Jump to content

Updated PECOTA: Orioles will finish 15 games out


Frobby

Recommended Posts

Someone at BP woke up yesterday and decided they hated us. The individual PECOTAs have been revised and almost every Oriole was downgraded offensively. To name a few:

Markakis .811 (was .877)

Jones .767 (was .851)

Reimold .790 (was .815)

Atkins .756 (was .811)

Wieters .790 (was .840)

They now have us scoring 767 runs, tied for 7th in the AL.

And their beta page still has Markakis at .877. I'm not going to worry too much about any of this until the full cards go live, hopefully soon. They're obviously still in the midst of some iterative processes.

Link to comment
Share on other sites

  • Replies 104
  • Created
  • Last Reply
We have a very long history of posters here who've characterized themselves as realists after saying worst-case scenarios were the most likely. I guess that's not too surprising after a decade of losing, but it leads to some projections that are frankly absurd.

I've had people tell me that, basically, they're expecting the 2010 Orioles to be bad because Tony Batista and Matt Riley sucked.

And just about all of those doom and gloom projections have been spot on for the past several years. So, unfortunately, those doom-and-gloomers have been the realists. We have been bad. Very bad. For way too long. And I can understand people who feel we haven't gotten better to a point where they are willing to say that we will contend or even flirt with .500 ball. Maybe we will be as good as people are predicting us to be. I certainly hope that we are even better. But, that doesn't quell the fact that in the back of my mind I still think there is a very real possibility of us not improving significantly from last year's dreadful performance.

Here's to a 100-win season from the Orioles!

Link to comment
Share on other sites

And their beta page still has Markakis at .877. I'm not going to worry too much about any of this until the full cards go live, hopefully soon. They're obviously still in the midst of some iterative processes.

I know PECOTA is a good measuring stick but really, how much can you trust a system that is changing so much week to week?

Its like their WARP3 system....Cal's WARP3 used to be close to 180 I believe and now its under 100. Of course, everyone has seen a huge drop but still, what kind of system has that many drastic changes that quickly?

Link to comment
Share on other sites

I know PECOTA is a good measuring stick but really, how much can you trust a system that is changing so much week to week?

I don't put any weight on the projections while they're still in flux. This year the process has been a little more turbulent than in the past. Hopefully they'll get the kinks worked out. If not, other systems will pass them by.

Its like their WARP3 system....Cal's WARP3 used to be close to 180 I believe and now its under 100. Of course, everyone has seen a huge drop but still, what kind of system has that many drastic changes that quickly?

WARP had a well-documented problem with setting the replacement level way, way, way too low. Like at generic AA player levels. So even guys like 2009 Robert Andino or Chad Moeller were rated as well above replacement level. Sometime last year they fixed it, and every single WARP total was rebaselined as a result.

This isn't something that's going to happen all the time. It was a one time fix of an obvious flaw.

Link to comment
Share on other sites

I don't put any weight on the projections while they're still in flux. This year the process has been a little more turbulent than in the past. Hopefully they'll get the kinks worked out. If not, other systems will pass them by.

To me, all of these projection systems are just for entertainment value anyway. It's like trying to project what the high temperature will be next March 10. You might say the "weighted mean" is 58 degrees, but the fact is it could be anywhere between about 35 and 80 degrees. Markakis' OPS could be .877, or .811, significantly higher or lower, or somewhere in between.

That said, PECOTA hasn't performed very well the last few years compared to some of the other systems out there, from what I've read.

Link to comment
Share on other sites

I don't put any weight on the projections while they're still in flux. This year the process has been a little more turbulent than in the past. Hopefully they'll get the kinks worked out. If not, other systems will pass them by.

But why are they still in flux? Why put out the projections to begin with, only to have to keep changing them?

Link to comment
Share on other sites

But why are they still in flux? Why put out the projections to begin with, only to have to keep changing them?

I don't know all the reasons, but I know that they're in the process of going from Nate Silver's method, which apparently involved a bunch of gigantic Excel spreadsheets, to a more modern database engine. They're also bringing in a bunch of new talent to help (and bringing Clay Davenport on full time). And the system is being tweaked to make it more accurate all the time. This year has been a perfect storm of things coming together to screw up the process.

Hopefully it'll be much better in the future. Or, like I said, other systems will be taken more seriously than PECOTA.

Link to comment
Share on other sites

I don't know all the reasons, but I know that they're in the process of going from Nate Silver's method, which apparently involved a bunch of gigantic Excel spreadsheets, to a more modern database engine. They're also bringing in a bunch of new talent to help (and bringing Clay Davenport on full time). And the system is being tweaked to make it more accurate all the time. This year has been a perfect storm of things coming together to screw up the process.

Hopefully it'll be much better in the future. Or, like I said, other systems will be taken more seriously than PECOTA.

