2008 Brewers ZiPS Projection

[TooLiveBrew]

Being pessimistic (or as they term it, "Mean"-istic) would seem to inherently miss breakout seasons. As a general rule, though, I like it - as it'd be more irresponsible imho for them to be projecting better than what many players end up doing.

[Ennder]

PECOTA is the most accurate, ZiPs is generally pretty mediocre.

 

As you would expect most of them are pretty iffy on pitching but thats because the people who compare the systems always do it wrong. ERA is a very weak stat and it is the one they usually use.

[Ennder]

PECOTA is the most accurate, ZiPs is generally pretty mediocre.

 

As you would expect most of them are pretty iffy on pitching but thats because the people who compare the systems always do it wrong. ERA is a very weak stat and it is the one they usually use.

[brewswinga2]

Top Near-Age Offensive Comps: Ken Keltner, Les Bell

 

There isn't a better near-age offensive comp than these two guys? They've never had a season as good as the one Braun just had and they played in the 1920s and 1940s.

[brewswinga2]

Top Near-Age Offensive Comps: Ken Keltner, Les Bell

 

There isn't a better near-age offensive comp than these two guys? They've never had a season as good as the one Braun just had and they played in the 1920s and 1940s.

[rluzinski]

I would agree on both counts, though they almost always miss the breakouts, which to me, is why projections are made.

 

Most projection systems aren't built to find breakouts. The foundation of most of them is simply a weighted average of the last three or four years, regressed and adjusted for age. Tangotiger's Marcel projection system only does that (he even ignores minor league numbers) and can hang with any projection system, including PECOTA. Maybe PECOTA is better at finding the break outs, but they don't consistently beat the other systems overall, from what I've seen at least. I don't think any projection system has proved itself far superior to the rest.

 

Projections are the good, safe bet but they can never be too accurate. A 600 AB sample of a player's true ability has a lot of error associated with it to begin with.

[rluzinski]

I would agree on both counts, though they almost always miss the breakouts, which to me, is why projections are made.

 

Most projection systems aren't built to find breakouts. The foundation of most of them is simply a weighted average of the last three or four years, regressed and adjusted for age. Tangotiger's Marcel projection system only does that (he even ignores minor league numbers) and can hang with any projection system, including PECOTA. Maybe PECOTA is better at finding the break outs, but they don't consistently beat the other systems overall, from what I've seen at least. I don't think any projection system has proved itself far superior to the rest.

 

Projections are the good, safe bet but they can never be too accurate. A 600 AB sample of a player's true ability has a lot of error associated with it to begin with.

[danzig6767]

Projections are the good, safe bet but they can never be too accurate.

 

When you're guessing at something in large volumes, being wrong quite often is almost inevitable.

 

[danzig6767]

Projections are the good, safe bet but they can never be too accurate.

 

When you're guessing at something in large volumes, being wrong quite often is almost inevitable.

 

[Ennder]

Maybe PECOTA is better at finding the break outs, but they don't consistently beat the other systems overall, from what I've seen at least

 

Each of the past 3 years they have shown the best correlation for OPS at least and by a pretty strong margin. I haven't seen a good study done on pitching since they all focus on ERA which is the wrong stat. I spend a lot of time looking at the various projections and by far PECOTA is the most valuable for rate stats and baseballforecaster is a pretty strong one for identifying breakout players. ZiPs is just marcels with a small tweak and those are weaker for sure, especially on fringe players.

 

I don't see anything too terrible in those projections. I think the pitcher ERA's are too good unless we move Braun or Weeks to a new position but I can't really argue with it. I expect Weeks and Hardy to be a little better than those numbers suggest but given their careers so far I can understand why they have conservative projections.

[Ennder]

Maybe PECOTA is better at finding the break outs, but they don't consistently beat the other systems overall, from what I've seen at least

 

Each of the past 3 years they have shown the best correlation for OPS at least and by a pretty strong margin. I haven't seen a good study done on pitching since they all focus on ERA which is the wrong stat. I spend a lot of time looking at the various projections and by far PECOTA is the most valuable for rate stats and baseballforecaster is a pretty strong one for identifying breakout players. ZiPs is just marcels with a small tweak and those are weaker for sure, especially on fringe players.

