247 creates composite football recruiting rankings (across services)

Submitted by turd ferguson on

I think this is pretty cool.  247Sports just put together a composite recruiting ranking that averages across the four major recruiting services. 

http://247sports.com/Season/2013-Football/CompositeRankings?Institution…

Here's where the Michigan recruits landed in 247's calculation of the four-site average:

16. Shane Morris
55. Dymonte Thomas
62. Kyle Bosch
74. Patrick Kugler
82. Chris Fox
114. Henry Poggi
116. Mike McCray
122. David Dawson
133. Logan Tuley-Tillman
137. Taco Charlton
146. Jourdan Lewis
153. Jake Butt
191. Ben Gedeon
203. DeVeon Smith
215. Gareon Conley
242. Wyatt Shallman
301. Maurice Hurst, Jr.
311. Jaron Dukes
355. Ross Douglas
541. Csont'e York
641. Khalid Hill

Also of note:

40. Leon McQuay
41. Laquon Treadwell
60. Derrick Green

I don't see Channing Stribling or Scott Sypniewski in the top 1250.

Michigan has 16 guys in the top 250.  By my count, OSU currently has 10, Notre Dame currently has 10, and MSU doesn't have anyone.  Keep in mind that their classes aren't as large as ours right now.


---------------------------------------------


For those who care, the details are below (and a little unclear).  I've tried this myself in a couple of diaries, and there are a lot of decisions that you have to make as you're doing it. 

 

247Composite Rating

The 247Composite Rating is a proprietary algorithm that compiles prospect "rankings" and "ratings" listed in the public domain by the major media recruiting services. It converts average industry ranks and ratings into a linear composite index capping at 1.0000, which indicates a concensus No. 1 prospect across all services.

The 247Composite Rating is the industry's most comprehensive and unbiased prospect ranking and is also used to generate 247Sports Team Recruiting Rankings.

More details

All major media services share an equal percentage in the 247Composite Rating.

The composite index equally weights this percentage among the services that participate in a ranking for that specific prospect.

A composite strength meter, indicated by red bars, illustrates the total number of industry services that have ranked the prospect. A full strength meter indicates the prospect has been ranked by all industry services participating in the composite.

All industry services have a different philosophy on number of "stars" distributed with each class. The 247Composite Rating assigns stars based on an approximate average distribution of stars from the industry.

markinmsp

July 30th, 2012 at 2:09 PM ^

 Gatorcountry (a UF fansite) has something similar that may interest you for comparison, Turd. They created a database with a ranking of each individual 2013 football recruit in each of the four major recruiting services: ESPNU, Scout, 247Sports and Rivals. They then come up with a composite score for each of the football recruits. Also they explain their methodology. Basically, it's an average of the four rankings with a slight scaled multiplier applied for players who are only ranked in one or two of the four services. (So again it uses the number a recruit is listed rather than some score they may attain in each service. Last Updated: June 19, 2012)  -    

   Composite Recruiting Rankings for 2013 football recruits

turd ferguson

July 30th, 2012 at 10:45 AM ^

For what it's worth, I think offer lists are informative but impossible to use for rankings.  The reasons:

  • Offers are generally self-reported, so if a recruit doesn't report some offers, misinterprets an interaction with a coach, or makes something up, it messes with the data.
  • Kids who commit early often don't have nearly as many offers (or as many good offers) as they would if they committed later.  For example, I'm sure that Shane Morris and Dymonte Thomas would have many more elite offers if they were still available now.
  • There are regional biases.  An elite recruit in Seattle will tend to get West Coast offers, since many Eastern, Southern, or Midwestern schools either won't know much about him or will figure that they won't have much chance of getting him.
  • It's kind of hard to know what to do with those lists even if have good data.  For example, look at Ohio State this year.  An OSU offer to a CB means a lot, because they've done really well with CB recruiting this season and don't have many spots.  An OSU offer to an offensive lineman doesn't mean nearly as much, because they've had trouble with o-line recruiting and have had to reach on some offers.
  • Other stuff

The details of what 247 did aren't clear, but I think this kind of thing is the best way to rate prospects.

Jon06

July 30th, 2012 at 11:59 AM ^

you provide a lot of reasons you can't use final offer sheets, but there's no reason you can't rank kids relative to local offers at the time of commitment--that would allow you to rank guys like morris relative to others in his area--and then use those rankings to construct informative rankings based on final offer sheets, since plenty of guys won't commit till the last minute. you'd also have to weight offers relative to a school's recruiting success at various positions at various points in time. this would be a pain in the ass, but i imagine the mathletes of the recruiting world would have no problem coming up with good rankings doing it. the biggest problem i can see is that you'd need an initial ranking of offers by program/position pair to do it, and all kinds of bias would creep in there. but that's nothing new for rankings in the CFB world.

y2mh

July 30th, 2012 at 3:39 PM ^

Have to agree with OP. Number of offers are affected by too many extraneous variables to be of any interest.