What data do they use to formulate these projections? How much of it is concrete, and how much of it is subjective?

In a much more primitive way, many of us on this board could make similar projections on our own and likely be within 3 games of PECOTA.

Just wondering.

Link to comment
Share on other sites

What data do they use to formulate these projections? How much of it is concrete, and how much of it is subjective?

In a much more primitive way, many of us on this board could make similar projections on our own and likely be within 3 games of PECOTA.

Just wondering.

Boiled down to its basics, PECOTA takes a bunch of player attributes (normalized hitting/pitching numbers, fielding, age, body types, and others), builds lists of comparable players, weights the comps, then uses them to generate projections which are normalized for park, run context, etc.

As an example, before last season it thought that 1978 Andre Dawson, 1988 Ellis Burks, 1975 Dwight Evans, and 1973 Gary Maddox were the four most-comparable players to 2009 Adam Jones. It used those guys' performances, along with weighted averages of a much longer list of players, to determine a projection for Adam Jones. I don't think any of it is subjective until you get to the projected playing time on the depth charts.

Link to comment
Share on other sites

Boiled down to its basics, PECOTA takes a bunch of player attributes (normalized hitting/pitching numbers, fielding, age, body types, and others), builds lists of comparable players, weights the comps, then uses them to generate projections which are normalized for park, run context, etc.

As an example, before last season it thought that 1978 Andre Dawson, 1988 Ellis Burks, 1975 Dwight Evans, and 1973 Gary Maddox were the four most-comparable players to 2009 Adam Jones. It used those guys' performances, along with weighted averages of a much longer list of players, to determine a projection for Adam Jones. I don't think any of it is subjective until you get to the projected playing time on the depth charts.

Do you have the comps for the 2010 season?

It would be interesting making some projections based on the comps and see how close they are to PECOTA's final numbers.

Link to comment
Share on other sites

Boiled down to its basics, PECOTA takes a bunch of player attributes (normalized hitting/pitching numbers, fielding, age, body types, and others), builds lists of comparable players, weights the comps, then uses them to generate projections which are normalized for park, run context, etc.

As an example, before last season it thought that 1978 Andre Dawson, 1988 Ellis Burks, 1975 Dwight Evans, and 1973 Gary Maddox were the four most-comparable players to 2009 Adam Jones. It used those guys' performances, along with weighted averages of a much longer list of players, to determine a projection for Adam Jones. I don't think any of it is subjective until you get to the projected playing time on the depth charts.

Gotcha. Okay, so another question: who determines the comps? I'm not saying that this is a bad system or anything, but scientific results based on subjective statistics still produce subjective results, IMO.

Again, I think that those player comparisons are pretty good, and likely be similar to what I'd pick out, but what determines the comps other than a bunch of (I'm sure very highly qualified) guys sitting around saying, "You know who he reminds me of?"

Link to comment
Share on other sites

Do you have the comps for the 2010 season?

It would be interesting making some projections based on the comps and see how close they are to PECOTA's final numbers.

Final PECOTA cards aren't out yet, and that's where the comp lists reside. Betas are there, but you need to subscribe to see either. When they come out I'll post a few, but it's probably against both site's policies to do more than that.

Gotcha. Okay, so another question: who determines the comps? I'm not saying that this is a bad system or anything, but scientific results based on subjective statistics still produce subjective results, IMO.

Again, I think that those player comparisons are pretty good, and likely be similar to what I'd pick out, but what determines the comps other than a bunch of (I'm sure very highly qualified) guys sitting around saying, "You know who he reminds me of?"

Not subjective, it's forumla based. Each player's stats are neutralized, then somehow scored for similarity to everyone else's. Similar processes are used for age, position, fielding ability, height, weight, translated minor league numbers, etc. No committee of random guys saying Chad Moeller reminds them of Al Pardo.

Link to comment
Share on other sites

But why are they still in flux? Why put out the projections to begin with, only to have to keep changing them?

From afar, it appears that they can't make up their mind about how to go about their business. Nothing wrong with having a model in a state of flux, but these guys are acting like bozo's by releasing one rough draft after another rough draft. They should stick with a system they have confidence in until they have more confidence in another one. This crap where they keep changing their published numbers every few days just makes them look like the Keystone Kops...

Link to comment
Share on other sites

BP is still fooling around....latest changes in bold

Someone at BP woke up yesterday and decided they hated us. The individual PECOTAs have been revised and almost every Oriole was downgraded offensively. To name a few:

Markakis .811 (was .877) .846

Jones .767 (was .851) .799

Reimold .790 (was .815) .827

Atkins .756 (was .811) .786

Wieters .790 (was .840) .827

Now they have us scoring 794 runs, up from 767 a week ago.

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.


×
×
  • Create New...