Free Trial Ad
Why Subscribe?
  • Player/Prospect News
  • Exclusive Insider Info
  • Members-Only Forums
  • Exclusive Videos
  • Subscribe Now!
Inbox
Reply to TopicPost New Topic
  Page of 2  Next >

A's Organization 2013 Oliver Projections

Posted: 02/10/2013 10:54 PM

A's Organization 2013 Oliver Projections 


I just noticed that the 2013 set of Oliver projections are up on Fangraphs.  From my quick brushing up, I understand they are similar to Marcels (three-year weighted averages) but include MLE's for minor-leaguers.  They also include fielding runs on a play-by-play system distinct from UZR and Plus/Minus; I have no idea how he projects fielding runs from minor-league data.

These look pretty similar to other systems, but the projections for those with little or no major league playing time make them interesting eyeballing.  For example, Andy Parrino is projected to produce a greater WAR than Jed Lowrie (takes playing time into account). 

You can also look up A's minor leaguers who don't show up on the page linked above.  According to Oliver, Michael Choice is just about ready for the big leagues, Miles Head is better than Josh Donaldson, and Addison Russell hits at Penny 2012 level right now.

Last edited 02/10/2013 10:54 PM by Qwerty75

Reply | Quote

Posted: 02/10/2013 11:10 PM

Re: A's Organization 2013 Oliver Projections 


seems like Steamer are more realistic.
Reply | Quote
Avatar

Posted: 02/11/2013 9:23 AM

Re: A's Organization 2013 Oliver Projections 


I think I could make better guesses than this projection system. Cespedes projected to have the lowest WAR among all of our starters? That's ridiculous.

Daric Barton, Shane Peterson, Michael Taylor, Parrino, Norris and Sogard ALL have higher projected WAR than Cespedes. Funny.

Not sure how it come up with a .266/.333/.448 line for Cespedes. It projects him to decline in every category. His only pro experience was last season and he shattered that production. Where do they come up with this shyt?

Norris has the 2nd highest projected WAR on the team? Odd.
Reply | Quote

Posted: 02/11/2013 9:53 AM

Re: A's Organization 2013 Oliver Projections 



BanditoB wrote: I think I could make better guesses than this projection system. Cespedes projected to have the lowest WAR among all of our starters? That's ridiculous.

Daric Barton, Shane Peterson, Michael Taylor, Parrino, Norris and Sogard ALL have higher projected WAR than Cespedes. Funny.

Not sure how it come up with a .266/.333/.448 line for Cespedes. It projects him to decline in every category. His only pro experience was last season and he shattered that production. Where do they come up with this shyt?

Norris has the 2nd highest projected WAR on the team? Odd.
Cespedes has only one year of data to go on and projections regress small samples to the mean heavily.  You'll see that other projection systems predict a similar drop from him next year.  

This is a case where the limitations of numbers-crunching come up.  The projection systems "see" the player's performance record and the data for the population (other ML outfielders, aging curves, etc.) in its calculations.  What we can "see" beyond the stats is his elite athleticism and readiness to make adjustments.  Those who want to take account of these qualitative factors will want to look at his ceiling and not the mid-level projection which takes what might be relevant scouting insights as statistical noise.  We're also fans of the team, so we're naturally disposed to be optimistic about his potential outcomes.
Reply | Quote
Avatar

Posted: 02/11/2013 10:12 AM

Re: A's Organization 2013 Oliver Projections 



Qwerty75 wrote:
BanditoB wrote: I think I could make better guesses than this projection system. Cespedes projected to have the lowest WAR among all of our starters? That's ridiculous.

Daric Barton, Shane Peterson, Michael Taylor, Parrino, Norris and Sogard ALL have higher projected WAR than Cespedes. Funny.

Not sure how it come up with a .266/.333/.448 line for Cespedes. It projects him to decline in every category. His only pro experience was last season and he shattered that production. Where do they come up with this shyt?

Norris has the 2nd highest projected WAR on the team? Odd.
Cespedes has only one year of data to go on and projections regress small samples to the mean heavily.  You'll see that other projection systems predict a similar drop from him next year.  