To add to it, are all offers created equal? Which offers do you count,… only FCS,… only AQ conferences,… top 50 teams? Do 10 offers from MAC and 4 from SunBelt schools outweigh 2-3 for a recruit that commits to LSU @ age 15?

Hands Free

July 30th, 2012 at 12:18 PM ^

Another variable to consider - which may skew the offer list data - is how open a player's recruitment may be?  A reason why Shane doesn't have as many big-time offers is likely due to the facts that he was committed early and is U-M all-the-way, so teams that may otherwise have offered him decided there was no point to do so.

MaizeAndBlueWahoo

July 30th, 2012 at 11:11 AM ^

I made some attempt at doing that a little while ago, actually, and it doesn't work very well.  Early commits have fewer offers than they should.  And now with actual offers not allowed to go out til August 1, the data is too fuzzy.  You'd need an unbiased list of offers, they're usually different across sites, and then you get a lot of times where a kid claims an offer and doesn't really have one.

NFG

July 30th, 2012 at 10:36 AM ^

 "and MSU doesn't have anyone."

 

I LoL'd. I don't look forward to the meltdown in the future when our class is only 11th in the country. How spoiled we are.

JCV16

July 30th, 2012 at 10:42 AM ^

To give more weight to services that have been shown to be better. Could use mgoblog's post from a while back, which I think showed scout best and rivals second

flysociety3

July 30th, 2012 at 10:52 AM ^

A little OT.... but, did anyone else see the 247 Tweet this morning that referred to Derrick Green's Committment to Michigan?

At around 715, 247 Tweeted that 'Michigan received its 24th committment from a Top247 Running Back,' and it linked to a story....

When I got to work about an hour later, the Tweet was deleted....

Think this means Green will commit later today, or that 247 is stupid?

Jmilan

July 30th, 2012 at 10:56 AM ^

From what I read it was just a mistake by 247. They made up the article in anticiapation of him committing at the BBQ, but somehow they accidently posted it. I wouldn't look too much into it. I have said this before and I will say it again Twitter will be the death of following recruiting closely if it ever happens.

UMgradMSUdad

July 30th, 2012 at 11:04 AM ^

Wow.  Looking through this list, it makes Michigan's and Ohio's classes stand apart even more from the rest of the B1G.  I counted 4 non UM or OSU B1G players in the top 250.  One each for the Illini and Wisc, and 2 for Neb.  Meanwhile UM alone has 16 in the top 250.

Edit: on a second look, there are 4 PSU and one Northwestern recruits in the top 250 as well.

Ball Hawk

July 30th, 2012 at 11:06 AM ^

Awesome work dude! Im surprised Dawson isnt at least in the top 100. Im surprised Shane isnt in the top 10. Anyways this is an awesome class!

RakeFight

July 30th, 2012 at 11:23 AM ^

I know middle of the road and lower TEs don't tend to get great rankings, but I'm surprised Hill doesn't even crack the top 600 with the pretty positive 7 and 7 camp reviews I see on him occasionally.

Leaders And Best

July 30th, 2012 at 11:24 AM ^

Because something seems off by them or they are still working out the bugs. It would be nice to see the detailed rankings of each service they are using.

For instance:

Ross Douglas is ranked the Composite #355 overall prospect (#28 CB) while Darian Hicks is ranked the Composite #256 overall prospect (#24 CB).  Rivals, Scout, ESPN, & 247Sports all have Ross Douglas ranked above Hicks in their indvidual rankings so it is unclear how Douglas is lower in the Composite.

turd ferguson

July 30th, 2012 at 11:37 AM ^

Wow, nice find on that.  That's really messed up.  In fact, unless it's a one-time typo, it's pretty incriminating evidence that these rankings suck. 

Rivals:  Douglas #22 CB, Hicks #62 CB

Scout:  Douglas #42 CB, Hicks #54 CB

247: Douglas #25 CB, Hicks #26 CB

ESPN: Douglas #30 CB, Hicks #59 CB

247 says that these are the four services used and that they receive equal weight in the composite rankings.  It's mathematically impossible that Hicks would come out ahead of Douglas.  That's probably worth an email.

