UZR projections, courtesy of Jeff Zimmerman

I'd make this into a fanpost, but I don't have the desire to go through and pick out the projections that intrigue me (like last season).

So have at it...

2 months ago baetown415 22 comments 1 recs |

Story-email Email Printer Print

Comments

Does Anyone Else Think These Are Being Way Too Regressed.

Looking at the spreadsheet I can not help but notice very few really bad or really good and way too many in the middle. Yes, i understand bell curves, but I also understand baseball performance and I just think this has narrowed the likely range of actual performaces way to close to both the mean and the average performance. The spead to me is just obviuosly much greater then these projections would predict.

by giantsrainman on Nov 14, 2009 11:20 PM PST reply actions 0 recs

I think the heavy regression is probably because of two reasons:

1. An attempt to strip out measurement errors from previous seasons

and

2. Because the range of performance in a year (from Franklin G to Adam D) is much wider than the range of true-talent level, which I believe is what the UZR projections are trying to say (since we should expect players to play at their true-talent level going forward, fluctuations be damned).

Aaron King is still my homeboy... iffy mechanics and all

McFAQ for all you newcomers out there.

GET THAT VORP AND WHIP SH!T OUTTA HERE!!!

by baetown415 on Nov 14, 2009 11:45 PM PST up reply actions 0 recs

I Understand The Goal Of Identifying True Talent Level

I just think that this data does not show enough varience in the range of true talent level. It ove regresses because intentionally or not it has assumed not enough varience in the true talent level of the players.

by giantsrainman on Nov 15, 2009 1:15 AM PST up reply actions 0 recs

Much like Bill James.

Matt Cain: throwing complete game shutouts since 06'. No big deal.

by cain1rstballothof on Nov 14, 2009 11:48 PM PST up reply actions 0 recs

It’s always best to project defensive numbers conservatively moving forward.

Bay City Ball

by xanthan on Nov 15, 2009 7:07 AM PST up reply actions 0 recs

Why? I would thinik the goal would be accuracy not conservativeness.

By accuracy I mean getting the most extreme players right in how extreme they are. To me being right with how extreme a player is by say 5% on 50% of the extreme players is much more useful then being right with how extreme a player is by say 20% on 75% of the extreme players.

by giantsrainman on Nov 15, 2009 1:26 PM PST up reply actions 0 recs

You can’t really project extreme players.

Bay City Ball

by xanthan on Nov 15, 2009 1:31 PM PST up reply actions 0 recs

In other words … If you aim for small you will miss small.

Who’s brain did you bring me?
Brain SabeanOranother.

by daveinexile on Nov 15, 2009 2:30 PM PST up reply actions 0 recs

Then I See No Value In Using These Projections In Constructing A Team

Fantasy or real.

by giantsrainman on Nov 15, 2009 3:11 PM PST up reply actions 0 recs

I think it would be useful , for us any ways, to use the projects as a baseline. I think he has the last couple years projections up so a curious fan could choose a given player and see if he is routinely beating ( or failing) the projection. That would let a fan target a player to see if his talent level is marked off the average and procced from there.

Who’s brain did you bring me?
Brain SabeanOranother.

by daveinexile on Nov 15, 2009 3:38 PM PST up reply actions 0 recs

Matt Cain: throwing complete game shutouts since 06'. No big deal.

by cain1rstballothof on Nov 15, 2009 3:56 PM PST up reply actions 0 recs

Whoops, +1 for GRM.

Matt Cain: throwing complete game shutouts since 06'. No big deal.

by cain1rstballothof on Nov 15, 2009 3:56 PM PST up reply actions 0 recs

Actually

I think they’re more useful on a team level than on the individual level. As you point out, the heavy regressions mean that they are over conservative with individual players in an attempt to counter random variation. If you pool a group of players together, however, the size of the sample will also serve to cancel some of the random variations, thus making the sum of the projections of an entire team more representative of reality than individual UZRs.

VAE PVTO DEVS FIO

by Bhaakon on Nov 15, 2009 6:53 PM PST up reply actions 0 recs

projections always predict more players to move towards the mean. It’s almost mathematically impossible not to.

This is one amongst many reasons why projections should be taken with a grain of salt. And some lime and tequila!

Please hit better, Randy Winn.

by oldjacket on Nov 15, 2009 8:04 AM PST up reply actions 0 recs

They need a Safe Harbor Statement

This Site contains certain forward-looking statements about UZR and its effect. Forward-looking statements may contain words such as "expect," "believe," "may," "can," "should," "will," "forecast" or similar expressions, and include the assumptions that underlie such statements. These statements are subject to known and unknown risks and uncertainties that could cause actual results to differ materially from those expressed or implied by such statements, , including but not limited to the performance of a player’s future defense, which is subject to many factors, including uncertain heath, conditioning, and team defensive positioning .We assumes no obligation to update the information on this Site.

Obviously any links in the above post are probably NSFW

by jctGamer on Nov 15, 2009 8:55 AM PST up reply actions 0 recs