Monday, November 5, 2012

The Guldan Rating, SB Nation

Patrick Guldan is the Managing Editor of SB Nation's "Columbus Crew community". After most Crew matches he posts a Notes and Grades summary of the previous game. In it he rates the Crew side of the match using the following system (in Mr. Guldan's words):
"Note on Grades: I have set 5 as an average performance. The player had both good and bad events on the field. Under 5 is a disappointing performance. Above 5 is a notable performance."

I've come to enjoy this feature over the past couple seasons as I like that he grounds player performance on average and works off that. A lot of player ratings out there (I see them most often with USMNT) use a 1-10 scale with no indication of average performance.

**I'm going to cut to the chase than explain lower down the post.**

What I did was take the number of times a player rated a 7 and above and divided that by total number of times rated.


55.6% : Federico Higuain : 9
46.2% : Jairo Arrieta : 13
35.3% : Danny O'Rourke : 17
33.3% : Andy Gruenebaum : 27
33.3% : Matt Lampson : 3
32.1% : Eddie Gaven : 28
30.0% : Milovan Mirosevic : 20
21.7% : Josh Williams : 23
21.4% : Dilly Duka : 14
20.0% : Kevan George : 5

17.9% : TEAM

16.7% : Chad Marshall : 18
16.7% : Chris Birchall : 12
15.4% : Sebastian Miranda : 26
12.5% : Cole Grossman : 8
11.1% : Justin Meram : 18
10.0% : Olman Vargas : 10
10.0% : Carlos Mendes : 10
9.1% : Emilio RenterĂ­a : 22
9.1% : Eric Gehrig : 11
9.1% : Nemanja Vukovic : 11
7.1% : Coaching : 28
5.6% : Tony Tchani : 18

0.0% : Bernardo Anor : 10
0.0% : Shaun Francis : 7
0.0% : Ethan Finlay : 11
0.0% : Kirk Urso : 5
0.0% : Aaron Schoenfeld : 8
0.0% : Julius James : 9
0.0% : Ben Speas : 1


Another thing I like about Guldan's system is that in the description it doesn't indicate if "10" is even the highest number or even if 1 is the lowest. Up until recently I'd never seen even an "8" given out by Guldan (I talked about that earlier this year).

29 total Crew players were rated 402 times (on average, about 14x a player). Plenty of times to draw valuable information out. Here's how the score distribution looked. On the left is his rating (LOW is bad, like - a 2; High is good, 9): On the right is the number of times that Score was handed out.

2 : 2
3 : 18
4 : 50
5 : 99
6 : 161
7 : 64
8 : 5
9 : 3

No "1" or "10" this year. Again, Guldan's system pivots off a 5 as average rating. What this reminds me of is a Net Promoter Score (NPS). It's a system that simply asks the question "Would you recommend this product to a friend". In the Net Promoter system average scores are thrown out. Which is exactly what I did. The reason being is to eliminate the 'passive' ratings. In the retail / maufacturing world 'passive' customers can be seen as the enemy. Why? I doesn't give you any actionable items. Without actionable tasks you waste a lot of time on nada.

If you have a customer nailing your ass to the wall... you want to focus on that. NOT the folks in the middle, NPS treats that as noise. Same goes for the higher ratings. Growing your company, you want to figure out what works. NPS can be valuable in this way.

What I did to derive a usable number from Guldan's rating system is take the total number of times a player scores a 7 and above and divide that by the total number of times the player was rated. I feel that shows best what Patrick Guldan views as quality.

Duka was rated a total of 14 times by Patrick Guldan. He (Duka) got rated 4 twice, a 5 four times, a 6 five times and a 7 three times. It looks like this:

4 : 2
5 : 4
6 : 5
7 : 3

Dilly didn't recieve any ratings others than those above. What I'm going to do is throw out the anything below 7... take the 3x at a "7" and divide it by total number of times rated, 14.

3 / 14 = 21.4% : Duka

Taking out the "detractors" pulls my number away from true NPS. I did this, in part, because Guldan's rating isn't on a 1-10 scale. Also, NPS has many critics, including me.

Taking the "average" rating of players leaves almost all at around 5 or 6. Removing the middle ground gives me more to go off of. Had a brief Twitter conversation with the great Steve Sirk (@stevesirk) about how the Crew players with the most time naturally find the middle of the road (Team averages). A NPS type score is a simple way of discounting average and bringing important items into focus.

Of course Guldan's rating system isn't the only thing to go off of, nor is only using "NPS" if you are a retail company. Guldan's ratings are an "outsider but almost insider" type rating that should only be taken as such. Meaning, he has contact with team officials and players but isn't part of the payroll. There are hints of favoritism here and there but generally his scores are honest. I'll post this on the right side blog though so you, dear informed reader, can make your own conclusions about things.

If you want to mess with Patrick Guldan's player ratings you can find them HERE. Mr. Guldan didn't rate every week (that I could find) but I pulled as much as there was available on his SB Nation Crew blog. Statistically; it's enough to evaluate how he views the performance of the players this year.

I'm posting it as part of my ongoing evaluation of Crew players.

I appreciate Patrick Guldan's player ratings and look forward to them again next year.

No comments: