I suspect all women basketball programs use (more than one) national and/or local rating agencies. No program has the manpower to visit all the high schools and tournaments to filter the top prospects, and I’m sure they don’t want to rely on alumni recommending their nieces, no matter how well meaning. The rating agencies fill a great need and are generally very accurate. Just look at the high school ratings of the WNBA draftees if you doubt the accuracy.
HOWEVER, that is not to say there are not flaws and problems! I’d like to ask your help with one of the problems. I collect the rating agencies’ data. Most of the time, the average of the ratings seems like the best indicator (until I find that one or more which are better). Sometimes, however, there are outliers—wild discrepancies among the rating services.
First, let’s play a game. I list 8 examples. How would you adjust (if at all) the data below for each of the women to better reflect an accurate rating of the player? Also, there are some questions below for help.
When ESPN HoopGurlz does not rank a player in the top 100, I give a ranking based on the grade. 91=101, 90=111, 89=121, etc. up to 151, which is the highest ranking any player gets.
From the 2020 recruiting class:
• Aaliyah Edwards (Connecticut): 23, 151, 151, 11 (note: BS listed the rank as 300 and Prospects listed as 4 star but not in the top 150) [Unadjusted Avg = 84]
• Amirah Abdur-Rahim (Notre Dame): 121, 66, 151,112 [Unadjusted Avg = 112.5]
• Alasia Hayes (Notre Dame): 44, 41, 130, 32, 50 (by Dan Olson Collegiate Girls Basketball Report [CGBR]) [Unadjusted Avg = 59.4]
• Shelby Calhoun (Virginia Tech): 111, 50, 57, 86, 15 (CGBR) [Unadjusted Avg = 63.75]
• Kennedy Todd-Williams (North Carolina): 111, 33, 56, 33, 14 (CGBR) [Unadjusted Avg = 49.4]
• Treasure Thompson (LSU): 111, 129, 81, 91, 64 (Sports Madness), 49 (CGBR) [Unadjusted Avg = 87.5]
• Katelyn Levings (Arizona State): 111, 43, 54, 92 [Unadjusted Avg = 75]
• Gabby Hutcherson (Ohio State): 111, 30, 12, 62 [Unadjusted Avg = 53.75]
So the questions become, should I trim outliers, as Dillon has suggested. Should I trim both high and low outliers? I tend to trim just high outliers. Should I leave mild outliers as they may indicate a better trend. Should I flag each outlier and use my personal judgment on what to do? Should I do nothing, i.e. (leave the data as is)?
I realize some of you don’t like the rating systems at all. That’s fine. (“She’s a player, not a number!”) But I’d rather save that discussion for another time. If you’d rather not play along, that’s fine.