In partnership with CBSSports.com
Online Now 203
Online now 120 Record: 2724 (1/16/2014)
You have no favorite boards.
The most viewed topics.
The most replied to topics.
The most up-voted topics.
The most down-voted topics.
The most up-voted posters.
The most down-voted posters.
The most followed posters.
The team rankings formula was a hot topic of discussion around signing day because 247Sports had Vandy rated so much lower than other sites. I thought you might want to know that we've made some adjustments to the formula.
One thing we take pride in at 247 is being self-evaluators in making sure we continue to provide the best product for the user. We found a couple of faults in our team ranking formula, one of which was the case study of Vanderbilt - a class with a lot of high three-stars but no five-star headliners. We realized that our formula unfairly penalized a school like Vanderbilt.
We also felt like the formula unfairly penalized a class like USC's and Stanford's that was extremely small but had very talented prospects all the way through.
Finally, the formula was very innacurate throughout most of the early stages of the recruiting cycle and didn't do an effective job of accurately ranking schools at a time when most programs have less than 10 commits.
After re-evaluating some of those issues and looking for solutions, the formula has been tweaked some and the new formula has been applied retroactively to the class of 2013. So Vanderbilt's new 247Composite team ranking moves from 30 to 26. That may not completely satisfy you but I wanted to make the point that your concerns didn't fall on def ears and moving forward the new formula will make classes with the makeup of Vanderbilt's last year - as well as the makeup of Vanderbilt's smaller class this year - less of outliers.
In order to create the most comprehensive Team Recruiting Ranking without any notion of bias, 247Sports Team Recruiting Ranking is solely based on the 247Composite Rating.
This post was edited by Barton S 13 months ago
Barton Simmons | 247Sports | National Recruiting Analyst | Twitter: @bartonsimmons
This is good to hear.
Electi a Deo. Pugnans dæmonium. Domine fortitudo mea, non timebo mala.
Thanks, Barton. I'm satisfied and appreciative. Speaks tons about 247's integrity and determination to get it as right as is humanly possible.
Good deal, Barton. As moto said, the integrity of 247 Sports--whether it's in evaluating their own processes, or ensuring we have a clean and troll-free environment--is a true differentiator.
Agreed, it's not so much the small rankings jump that we received, but more the effort that was put forth to fix a problem and make a fairer, more accurate system. Thanks.
Thanks for explaining this and it's good to know our argument didn't fall on deaf ears.
follow me on twitter @country_cookin
Awesome! I'm glad our thorough analysis of the ranking formula helped prompt some improvements. Hope this quiets some of the conspiracy theories.
I'm curious how it was tweaked. Did you increase the standard deviation of the Gaussian?
I don't want to get too much into the nuts and bolts because i'm not a mathmetician and I'll misspeak but essentially we did two things:
We widened the curve by increasing the soft cap on classes but we also flattened the curve by subtracting 70 from each players rating. So a player rated as a 72 now counts for 2 points instead of 72 points. A 90-rated player is 20 points not 90, etc.
it's good to see professional efforts following collegiate sports! thanks for the update and explanation, much appreciated!
I don't understand the subtraction of 70. In your example, the higher rated player now has 10 times the points of the lesser rated player, whereas before it was just a matter of degree. Isn't that a huge benefit to the higher ranked player? I ain't no math major either!!
Yeah, moto.....yeah. I saw that too.
Remember that 247's philosophy is that higher-ranked players should be more strongly emphasized in the class rankings than lower-ranked players, because they have a greater chance of having a greater impact. That's why they originally set the the soft cap at about 16 for the Gaussian distribution--so that the lower ranked players have significantly less value.
But the problem with that was that they were de-valuing large classes of consistently-graded players, like Vanderbilt's class of 27 with a ton of guys in the 84-88 range--and to make matters worse, 92% of signing classes had more than 16 players, so therefore 92% of the signing classes were potentially being rated disproportionately.
So the solution to fix that problem is to extend the soft cap out--I think the discussion was to move it to around 23 or 24. But when you do that, you now decrease the effect of "higher-rated players should be weighted more heavily"--basically, you're stretching out the weighting over 23 or 24 points rather than 16 points, and the weighting between player 1 and player 10 has decreased. And this is counter to 247's mentality that the higher-rated players should be weighted more heavily.
So, their solution to that problem is to scale the scores non-linearly to restore the heavier weighting of the higher ranked players. Now, a player rated 90 is scaled to 20, and a player rated 100 is scaled to 30. The player rated a 90 now has 67% the value of the player rated 100, whereas it was 90% the value before.
It's actually a very clever way for 247 to stick to their guns--which I applaud them for--while still accounting for the class-size loophole in their original algorithm.
This post has been edited 2 times, most recently by quadradore 13 months ago
Thanks, Quad -- U be smart!!
Ha--let's stick with "educated"....you take enough stats classes, it starts to stick eventually
This post was edited by quadradore 13 months ago
247Sports In partnership with CBS Sports