 

I don't see anything too terrible in those projections. I think the pitcher ERA's are too good unless we move Braun or Weeks to a new position but I can't really argue with it. I expect Weeks and Hardy to be a little better than those numbers suggest but given their careers so far I can understand why they have conservative projections.

[rluzinski]

I'd like to see a link showing PECOTA having the best correlations by a "pretty strong margin" for 3 years running. Am I to assume it came from BP? That runs contrary to the couple of studies I've ran across, recently. Depending on the exact test used, the "winner" changed. With respect to straight linear correlation, several were pretty darn close to each other. Can't find a link ATM, though.

[rluzinski]

I'd like to see a link showing PECOTA having the best correlations by a "pretty strong margin" for 3 years running. Am I to assume it came from BP? That runs contrary to the couple of studies I've ran across, recently. Depending on the exact test used, the "winner" changed. With respect to straight linear correlation, several were pretty darn close to each other. Can't find a link ATM, though.

[TooLiveBrew]

I expect Weeks and Hardy to be a little better than those numbers suggest but given their careers so far I can understand why they have conservative projections.

Well said. I'd imagine that, since they just crunch numbers (as in, PT-wise), the averages from recent years (small sample) will say that JJ & Rickie will miss decent portions of their forecasted 2008.

[TooLiveBrew]

I expect Weeks and Hardy to be a little better than those numbers suggest but given their careers so far I can understand why they have conservative projections.

Well said. I'd imagine that, since they just crunch numbers (as in, PT-wise), the averages from recent years (small sample) will say that JJ & Rickie will miss decent portions of their forecasted 2008.

[Ennder]

I'd like to see a link showing PECOTA having the best correlations by a "pretty strong margin" for 3 years running. Am I to assume it came from BP? That runs contrary to the couple of studies I've ran across, recently. Depending on the exact test used, the "winner" changed. With respect to straight linear correlation, several were pretty darn close to each other. Can't find a link ATM, though.

 

http://lanaheimangelfan.blogspot.com/search?q=pecota

 

I finally tested Pecota vs the others. For hitters with minimum of 500 AB, Pecota had an r = .736. Very impressive, beating Ron Shandler's second place of .702. Their method of finding comparable players might just be on to something

 

I've added in Marcel, Bill James, and Ron Shandler's projection OPS to the mix. I'm only looking at 114 players who had 500 or more AB, and I had to eliminate a few (Dan Uggla and Hanley Ramirez among others) because not all systems projected minor leaguers. Kenji Johjima had to go, as only ZIPs and CHONE even tried to project him (very well I might add). I don't have access to Baseball Prospectus 2006, so I'll look at Pecota in a week or so.

 

Here's the results:

 

Shandler .702

James .685

ZIPs .684

Chone .677

Marcel .664

 

I don't have links for the other years but I know for a fact it is 3 years in a row that PECOTA has shown the best correlation for hitting. Pitching is more of a crapshoot but they always look at ERA which is such a random stat that I wouldn't expect a strong correlation.

[Ennder]

I'd like to see a link showing PECOTA having the best correlations by a "pretty strong margin" for 3 years running. Am I to assume it came from BP? That runs contrary to the couple of studies I've ran across, recently. Depending on the exact test used, the "winner" changed. With respect to straight linear correlation, several were pretty darn close to each other. Can't find a link ATM, though.

 

http://lanaheimangelfan.blogspot.com/search?q=pecota

 

I finally tested Pecota vs the others. For hitters with minimum of 500 AB, Pecota had an r = .736. Very impressive, beating Ron Shandler's second place of .702. Their method of finding comparable players might just be on to something

 

I've added in Marcel, Bill James, and Ron Shandler's projection OPS to the mix. I'm only looking at 114 players who had 500 or more AB, and I had to eliminate a few (Dan Uggla and Hanley Ramirez among others) because not all systems projected minor leaguers. Kenji Johjima had to go, as only ZIPs and CHONE even tried to project him (very well I might add). I don't have access to Baseball Prospectus 2006, so I'll look at Pecota in a week or so.

 

Here's the results:

 

Shandler .702

James .685

ZIPs .684

Chone .677

Marcel .664

 

I don't have links for the other years but I know for a fact it is 3 years in a row that PECOTA has shown the best correlation for hitting. Pitching is more of a crapshoot but they always look at ERA which is such a random stat that I wouldn't expect a strong correlation.