- FanGraphs Baseball - http://www.fangraphs.com/blogs -

Five-Tool Players by the (Nerdiest Possible) Numbers

Posted By Carson Cistulli On August 18, 2011 @ 9:00 am In Daily Graphings,Research | 58 Comments

If there’s still such a thing as newstands anymore, the issue of Baseball America at your local one (i.e. your local newstand) is that publication’s annual “Tools” edition. No, it’s not (as you might suspect from the title) an issue dedicated entirely to relief pitchers with questionable taste in facial hair. Rather, it’s in this edition of the magazine that the editors of Baseball America attempt to isolate the players — major- and minor-leaguers — with the best baseballing tools (hitting for average, hitting for power, speed, etc.).

Beyond the results of a survey to which each of the league’s 30 managers responded, the issue also includes an attempt by author Matt Eddy to find five-tool players “by the numbers” (subscription required, I think).

After a brief discussion of what a “plus” tool might look like when quantified — and also some notes on the obvious limits of such an endeavor — Eddy suggsts this as a methodology:

For the sake of this exercise, let’s identify an above-average hitter as one who bats at least .285/.360/.460 with an isolated power of .175. That’s a 110 percent bump across the board (and then rounded down slightly to please the eye).

To this, Eddy also adds a speed component (more than 20 stolen bases) and runs his criteria through Baseball Reference’s Play Index for all player seasons 2000-10 — the results of which you can find here. (Note: it appears as though Eddy’s power criteria in that search is actually 20-plus homers and not a floor of a .175 ISO, but the results come out similarly.)

The big winner using this methodology is Bobby Abreu, who meets all of Eddy’s criteria in seven of 11 possible seasons. Hanley Ramirez qualifies in four seasons, while Alex Rodriguez finishes third with three “five-tool” seasons.

Eddy’s experiment is an interesting one, both in and of itself, and also for its potential to be nerd-ified — which, this being FanGraphs, that’s what I’ve endeavored to do in what follows.

For every player in the FanGraphs Era (2002-10), I’ve attempted to build on Eddy’s methodology using some numbers that might be more amenable to a FanGraphs reader’s tastes. Defense, as Eddy notes, is tougher to quantify, but I’ve submitted a way to deal with it that’s at least somewhat satisfying.

Here are the criteria for each of the five tools in this particular exercise:

Tool: Hit for Average
Stat: Contact Rate (Contact%)
Notes: I also considered strikeout rate (K%) for this, but felt that, if the goal is really to isolate a hitter’s contact ability, then contact rate is (n’doy) the best way to do that. While contact rate is a different thing than hitting for average, it also (i.e. contact rate) becomes reliable in a much smaller sample. In fact, batting average doesn’t even become a reliable marker of skill within 750 plate appearances, making a single season’s worth of data something less than entirely meaningful.

Tool: Hit for Power
Stat: Home Runs per Batted Ball (HR/(PA-K-BB-HBP))
Notes: For power, I considered at least two other metrics — namely, isolated power (ISO) and home runs per fly ball (HR/FB). The problem with ISO is that it gives credit to speed, because faster players are more likely to turn singles into doubles and doubles into triples. That’s fine, but if we’re trying to isolate speed separately, we don’t want to credit it here, as well. HR/FB, as I say, was another possibility, but the thing that it ignores is a player’s proclivity for hitting fly balls in the first place. Part of the mechanics of hitting a home run is the ability to create loft. Home runs per batted-ball rewards this ability.

Tool: Speed
Stat: Speed Score (Spd)
Notes: There’s no totally great way to do adjudge speed. I considered finding runs added via stolen bases (using linear weights) and general baserunning runs together, but there are plenty of players who are smart baserunners but who aren’t necessarily fast. If the goal is to isolate the speed tool, Speed score is the best (if not ideal-est) way to do that.

Tool: Fielding (and Arm)
Stat: UZR + Positional Adjustment
Notes: If I’ve had one intelligent thought in the past week (or, granted, maybe longer), it was to include not only UZR in the fielding criteria but also to add in each player’s respective positional adjustment. One of the draws of the five-tool player is that he’s theoretically able to play a position somewhere on the right side of the defensive spectrum. In other words, a player with a +5.0 true-talent UZR at shortstop is very different than one with a +5.0 true-talent UZR at first base — about 20 runs different, in fact. Adding in the positional adjustment is also akin to adding a sort of regression to the defensive numbers, thus giving less overall weight to UZR, which needs some three seasons to become reliable, alone. The addition of positional adjustment also goes some way to crediting both fielding ability and arm together. Like the above categories, it’s not necessarily ideal, but it works well for the purposes of the exercise, I think. (Note: so’s not to credit players who appeared in more games, I divided the sum of UZR and positional adjustment by games played).

To decide what constituted a “plus” tool, I calculated the above stats for each of the 1397 qualified players between 2002 and ’10. From there, I found the z-score (standard deviations from the mean) for each player in each respective category. The goal was to produce a number of player seasons roughly equivalent to the 40 or so Eddy found (or slightly fewer, because Eddy’s study included two more years.)