This is a case where the limitations of numbers-crunching come up.  The projection systems "see" the player's performance record and the data for the population (other ML outfielders, aging curves, etc.) in its calculations.  What we can "see" beyond the stats is his elite athleticism and readiness to make adjustments.  Those who want to take account of these qualitative factors will want to look at his ceiling and not the mid-level projection which takes what might be relevant scouting insights as statistical noise.  We're also fans of the team, so we're naturally disposed to be optimistic about his potential outcomes.
Sorry for my ignorance, but what is "the mean" when we're talking about Cespedes? He was an all-time great in Cuba, and the MVP of a division winner as a rookie in MLB.
Reply | Quote

Posted: 02/11/2013 10:23 AM

Re: A's Organization 2013 Oliver Projections 



BanditoB wrote:
Qwerty75 wrote:
BanditoB wrote: I think I could make better guesses than this projection system. Cespedes projected to have the lowest WAR among all of our starters? That's ridiculous.

Daric Barton, Shane Peterson, Michael Taylor, Parrino, Norris and Sogard ALL have higher projected WAR than Cespedes. Funny.

Not sure how it come up with a .266/.333/.448 line for Cespedes. It projects him to decline in every category. His only pro experience was last season and he shattered that production. Where do they come up with this shyt?

Norris has the 2nd highest projected WAR on the team? Odd.
Cespedes has only one year of data to go on and projections regress small samples to the mean heavily.  You'll see that other projection systems predict a similar drop from him next year.  

This is a case where the limitations of numbers-crunching come up.  The projection systems "see" the player's performance record and the data for the population (other ML outfielders, aging curves, etc.) in its calculations.  What we can "see" beyond the stats is his elite athleticism and readiness to make adjustments.  Those who want to take account of these qualitative factors will want to look at his ceiling and not the mid-level projection which takes what might be relevant scouting insights as statistical noise.  We're also fans of the team, so we're naturally disposed to be optimistic about his potential outcomes.
Sorry for my ignorance, but what is "the mean" when we're talking about Cespedes? He was an all-time great in Cuba, and the MVP of a division winner as a rookie in MLB.
My guess is that it's the major-league average for players at his position(s).  So that would be CF, LF, and DH according to playing time at the respective positions.  Not sure how this system handles it; it might simply go by the position where he spent most time (LF) or OFs as a whole.  It looks like Oliver only looks at minor leagues in the US system, so it doesn't incorporate data from the Cuban leagues, nor would it use data from NPB for Japanese players.

In Cespedes' case, I think what hurts him most in projections is his performance in defensive metrics.  He was *so* bad last year that even regressing his fielding runs to the mean is a good-sized negative.  Those who put stock in his ability to improve in the field with more experience based on learning ability and athleticism will guesstimate him as around average or better in an OF corner.  Projection systems based simply on the statistical record won't give him that benefit of the doubt.

Last edited 02/11/2013 10:26 AM by Qwerty75

Reply | Quote

Posted: 02/11/2013 10:35 AM

Re: A's Organization 2013 Oliver Projections 



zitoforpres wrote: seems like Steamer are more realistic.
From the looks of it, Oliver uses minor league data more than Steamer.  This shows up in the projections for players with a short major league record who've performed better than one would expect from their minor league numbers.
Reply | Quote

Posted: 02/11/2013 8:46 PM

Re: A's Organization 2013 Oliver Projections 


I actually am pretty sure it simply uses league mean rather than position specific. Cespedes certainly has upside that no projection system will predict going forward.  
Qwerty75 wrote:
BanditoB wrote:
Qwerty75 wrote:
BanditoB wrote: I think I could make better guesses than this projection system. Cespedes projected to have the lowest WAR among all of our starters? That's ridiculous.

Daric Barton, Shane Peterson, Michael Taylor, Parrino, Norris and Sogard ALL have higher projected WAR than Cespedes. Funny.

Not sure how it come up with a .266/.333/.448 line for Cespedes. It projects him to decline in every category. His only pro experience was last season and he shattered that production. Where do they come up with this shyt?

Norris has the 2nd highest projected WAR on the team? Odd.
Cespedes has only one year of data to go on and projections regress small samples to the mean heavily.  You'll see that other projection systems predict a similar drop from him next year.  

This is a case where the limitations of numbers-crunching come up.  The projection systems "see" the player's performance record and the data for the population (other ML outfielders, aging curves, etc.) in its calculations.  What we can "see" beyond the stats is his elite athleticism and readiness to make adjustments.  Those who want to take account of these qualitative factors will want to look at his ceiling and not the mid-level projection which takes what might be relevant scouting insights as statistical noise.  We're also fans of the team, so we're naturally disposed to be optimistic about his potential outcomes.
Sorry for my ignorance, but what is "the mean" when we're talking about Cespedes? He was an all-time great in Cuba, and the MVP of a division winner as a rookie in MLB.
My guess is that it's the major-league average for players at his position(s).  So that would be CF, LF, and DH according to playing time at the respective positions.  Not sure how this system handles it; it might simply go by the position where he spent most time (LF) or OFs as a whole.  It looks like Oliver only looks at minor leagues in the US system, so it doesn't incorporate data from the Cuban leagues, nor would it use data from NPB for Japanese players.

