That is one sweet graphic

I think something like this, which needs very little explanation to be digested by the end-user, is an excellent direction for sabermetrics.

And there he is!

This stuff is gold. Not only does it give us a quick way to compare players, it makes it easier for the non-baseball folk to begin to understand the differences in what makes each player valuable (or not) to the team.

Can you make on of Adam Dunn and no deflect his power the -27%?. Just keep it a straight line. Wondered what that one would look like.

The other problem would be that the area shaded is not the % they are at. Look at Billy Butler’s OBP of 49%, but only ~25% of his area is colored. Can you redo his where the point is at .7 (square root of .5). If you put it at that point about half of that triangle would be filled in. Email me if you have any questions.

Are busy arguing about whether or not string theory (or one of its many variants) is the answer to the search for the unified theory of everything.

This is pretty cool though. I hope you guys include some of the feedback from the comments to make this even better.

Very interesting stuff with a lot of potential though, I like it.

But I agree with the posters above that some tweaks could probably make it even better.

And props to me for remembering this type of chart and getting all excited about it in Tommy’s DBS piece (though I claim no credit whatsoever).

I think using a distribution of all players for defense makes sense. That prevents the ridiculous negative percentiles (does it? I may not have read the explanation as carefully as I should have). Also, I would like to see fielding done with positional adjustments, so that we can glean even more information from it at a glance. Then again, I’m sure it would make people mad to see Pujols as below average since he’s playing at first base.

You could zero out the lowest negative number to to have the scale make some sort of intuitive sense

Well rounded

Awesome stuff, guys!

Catcher defense is probably the most confounding issue I’ve encountered to date (shocking, I know). When they translate into the database, they come across just as a blank … which, of course, Excel translates into a zero, which is actually pretty close to league average, since we’re using a metric that measures runs above and below that. Same for EqBRR.

A lot of the problems with fielding going off the chart (-27% for Dunn, 117% for Zobrist) is their inability to qualify at any one position for UZR/150. We wanted to use that particular metric to help downscore some predominately DH / immobile types, but I’m not sure at this point if it’s working, and am tinkering with some other options. We also have 3 rate statistics (OBP, ISO, UZR/150) and 1 counting stat (EqBRR) which works pretty well this late in the season, but is annoying me.

Also, nothing on the grid is normalized for league, park, or position. After the full database is complete, I will start the positional translations (oh please, someone give me some ideas for catchers) to see how that will work. I believe Justin and I are both opposed to weighting the graph itself (avoiding getting something that looks like this: http://www-personal.umich.edu/~mejn/election/2008/statepopredblue1024.png), but I’d be a lot more interested in a scoring system that can evaluate the grid and score him in a way that would evaluate his best use (if any at the major league level): For example, a high middle-infield defense score and speed rating would rate as a Pinch Runner / Defensive Replacement, and where they would rank vs. other players of their own type – Top Tier, Middle Tier, Lower Tier. Simple, but potentially very effective I think. For example, I think that Adam Dunn would hopefully profile and auto-score as an Upper Tier Designated Hitter, and then perhaps in what percentile for that position. Perhaps even positional suggestions? A player with a great bat struggling at a position too far on the left of the defensive spectrum could profile as a Middle Tier Second Baseman vs. say a Lower Tier Shortstop.

Thanks for the advice, comments, and expertise. We’ll keep tinkering with this over the next couple of weeks to incorporate a lot of these ideas and see if we can come up with something cool.

Have each quadrant always be shaped like a kite, intersecting the borders of its region at the 25th percentile no matter what, and the tip of the kite extending towards 100% (or back towards 0% and beyond). This would be more like a rotating bar graph, but still look a little cooler. Not sure it’d look as cool as the spider plot, but maybe worth a try.