To do this, I went through each of the first three categories and chopped off all the player seasons with z-scores below 0.10 in the relevant category. This process revealed immediately how difficult it is find players who possess all the skills simultaneously.

To wit: eliminating just the players with a contact or power z-score below 0.10 leaves only 125 player seasons — or roughly 9% of the original sample. Cutting speed z-scores below 0.10 leaves only 36 players. All 36 of these players are included below.

I’ve sorted the qualifying players by the defensive component so that the reader can draw his own conclusions about the players who finished below the 0.10 threshold. The 20 players above the Xs qualify as five-tool players across the board, including defense; the other 16 players finished below the threshold.

Year Name Team Con Pow Spd Fld+
2008 Chase Utley Phillies 0.48 0.76 0.95 1.83
2007 Chase Utley Phillies 0.55 0.17 0.95 1.49
2010 Troy Tulowitzki Rockies 0.76 0.89 0.83 1.41
2008 Carlos Beltran Mets 0.81 0.29 1.25 1.26
2006 Carlos Beltran Mets 0.45 2.05 0.89 1.21
2009 Chase Utley Phillies 0.46 0.85 1.37 1.20
2009 Ian Kinsler Rangers 1.05 0.68 1.67 1.18
2006 Eric Byrnes Diamondbacks 0.50 0.37 1.25 1.04
2008 Grady Sizemore Indians 0.22 0.79 1.31 0.95
2007 Jimmy Rollins Phillies 0.91 0.10 2.68 0.89
2006 Chase Utley Phillies 0.32 0.61 0.95 0.87
2009 Troy Tulowitzki Rockies 0.39 1.08 1.43 0.87
2006 Vernon Wells Blue Jays 0.32 0.61 0.59 0.83
2005 Felipe Lopez Reds 0.48 0.13 0.77 0.80
2007 David Wright Mets 0.20 0.61 0.53 0.79
2008 David Wright Mets 0.31 0.72 0.24 0.69
2009 Nate McLouth - – - 0.43 0.14 0.65 0.37
2004 Carlos Beltran - – - 0.12 1.16 2.38 0.33
2003 Carlos Lee White Sox 0.60 0.50 0.42 0.31
2006 Ray Durham Giants 1.12 0.55 0.71 0.26
x x x x x x x
2005 Vladimir Guerrero Angels 0.29 0.86 0.47 -0.05
2010 Aubrey Huff Giants 0.26 0.35 0.59 -0.11
2004 Tony Batista Expos 0.48 0.55 0.36 -0.15
2003 Gary Sheffield Braves 0.34 1.12 0.59 -0.17
2006 David Wright Mets 0.45 0.38 0.95 -0.22
2009 Grady Sizemore Indians 0.26 0.28 1.07 -0.46
2007 Ian Kinsler Rangers 0.62 0.15 1.37 -0.55
2005 Albert Pujols Cardinals 0.64 1.26 0.53 -0.67
2002 Raul Ibanez Royals 0.24 0.45 0.18 -0.68
2008 Nate McLouth Pirates 1.17 0.23 1.43 -0.74
2009 Johnny Damon Yankees 0.48 0.31 0.95 -0.83
2002 Brian Giles Pirates 0.91 1.70 0.30 -0.96
2007 Gary Sheffield Tigers 0.13 0.52 0.89 -1.03
2007 Hanley Ramirez Marlins 0.50 0.30 1.84 -1.04
2004 Bobby Abreu Phillies 0.38 0.76 0.83 -1.56
2006 Carlos Lee - – - 0.83 0.78 0.30 -1.60

The results pass the sniff test, I think. By this methodology, four players have recorded multiple “five-tool” seasons since 2002: Chase Utley (four), Carlos Beltran (three), Troy Tulowitzki (two), and David Wright (two, with a third below the Xs). It’s possible that Utley’s arm might preclude him from being a true “five-tool” player, but I’m leaving him here because (a) I know literally nothing about his arm strength, and (b) he’s awesome.

Some other notes on the results:

• The average WAR for the 20 qualifying five-tool players is 6.3.

• Bobby Abreu, who dominated Eddy’s list, appears here only once — and even then he doesn’t qualify defensively. He qualifies in the speed category in every season but one, but power (the way it’s defined here, at least) holds him back: he’s only posted two seasons with a power z-score greater than 0.10 since 2002.

Carlos Lee was fast once. Maybe. Anyway, between 2003 and ’06 he averaged 15 stolen bases and just four caught stealings.

Aubrey Huff probably wasn’t ever fast, but he’s managed five triples in two different seasons (2007, ’10).

Tony Batista slashed just .241/.272/.455 (.225 BABIP) in his quasi-five-tool season, finishing with just a -0.4 WAR.


Article printed from FanGraphs Baseball: http://www.fangraphs.com/blogs

URL to article: http://www.fangraphs.com/blogs/finding-five-tool-players-using-advanced-stats/

Copyright © 2009 FanGraphs Baseball. All rights reserved.