In Cespedes' case, I think what hurts him most in projections is his performance in defensive metrics.  He was *so* bad last year that even regressing his fielding runs to the mean is a good-sized negative.  Those who put stock in his ability to improve in the field with more experience based on learning ability and athleticism will guesstimate him as around average or better in an OF corner.  Projection systems based simply on the statistical record won't give him that benefit of the doubt.
Reply | Quote

Posted: 02/11/2013 10:22 PM

RE: A's Organization 2013 Oliver Projections 


enough of the pecota and oliver projections.

let's get some zfp, oakman, greenmachine, and emkey projections.

oh wait emkey isn't here anymore. at least we have zfp...
Reply | Quote

Posted: 02/12/2013 12:19 AM

Re: A's Organization 2013 Oliver Projections 



vignette17 wrote: I actually am pretty sure it simply uses league mean rather than position specific. Cespedes certainly has upside that no projection system will predict going forward.  
Qwerty75 wrote: My guess is that it's the major-league average for players at his position(s).  So that would be CF, LF, and DH according to playing time at the respective positions.  Not sure how this system handles it; it might simply go by the position where he spent most time (LF) or OFs as a whole.  It looks like Oliver only looks at minor leagues in the US system, so it doesn't incorporate data from the Cuban leagues, nor would it use data from NPB for Japanese players.
Thanks for the clarification.  League average would explain why the larger falloff in performance than in other projections.  Is Oliver alone in regressing to the league average rather than a positional average?  I figured including WAR in the projections would lead him to use a positional average for consistency with respect to positional replacement level.
Reply | Quote

Posted: 02/12/2013 6:43 AM

Re: A's Organization 2013 Oliver Projections 



Qwerty75 wrote:
vignette17 wrote: I actually am pretty sure it simply uses league mean rather than position specific. Cespedes certainly has upside that no projection system will predict going forward.  
Qwerty75 wrote: My guess is that it's the major-league average for players at his position(s).  So that would be CF, LF, and DH according to playing time at the respective positions.  Not sure how this system handles it; it might simply go by the position where he spent most time (LF) or OFs as a whole.  It looks like Oliver only looks at minor leagues in the US system, so it doesn't incorporate data from the Cuban leagues, nor would it use data from NPB for Japanese players.
Thanks for the clarification.  League average would explain why the larger falloff in performance than in other projections.  Is Oliver alone in regressing to the league average rather than a positional average?  I figured including WAR in the projections would lead him to use a positional average for consistency with respect to positional replacement level.
I think Marcel is the only "major" projection system that regresses toward the overall league mean (excluding pitchers batting)--which is an intentional limitation.

Here's how Oliver determines the mean:

The values that each player is regressed to are determined by information about the player other than his performance—his age, position, and level played. A 19-year-old in Double-A will be regressed to a higher mean than a 23-year-old at that same level, as the team presumably considered the 19-year-old to have more talent in aggressively promoting him. Similarly, a first baseman will be regressed to a higher home run rate than a shortstop, as we know that, on average, first basemen hit more home runs than shortstops.
Reply | Quote

Posted: 02/12/2013 7:30 AM

Re: A's Organization 2013 Oliver Projections 



AthertonA wrote:
Qwerty75 wrote:
vignette17 wrote: I actually am pretty sure it simply uses league mean rather than position specific. Cespedes certainly has upside that no projection system will predict going forward.  
Qwerty75 wrote: My guess is that it's the major-league average for players at his position(s).  So that would be CF, LF, and DH according to playing time at the respective positions.  Not sure how this system handles it; it might simply go by the position where he spent most time (LF) or OFs as a whole.  It looks like Oliver only looks at minor leagues in the US system, so it doesn't incorporate data from the Cuban leagues, nor would it use data from NPB for Japanese players.
Thanks for the clarification.  League average would explain why the larger falloff in performance than in other projections.  Is Oliver alone in regressing to the league average rather than a positional average?  I figured including WAR in the projections would lead him to use a positional average for consistency with respect to positional replacement level.
I think Marcel is the only "major" projection system that regresses toward the overall league mean (excluding pitchers batting)--which is an intentional limitation.

