280   .362   .454Toronto Blue Jays      80        82        813    831    .271   .342   .449Tampa Bay Devil Rays   78        84        806    834    .275   .338   .446Baltimore Orioles      74        88        754    823    .274   .333   .423

And:

I know those 74 wins won't be popular around here, but being completely objective (and not assuming any huge breakouts) I'm not sure how you can project them to any better than about .500.

My own guess is:

90th percentile - 87 wins

75th - 83 wins

50th - 78 wins

25th - 74 wins

10th - 70 wins

I think they're spot on. I expect about that many wins (maybe a few more). I also expect the pitching to improve and the offense to prove extremely weak again.

[davearm]

Interesting that PECOTA predicts evey one of our starters to decline except Markakis who just treads water. If we just win 74 this year the only man left standuing will be Angelos IMO.

That's because every one of the O's starting position players except Markakis and CPat is on the wrong side of 30.

It only makes sense to predict a decline for the thirtysomething guys.

[DrungoHazewood]

That's pretty interesting. I didn't realize that PECOTA believes that a Roberts type of player is on the downside. What other players like Roberts are they referring to? Who's his PECOTA twin brother from a past?

I was kind of generalizing about how they do projections. His top comps are Ray Durham ('01), Don Buford ('66), Bill Doran ('86), Roberto Alomar ('97), and Jim Gilliam ('58). A lot of his comps had hiccups in mid-career (Buford, Doran and Alomar had some of their worst seasons directly following the year that compared to Roberts' '06) then bounced back, so like I told Frobby, I think there's a little flukey quality to Roberts' projection.

[Frobby]

That's because every one of the O's starting position players except Markakis and CPat is on the wrong side of 30.

It only makes sense to predict a decline for the thirtysomething guys.

Roberts is 29. Huff and Gibbons are 30. I think the declines forecast for those players seem illogical. The fact that they have Markakis stagnant also seems illogical.

[Baltimoron]

Roberts is 29. Huff and Gibbons are 30. I think the declines forecast for those players seem illogical. The fact that they have Markakis stagnant also seems illogical.

Subtle nitty point but PECOTA is an "it". BP is a group of writers, one of whom, Nate Silver, came up with a projection program that was first published in BP 2003 (this is its 5th year of widespread publication.)

BP is a lot more than Nate's PECOTA program, and PECOTA doesn't speak for BP. For example, Kevin Goldstein, the guy who did the top 100 prospects and is the big draft/minor league writer, eschews a lot of PECOTA stuff, especially as it pertains to the players he covers.

Nate himself thinks the PECOTA projection for Nick is wrong and pretty fairly recognizes its limitations, although the tenor of the bulk of what he writes is in defense/support of PECOTA, as it (and other projection systems) aren't to big outside of fantasy baseball (as a side note BP apparently has at least 80 subscriptions from major league front office personal, so they are making inroads).

Most BP writers refer to PECOTA, as its still very good, but also color those projections with their own observations and opinions, as should we all.

[TradeAngelos]

Actually PECOTA last year predicted every team except the Indians, Astros, Cubs and Reds within one spot of their eventual place in the final division standings.

In the AL West, not only did PECOTA pick the correct standings, they predicted the A's, and the Rangers win totals dead-on, and they were only one off on the Mariners win total for the season.

Given PECOTA can't predict in season trades or injuries, the last year predictions seem pretty darn good to me.

For every team in the major, PECOTA projected them within 12 wins of their actual win total. The average difference of projection to actual wins was about 5 wins.

That is stunning that it was an average of about 5 wins give or take they were off,and I challenge anyone and everyone on this board to even come close to that.Please post all your predictions and at the end of the season we will see what the average was,I would guess that most here couldnt come within 10 games on average.

[DrungoHazewood]

Roberts is 29. Huff and Gibbons are 30. I think the declines forecast for those players seem illogical. The fact that they have Markakis stagnant also seems illogical.

I can agree that Markakis' and Roberts' projections are probably too conservative. We've talked about reasons for each of those.

But Huff and Gibbons? Those look a lot more reasonable.

Huffs' top 10 comps: Dick Sisler, Pat Putnam, Sean Casey, Chris Chambliss, Paul O'Neill, Eddie Robinson, Larry Sheets, Leon Durham, Sid Bream, Hideki Matsui. Several players who had a few years left, like Chambliss and O'Neill. But 2/3 of the group fell off a cliff after their 30th birthday.

Gibbons' are much uglier: Troy O'Leary, Robert Fick, Jim Spencer, Wes Covington, Pat Putnam, Jerry Lynch, Johnny Blanchard, Gordy Coleman, Gus Bell, and Dick Sisler. A few of those players had 3-4 good partial platoon seasons left. But most of them were bad or just disappeared. Logically Jay Gibbons is a DH with no defensive value who can't stay healthy, has a career .319 OBP, and has slugged .500 once in his life. Logically he's barely preferable to freely available talent like Knott, House, and Dubois.

[Frobby]

I can agree that Markakis' and Roberts' projections are probably too conservative. We've talked about reasons for each of those.

But Huff and Gibbons? Those look a lot more reasonable.

Huffs' top 10 comps: Dick Sisler, Pat Putnam, Sean Casey, Chris Chambliss, Paul O'Neill, Eddie Robinson, Larry Sheets, Leon Durham, Sid Bream, Hideki Matsui. Several players who had a few years left, like Chambliss and O'Neill. But 2/3 of the group fell off a cliff after their 30th birthday.

Gibbons' are much uglier: Troy O'Leary, Robert Fick, Jim Spencer, Wes Covington, Pat Putnam, Jerry Lynch, Johnny Blanchard, Gordy Coleman, Gus Bell, and Dick Sisler. A few of those players had 3-4 good partial platoon seasons left. But most of them were bad or just disappeared. Logically Jay Gibbons is a DH with no defensive value who can't stay healthy, has a career .319 OBP, and has slugged .500 once in his life. Logically he's barely preferable to freely available talent like Knott, House, and Dubois.

Gibbons has had exactly one season where his OPS was below .780. I think it's highly unlikely he ends up below that this year.

[Sports Guy]

Gibbons has had exactly one season where his OPS was below .780. I think it's highly unlikely he ends up below that this year.

And that is with him missing playing time do to unjury and getting too many ab's.

If he can stay healthy and Perlozzo can understand how to use a bench, Gibbons could/would have a 800+ OPS.

[DrungoHazewood]

Gibbons has had exactly one season where his OPS was below .780. I think it's highly unlikely he ends up below that this year.

And he's had one year over an .800. Ever.

I hope you're right, but it often just doesn't work that way. Gibbons is a guy who's never healthy, is adjusting to DHing, and he's 30. They're trying to find ways to keep him out of the field, and eventually somebody will notice that an average DH has an OPS 40 or 50 points above his career average, and they'll start docking his playing time. They're already hinting around at him losing playing time to the other corner players backed up on the roster.

Power hitters with no defensive value and injury histories tend to fall off a cliff in their early 30s. I'd love to hear a logical reason why Gibbons is going to buck that trend.

[DrungoHazewood]

If he can stay healthy and Perlozzo can understand how to use a bench, Gibbons could/would have a 800+ OPS.

I might be able to see Gibbons as Jim Dwyer, circa 1983-87. But we all know Perlozzo is very unlikely to use a player in that way.

[TGO]

PECOTA projects them to increase that output by almost 120 runs? that is not just improvement, that is a downright shocking turnaround. who injects this life? Delmon Young? Possibly. The cat from Japan?(points if you tell me the song.) Im not sold on that.

Ziggy Stardust