Brewer PECOTA stats

[TooLiveBrew]

My understanding is fine. But hey, you argue that it isn't supposed to predict 2007 at all[/i], and in the next sentence say that PECOTA projected them as impact players for 2007.

 

No, he said it doesn't predict 2007 at all by your framing of the argument.

[RobertR]

FWIW, I believe that PECOTA probably underrates Sheets's IP and WARP numbers because it likely can't distinguish the difference between types of injuries. I'm ready to throw up my hands and say don't expect more than 180 IP out of Sheets because he'll hurt something freakishly at some point, but I also distinguish that he doesn't have chronic shoulder and elbow injuries and we shouldn't necessarily expect his IP to decline from its current level.

 

Hint to fantasy leaguers.

 

Robert

[Ennder]

Whatever, you are pointing at a stat that isn't supposed to predict the current season at all for your argument. The system didn't fail, your

understanding of what stats to look at did. PECOTA projected all three of those players as impact players for 2007 and all three were that is a success.

My understanding is fine. But hey, you argue that it isn't supposed to predict 2007 at all, and in the next sentence say that PECOTA projected them as impact players for 2007.

 

But hey, I guess almost all those players rated as higher than Tulo will be "impact players" sometime during their careers. I wonder what percentage of them will make major league rosters.

No I said the upside stat is not supposed to be predicting 2007, their OPS projection etc were pretty much spot on. Those stats are SUPPOSED to predict 2007.

 

The upside stat is exactly what it sounds like, the players upside. I would expect only a small percentage of players to actually reach that number. It is just a very obscure stat for you to be pulling out and saying the system is bad because of.

[igor67]

AKA Pete it looks like you really do misunderstand what PECOTA gives in terms of projections that no other system does. Namely it gives you more than just the mean it let's you estimate the likely variance in player performance. A projection system is not going to reliably predict player's actual performance beyond a certain point. It's way more useful to have a mean (50th percent) projection and know what the down and upside of a player looks like (25th and 75th percentiles). Some players are going to have fairly narrow bands others will have big spreads.

[TooLiveBrew]

I know the BBTF guys do utilize optimistic & pessimistic lines, but they publish their mean projections. That is, except for the guys they 'spotlight.' I'd guess if you subscribe to their site you can access the rest of the data, but that's just a guess.

[rluzinski]

Namely it gives you more than just the mean it let's you estimate the likely variance in player performance.

 

Has BP ever gone back to prove the predictive value of those reported variances? I don't think they have.

[pete]

The PECOTA top 100 prospects list isn't an obscure stat. I merely pointed out how poorly it performed. If you want to say that it is claimed to be for careers, you should just quote the relative material or give a url. I don't really think it matters much as the first year is so bad.

 

And for crying out loud why are you talking about the upside stat?

 

Better yet, is there a peer reviewed analysis of PECOTA methodologies? How about its track record? I know Nate Silver loves it, but I think he is talking about his bank account.

[Ennder]

I'm referring to the upside stat because the numbers you pulled out are their upside ranking and they weren't even presented as an actual prospect list. That is why I was so confused by your statements.

 

• The Upside rating is designed to evaluate players in terms of value in their pre-free agency seasons. They are not, strictly speaking, intended to speak to who is going to have the most valuable career.

So there is no way for you to even begin to judge how good those rankings are until these guys hit free agency and you have to keep in mind these are not park adjusted and do include defense. Tulo was a below average hitter outside of Coors last year and Braun's defense was bad enough it offset a huge portion of his offense. This is also a newer part of their system so it is like looking at 'new technology'. The point is the system itself was not meant to try to predict which rookies would have the best 2007, if you wanted to know that you'd have to look at individual projections.