Friday, September 2, 2016

Binary Components on the 20-80 Scale

Kiley McDaniel explains the 20-80 scouting scale in this article for fangraphs:

"The invention of the scale is credited to Branch Rickey and whether he intended it or not, it mirrors various scientific scales. 50 is major league average, then each 10 point increment represents a standard deviation better or worse than average. In a normal distribution, three standard deviations in either direction should include 99.7% of your sample, so that’s why the scale is 20 to 80 rather than 0 and 100."

Here are the major league averages for each binary component, 2006-2015:

$BB $SO $HR $H $XBH $3B $SA $SB
 9% 21%  4% 30% 25%  9%  9% 73%

Some brief refresher definitions, if you didn't read the binary components post:

$BB - percentage of all plate appearances that ended in a walk or a HBP
$SO - pct. of all PA ending with a thrown strike (hit, missed or called) that were strikeouts
$HR - pct. of all balls hit fair that resulted in a homerun
$H - pct. of all non-HR balls hit fair that resulted in hits - aka BAbip (batting average on balls in play)
$XBH - pct. of all non-HR hits that were doubles or triples
$3B - pct. of non-HR XBH that were triples
$SA - pct. of times on first (1B + BB + HBP) where the batter attempted to steal
$SB - stolen base success rate

Using the standard deviations from the hitter controlled/pitcher controlled study, I found scores for every batter with at least 1,000 PA, and every pitcher with at least 1,000 BF, from 2006 to 2015.

As Kiley explained, 50 is average, 60 is "plus", 70 is "plus plus", and 80 is very rare.

Here are the top ten batters in OPS (min. 1000 PA) , along with their component scores:

Rk Player OPS $BB $SO $HR $H $XBH $3B $SA $SB
1  Miguel Cabrera  .980  62  54  67  77  54  38  42  42
2  Manny Ramirez  .962  74  48  68  73  55  40  40  29
3  Joey Votto .957  77  47  63  81  59  42  46  45
4  Mike Trout .956  67  40  70  81  61  65  59  66
5  Albert Pujols  .946  63  66  68  44  57  36  46  52
6  David Ortiz .940  68  52  72  50  71  38  41  31
7  Paul Goldschmidt .930  68  41  66  79  67  42  53  59
8  Jim Thome .922  77  32  84  57  56  37  40  19
9  Lance Berkman  .915  74  50  67  56  55  44  46  42
10 Chipper Jones  .913  68  58  59  63  54  43  43  66
   Average 70  49  68  66  59  42  46  45

The typical elite hitter has plus-plus patience, power and BAbip and a plus $XBH rate, but he makes just average contact and has below-average rates of $3B, $SA, and $SB.

There are exceptions - Pujols has plus contact and below average BAbip. Ortiz has a plus-plus $XBH but an average BAbip. Jim Thome is around 80 in $BB and $HR (patience and power), but is average or below average at everything else. Goldschmidt is an above-average base-stealer, and Chipper had a plus $SB. And then there's Mike Trout - plus or better at everything except for his $SO rate, which is below average (or it was, before he corrected it along with his lagging $SA in 2016).

Scouts love five-tool players (hitting, power, running, fielding, and throwing) on the traditional 20-80 scale. Are there eight-tool hitters??

Turns out, two players scored 50 or better in every single component:

Player $BB $SO $HR $H $XBH $3B $SA $SB Avg Tools
Andrew McCutchen  65  51  56  70  57  58  59  52  8
Ryan Braun 50  50  66  70  59  54  59  58  8

The 2011 and 2013 NL MVPs. But Braun is exactly average (50) at the first two components, $BB and $SO.

Players with the most "plus" (60 or higher) tools:

Player $BB $SO $HR $H $XBH $3B $SA $SB Plus Tools
Mike Trout  67  40  70  81  61  65  59  66 6
A.J. Pollock  45  58  48  65  61  62  65  62 5
Kenny Lofton  51  69  38  60  40  87  65  63 5
Ben Revere  35  68  33  63  17  85  74  61 5
Ichiro Suzuki 39  66  38  68  22  67  63  66 5

Trout just missed the cutoff in $SA, and with his increased emphasis on base-stealing in 2016, he would no-doubt be over 60 in that category as well if 2016 stats were included, making him a legit "seven-tool" hitter.

Lofton, Revere, and Ichiro are all cut from the same cloth - below average or poor in $BB, $HR, and $XBH (although the criminally-underrated Lofton had an average $BB), plus or better at everything else. Essentially these guys make contact, leg out base hits, turn their relatively few extra-base hits into triples, and steal bases.

Players with the most "plus plus" (70 or higher) tools:

Player $BB $SO $HR $H $XBH $3B $SA $SB Plus Plus Tools
Starling Marte 46  42 51  77 51  70 78  51 3
Dee Gordon 36  57 35  74 29  80 95  55 3
Juan Pierre 41  74 33  50 25  76 82  55 3
Jack Cust 81  19 73  71 46  38 41  25 3

These players all excel at a few components but are average or worse at the rest. Marte, Gordon, and Pierre are all burners with elite rates of base hits ($H), triples ($3B), and stolen base attempts ($SA). Jack Cust had extremely good patience, power, and BAbip, but just couldn't make enough contact to be a good major league hitter.

Finally, three players scored 80 or better in more than one component:

Player $BB $SO $HR $H $XBH $3B $SA $SB 80 Tools
Dee Gordon 36  57 35  74 29  80 95  55 2
Russell Branyan 63  25 84  42 65  39 42  88 2
Jarrod Dyson 49  51 35  55 39  99 104  69 2

Russell the Muscle's power was legit, but his $SB score of 88 is not - he was just successful in all nine of his steal attempts.

I'll do pitcher scores in the next post.

2 comments:

  1. I guess I don't like the 8 outcomes being weighted evenly when they happen at very different rates. Triples are a rare occurrence so a player's ability and value don't factor in very much on their triples production. I like turning them into "tools" on the 80-20 scale as you've done but perhaps it would be better to try and split out parts based on value so that each "tool" would be of similar value.

    ReplyDelete
    Replies
    1. That's a good point. I like them on the 20-80 scale, too, because then 50 is always league average. I think making the lists of 5-plus-tool players, 3-plus-plus-tool, etc., like I did, was probably misleading because it treats all the components as having equal value, which of course they don't.

      Delete