DiamondView: 2009 AL East Review

by Justin Bopp on Jan 8, 2010 7:00 AM EST in Visuals Above Replacement

Our AL East DiamondView journey comes to a close, and with it, any mystery why the standings turned out the way they did in 2009. With unexpected precision, this method shows that the teams with the better starters in the ON BASE and POWER categories came in the exact same order as they appear in the standings. While the obvious remains obvious, this visual is certainly enlightening:

Al-east-standings_medium

Can anybody find a correlation in this?

Team composites after the jump.

Batter-dvc2-team-yankees_medium

Batter-dvc2-boston-team-avg_medium

Batter-dvc2-tampa-team-avg_medium

Batter-dvc2-toronto-team-avg_medium

Batter-dvc2-orioles-teamavg_medium

AL East: Yankees | Red Sox | Rays | Blue Jays | Orioles

Notes & Background:

The DiamondView Composite Player Evaluation method/tool is an attempt by the author to visualize the five classic player tools into comparable shapes. The original hypothesis was that one would eventually be able to look at an unlabled DVC and be able to describe which position he probably plays, with speedy/good glove guys probably in center or short, and slower power guys in left and first--just like everybody would think.

The data actually does bear this out, but often has surprising returns, and is rarely perfect. If you review the linked articles above and look for the position average for each, you'll see that 1st points northeast, 2nd points southeast, 3rd and RF are both relatively squarish, catchers are pretty much awful at everything, center points south, and LF is squarish pointing slightly east.

Currently, the data is collected as follows:

On Base: OBP | Power: ISO | Base Running: EQBRR | Fielding: UZR/150

All data is for 2009 and has some minimum number of PA to make the DiamondView cut. The data is neither park adjusted nor position adjusted. The players are picked for these reviews based on a combination of PT and availability of their scores in my DB. The position averages are figured using the average of players that made the DiamondView minimum PA. The team average is the average of the starters listed. The graph is a combination of programs and effects.

Opportunities and Intentions to Improve:

1. As noted above, the data isn't properly adjusted to reflect park effects nor positional value. Both would be needed for a slightly improved review of the time period.

2. Projected data will be needed to make these relevant for the upcoming year. Please advise which data set for each category you believe would be the best fit for this purpose.

3. The fielding stat (UZR/150) leaves a lot to be desired. It doesn't help with catchers at all. I'd like to hear your opinions on the best way to combine available data (Nate Silver style?).

4. An insightful commenter CoachOfEarl suggested rotating the stats to add an additional layer of subtext to the visual: power at home plate to represent scoring, on base at 1st to represent getting on base, fielding at the top to represent the rangy SS, 2Bs, and CFs, and base running at 3rd to represent higher-speed guys that make it that far. Does that make more sense than the current setup to you guys, and if so, do you think it's too late to change? Do you care? I think it's a great idea but don't want to break any subconscious visual commitment that we've already made over the course of this project.

5. Yes, Pitching DiamondView is in the works. I'm currently settling on the proper data sets to best represent every significant factor in evaluating pitching talent as well as a slight tweak to the visual so it won't be mistaken for the batter version.

Please leave your thoughts and further suggestions below.

Tweet Comment 18 comments | 0 recs |

Do you like this story?

Comments

Great stuff

I especially love the way the Rays and the Yankees are basically the opposite.

Suggestions:

1) If you’re going to be using projections, park adjusted data isn’t really necessary because projections will already do that. For positional values, just add in the positional adjustment with defense (so a +5 shortstop is actually +12.5 on defense).

2) I would use CHONE because A) it is probably the best, and B) it includes a fielding projection.

3) Use either CHONE’s fielding projections (which include catcher defense), or Steve Sommer’s.

4) I don’t really like it. That would make the graph to cluttered and it’s not really useful.

by vivaelpujols on Jan 8, 2010 9:18 AM EST reply actions

Thanks.

Rays vs. Yankees vs. Red Sox: aaaand this is why I’d like to do a pitching visual. From this alone, one might (incorrectly) assume that a team only need the most home runs, fielding be damned. We all know power is more valuable, of course, but I’d say these three are probably close enough that the pitching composite would speak volumes.

1. That makes sense.

2/3 : Going to a single source for the data would be optimal. CHONE might be the one to rely on, then.

4. I’m not sure you understood the concept, it was really just about rotating the entire thing clockwise once. Either way, comments here tell me it’s not really important anyway.

See Data Differently.
beyondtheboxscore.com | Twitter: @ justinbopp

by Justin Bopp on Jan 8, 2010 12:18 PM EST up reply actions

I agree that CHONE is probably best

Unless you wanted to do a composite, maybe combining like, CHONE, Marcel, ZiPS, Bill James and FanGraphs’ Fan Projections.

As for fielding, I agree that Steve Sommer’s defensive projections are probably the best out there. I’m still interested to see what BPro comes up with for their new defensive metric, as Colin talked about it’s development.