Here's how Oliver determines the mean:

The values that each player is regressed to are determined by information about the player other than his performance—his age, position, and level played. A 19-year-old in Double-A will be regressed to a higher mean than a 23-year-old at that same level, as the team presumably considered the 19-year-old to have more talent in aggressively promoting him. Similarly, a first baseman will be regressed to a higher home run rate than a shortstop, as we know that, on average, first basemen hit more home runs than shortstops.
This is kinda strange.  This would seem to overrate players young for leagues and first basemen.  Aggressive promoters like the Tigers and Mets would have overrated prospects, and Josh Donaldson could change his projection just by moving to 1B.  That can't be right, can it?
Reply | Quote
Avatar

Posted: 02/12/2013 8:29 AM

Re: A's Organization 2013 Oliver Projections 



WaddellCanseco wrote:
AthertonA wrote:
I think Marcel is the only "major" projection system that regresses toward the overall league mean (excluding pitchers batting)--which is an intentional limitation.

Here's how Oliver determines the mean:

The values that each player is regressed to are determined by information about the player other than his performance—his age, position, and level played. A 19-year-old in Double-A will be regressed to a higher mean than a 23-year-old at that same level, as the team presumably considered the 19-year-old to have more talent in aggressively promoting him. Similarly, a first baseman will be regressed to a higher home run rate than a shortstop, as we know that, on average, first basemen hit more home runs than shortstops.
This is kinda strange.  This would seem to overrate players young for leagues and first basemen.  Aggressive promoters like the Tigers and Mets would have overrated prospects, and Josh Donaldson could change his projection just by moving to 1B.  That can't be right, can it?

My assumption is that Oliver is doing a form of crowdsourcing, but the crowd is MLB front offices. They're weighting in what teams think of the players, under the assumption that teams are acting rationally for the good of their team and, therefore, positional changes and promotions provide meaningful information about a player.
Reply | Quote

Posted: 02/12/2013 8:42 AM

Re: A's Organization 2013 Oliver Projections 



WaddellCanseco wrote:
AthertonA wrote:
 
I think Marcel is the only "major" projection system that regresses toward the overall league mean (excluding pitchers batting)--which is an intentional limitation.

Here's how Oliver determines the mean:

The values that each player is regressed to are determined by information about the player other than his performance—his age, position, and level played. A 19-year-old in Double-A will be regressed to a higher mean than a 23-year-old at that same level, as the team presumably considered the 19-year-old to have more talent in aggressively promoting him. Similarly, a first baseman will be regressed to a higher home run rate than a shortstop, as we know that, on average, first basemen hit more home runs than shortstops.
This is kinda strange.  This would seem to overrate players young for leagues and first basemen.  Aggressive promoters like the Tigers and Mets would have overrated prospects, and Josh Donaldson could change his projection just by moving to 1B.  That can't be right, can it?
The A's could move Donaldson to 1B to get a better Oliver projection, but there's no incentive to game the system like that.  The question is whether or not it makes for a better projection.  If you have a SS and a 1B who have the same (park/league adjusted) batting line in A ball, does knowing their primary fielding position provide useful information in projecting those players going forward?

As Minstrel notes, this is basically relying on team decisions to provide additional, useful information.  I have always thought this was necessary, particularly for prospects and players with injury histories.  

I agree there can be issues with teams using different promotion philosophies--whether it's aggressive promotion of younger players or avoiding sending pitching prospects to rough environments (the Dodgers have often been said to avoid sending their top pitching prospects to the PCL).
Reply | Quote

Posted: 02/12/2013 9:19 AM

Re: A's Organization 2013 Oliver Projections 


Every time you introduce some element into the equation it will help for some cases and hurt for some cases. So, yes regressing by position isn't perfect but, hopefully they tested it and showed that it helps more than it hurts.

Marcel doesn't bother with that because one of it's primary features is simplicity.

You might be able to make an even better projection by somehow weighting scouts into the equation. If someone grades out to 70 power, you might regress them to a higher homer level than someone with a 30 grade. Again, this is not going to be perfect but, I wonder if it might not be better than regressing based on position.
Reply | Quote

Posted: 02/12/2013 9:45 AM

Re: A's Organization 2013 Oliver Projections 


Dan Szymborski, the guy who created/runs ZIPS, has said that he's tried to incorporate prospect rankings and draft position into his projections, but that they don't add anything useful.

From #118 here: As for pedigree, I haven't found much in the way of worth regarding regression (I checked using variables for both BA prospect position and draft pick # as inputs for estimating regression and I found them to be worthless).

I find that kind of shocking, especially for guys without much pro experience.  If a non-prospect and a 1st round pick put up the same level of performance in their first full season, are they really equally likely to succeed going forward?

Last edited 02/12/2013 9:46 AM by AthertonA

Reply | Quote

Posted: 02/12/2013 9:59 AM

Re: A's Organization 2013 Oliver Projections 



AthertonA wrote:
WaddellCanseco wrote:
AthertonA wrote:
 
I think Marcel is the only "major" projection system that regresses toward the overall league mean (excluding pitchers batting)--which is an intentional limitation.

Here's how Oliver determines the mean:

The values that each player is regressed to are determined by information about the player other than his performance—his age, position, and level played. A 19-year-old in Double-A will be regressed to a higher mean than a 23-year-old at that same level, as the team presumably considered the 19-year-old to have more talent in aggressively promoting him. Similarly, a first baseman will be regressed to a higher home run rate than a shortstop, as we know that, on average, first basemen hit more home runs than shortstops.
This is kinda strange.  This would seem to overrate players young for leagues and first basemen.  Aggressive promoters like the Tigers and Mets would have overrated prospects, and Josh Donaldson could change his projection just by moving to 1B.  That can't be right, can it?
The A's could move Donaldson to 1B to get a better Oliver projection, but there's no incentive to game the system like that.  The question is whether or not it makes for a better projection.  If you have a SS and a 1B who have the same (park/league adjusted) batting line in A ball, does knowing their primary fielding position provide useful information in projecting those players going forward?

As Minstrel notes, this is basically relying on team decisions to provide additional, useful information.  I have always thought this was necessary, particularly for prospects and players with injury histories.  

I agree there can be issues with teams using different promotion philosophies--whether it's aggressive promotion of younger players or avoiding sending pitching prospects to rough environments (the Dodgers have often been said to avoid sending their top pitching prospects to the PCL).
Right, the A's wouldn't move Donaldson to 1B to get a better Oliver projection, but they might move him there if he were a terrible 3B, which he isn't.  But moving him to 1B shouldn't affect his hitting projection.  Their reasoning seems circular.  1B guys hit better because the crappy hitting 1B are out of baseball, not because moving to 1B makes you a better hitter.  They're saying that a guy who moves to 1B must be a better hitter because if he weren't, he'd be out of baseball.  That seems strange.
Reply | Quote

Posted: 02/12/2013 10:06 AM

Re: A's Organization 2013 Oliver Projections 


Hmm, I wonder why there might be correlation between position and regression then? Prospect ranking and position both reflect information based on human evaluation so, you'd think that if one added information then both would.
Reply | Quote

Posted: 02/12/2013 10:13 AM

Re: A's Organization 2013 Oliver Projections 



WaddellCanseco wrote:

Right, the A's wouldn't move Donaldson to 1B to get a better Oliver projection, but they might move him there if he were a terrible 3B, which he isn't.  But moving him to 1B shouldn't affect his hitting projection.  Their reasoning seems circular.  1B guys hit better because the crappy hitting 1B are out of baseball, not because moving to 1B makes you a better hitter.  They're saying that a guy who moves to 1B must be a better hitter because if he weren't, he'd be out of baseball.  That seems strange.
If there's no correlation between prospect ranking and regression, it's hard to believe there should be correlation between position and regression. What am I missing here?
Reply | Quote

Posted: 02/12/2013 3:45 PM

Re: A's Organization 2013 Oliver Projections 



voxhoo wrote:
WaddellCanseco wrote:

Right, the A's wouldn't move Donaldson to 1B to get a better Oliver projection, but they might move him there if he were a terrible 3B, which he isn't.  But moving him to 1B shouldn't affect his hitting projection.  Their reasoning seems circular.  1B guys hit better because the crappy hitting 1B are out of baseball, not because moving to 1B makes you a better hitter.  They're saying that a guy who moves to 1B must be a better hitter because if he weren't, he'd be out of baseball.  That seems strange.
If there's no correlation between prospect ranking and regression, it's hard to believe there should be correlation between position and regression. What am I missing here?
The entire population of batters is already differentiated into groups with different performance levels (fielding position), of whose value relative to the others has been previously established and can be re-calibrated for accuracy.  It adds granularity to the data by putting them into stable bins.  My guess on what has come out of tying prospect ranking to historical performance data is that: 1) not enough differentiation could be seen between different prospect-player groups, or 2) the fact that the data isn't comprehensive across the player population is hard to account for.  You'll only have ranking for 100 players each year (does BA do 200?), so it doesn't apply across the entire population like fielding position does.
Reply | Quote
Reply to TopicPost New Topic
  Page of 2  Next >