First it was Paige, now it's Price.
You can find me at http://www.beyondtheboxscore.com/

by Satchel Price on Jan 8, 2010 9:57 AM EST via mobile reply actions

I'd leave out Marcel, as any good projection will start with its methods and improve them.

Bill James is kinda sketchy, so I’d probably go with a mix of CHONE and ZiPS, and maybe something from the fans.

Beyond the Boxscore Not a member? Sign up.

by Sky Kalkman on Jan 8, 2010 10:05 AM EST up reply actions

Yeah, BJ and Marcel are probably the best to leave out

I’m not really clear on how James’ projections are made, but I agree that the results are often questionable. I was merely pointing out that a composite projection could be better than using a single system, and those were the systems I thought of off the top of my head.

CHONE and ZiPS are definitely the best ones, and I personally just have a mild fascination with the Fan Projections, I think that it was a great idea by Appelman and Co., it’s interesting to see the differences between what people think players will do compared to what mathematical formulas say.

Basically, I think that using multiple projection systems would be better than just using one, although obviously it would be CHONE or ZiPS if I had to choose one.

First it was Paige, now it's Price.
You can find me at http://www.beyondtheboxscore.com/

by Satchel Price on Jan 8, 2010 11:01 AM EST up reply actions

For the sake of discussion...

It might be that the Bill James projections have some value if you adjusted them to be less optimistic. They might contain information not contained elsewhere. Who knows.

Beyond the Boxscore Not a member? Sign up.

by Sky Kalkman on Jan 8, 2010 11:44 AM EST up reply actions

Isn't the uncertainty the problem itself?

See Data Differently.
beyondtheboxscore.com | Twitter: @ justinbopp

by Justin Bopp on Jan 8, 2010 12:23 PM EST up reply actions

I doubt BPro's metric is going to be any better than UZR or +/-

by vivaelpujols on Jan 8, 2010 11:40 AM EST up reply actions

I'm not sure this visual would be the best place to use a composite of various competing defensive metrics.

That said, I’ve noticed an all-too-quick dismissal of UZR for higher-level comparison purposes such as the diamondview. The key for comparison is a consistently applied measure, improved with a park adjustment.

See Data Differently.
beyondtheboxscore.com | Twitter: @ justinbopp

by Justin Bopp on Jan 8, 2010 12:22 PM EST up reply actions

Thanks for the love

on the D projections (both you and VEP). Justin, if you decide to use them be aware that it’s UZR/150 so you (or whomever applies them for you) would need to project playing time too (ie Defensive Games).

Personally I use mine for as many guys as it’s applicable (figure I should probably have enough confidence to use my own right?), Jeff’s are next because he has some position combos that I don’t have and guys that didn’t have FSRs, and then CHONE for Catchers and the guys with no MLB experience.

Beyond the Box Score | Play a Hard Nine | Twitter

by stevesommer05 on Jan 8, 2010 10:02 PM EST up reply actions

Ben Zobrist is a beast!

Representin' the West Side of Mulberry!

by SeanDubbs on Jan 8, 2010 10:22 AM EST reply actions

I'd rather the view be left as-is.

I think we’ve all started to emphasize to ourselves that area covered is not directly proportional to total contributed value on the y axis, and what that means in the context of DiamondView. It may not be perfect, but unless someone can come up with a significant improvement, I’d rather not futz about with it.

by jwiscarson on Jan 8, 2010 10:27 AM EST reply actions

This is heard loud and clear.

See Data Differently.
beyondtheboxscore.com | Twitter: @ justinbopp

by Justin Bopp on Jan 8, 2010 12:24 PM EST up reply actions

loved this series....

any chance we get to see the rest of the teams?? bench players and propects would be nice too. Thanks for the hard work

by I miss Jack Buck on Jan 8, 2010 11:47 AM EST reply actions

Thank you. Rest of the teams should be coming

…time being the issue. Starting some MBA classes on Tuesday so the output will probably slow down a bit (the mediocre writing actually takes me longer than the charts themselves).

1. The original plan was to review 2009 by a division-by-division basis, bouncing between the AL and NL weekly. That would put us around mid Feb to go through each division, assuming I can keep that pace, which is wildly optimistic unless I just post the charts with less naive commentary.

2. Bench players and prospects might be a possibility for the projected version of this, assuming the player is covered in the projected data set. As is, the comparison is strictly comparison based and requires some minimum number of PA for viability purposes.

3. I would consider stopping the 2009 review and jumping straight to the 2010 projection if that would be more appreciated, but it would take a little more time to BtB-crowd-source the data. If anybody would prefer just jumping to the 2010 projection now, let’s hear it.

4. My secret dream for this is a flash-based/java-based/php/w/e tool that anybody could access and look up whoever they’d like whenever they choose. Some visual compromise would be required but the possibilities are endless. If anybody out there has that ability and would like to work on something together, contact me and let’s see what we can work out.

See Data Differently.
beyondtheboxscore.com | Twitter: @ justinbopp

by Justin Bopp on Jan 8, 2010 12:41 PM EST up reply actions