Leaders And Best

July 30th, 2012 at 11:47 AM ^

They only used inputted the data from each services top prospects lists(Rivals 250, ESPN 300, Scout 300) into their system. Because Hicks is not ranked in any those lists, he maintains only his 247Sports ranking (which turns out to be his highest). Douglas is ranked by a lot of those lists and that actually ends up bringing his ranking down because it is below his 247Sports ranking.

I don't think they used the position ranking lists at all.

Just a theory.

turd ferguson

July 30th, 2012 at 11:49 AM ^

Interesting idea.  If they did that, that's really stupid.

I'm not sure how they could rank so many guys with that method, though.  I think they ranked 2000+.

Also, I just emailed them to ask about it, and I'll post their response here if/when it comes.

woodfeld

July 30th, 2012 at 11:58 AM ^

I think that's exactly what they did.  That's why Hicks only has 1 red box.  They say in their description that the number of red boxes is how many sites were used in the ranking.  Since Hicks falls outside the top 250/300, etc of the other lists, they just use the one they have...theirs.

It also explains why Northwestern's center is ranked #138

turd ferguson

July 31st, 2012 at 2:54 PM ^

In case anyone is interested, 247 modified their super-secret ranking algorithm, which took care of some of the problems that we talked about here.  I emailed them about this yesterday but didn't get a response.

It's still possible that these rankings are stupid - again, hard to tell - but this seems like a significant improvement.  For example, Tyler Lancaster, who is ranked by 247 (at #159) but completely unranked by the other three services, slipped from #135 in the composite rankings to #715.  Oops.  Ross Douglas, who is rated higher than MSU's Darian Hicks on every service (as a CB) is now rated higher than him in the composite rankings too (Douglas #350; Hicks #537).

An improvement, no doubt.

Leaders And Best

July 30th, 2012 at 12:24 PM ^

Jake Butt  Rivals #5 TE (5.9 ranking, #118 overall), 247sports #10 TE (90 ranking, not ranked in top 247), ESPN #5 TE (83 ranking, #174 overall), Scout #6 TE (4 star, #137 overall)

Josh McNeil Rivals #10 TE (5.8 ranking, not ranked in top 250), 247sports #8 TE (92 ranking, #184 overall), ESPN #4 TE (83 ranking, #131 overall), Scout #7 TE (4 star, #151 overall)

247Sports Composite ranking

Josh McNeil #150 overall, #6 TE

Jake Butt #153 overall, #7 TE

Although they are both close as you would expect based on the raw data, I still think you would clearly expect Butt to be ranked higher than McNeil when you compare the 4 ratings. Butt's position average ranking and national average rank is better than McNeil's (unless they are also using some other ranking system besides the main 4).

BlueHills

July 30th, 2012 at 11:25 AM ^

Interesting and good work.

In the case of college athletes, though, you have to take your hat off to programs like Wisconsin, Iowa and MSU that build very good teams with only a few of the top recruits.

This tells me that some kids with a lot of potential in high school are simply under-coached, in the sense that they don't learn fundamentals as well, and therefore may be less highly ranked.

The coaches in these programs obviously know how to find raw talent, and then, how to teach these kids to perform at a high level. For these programs, recruiting rankings are less critical to success than many of us might like to admit (I know there are stats that point to top teams being built on top recruits, I'm talking about the exceptions).

I guess I'm saying that maybe we make too much of this ranking business, and I'm as guilty of being excited about getting highly rated recruits as anyone.

Genzilla

July 30th, 2012 at 11:26 AM ^

Of the top 13 players in Ohio in the composite rankings, 7 are going to UM and 4 are going to OSU.



UM has the top 3 players in Michigan and 6 of the top 12.



UM has the #4 and #6 players from Illinois and is the favorite to get the #2 player in the state (LT)



UM has the #4 player in PA



The composite rankings show just how well UM is getting the top midwest talent, especially in MI and OH.

Genzilla

July 30th, 2012 at 11:59 AM ^

QB: Shane Morris (#2 Pro Style QB)

RB: Deveon Smith (#14 RB) Wyatt Shallman (#1 FB)

-Potential RB: Derrick Green (#7 RB)

WR: Jaron Dukes (#40 WR) C'sonte York (#70 WR)

-Potential WR: Laquon Treadwell (#4 WR)

TE: Jake Butt (#7 TE) Khalid Hill (#28 TE)

T: Chris Fox (#3 OG) Logan Tuley-Tillman (#10 OT)

G: Kyle Bosch (#2 OG) David Dawson (#6 OG)

C: Patrick Kugler (#2 C)



DE: Taco Charlton (#5 WDE)

DT: Henry Poggi (#10 DT) Maurice Hurst Jr. (#24 DT)

WILL: Ben Gedeon (#9 ILB)

MIKE: Mike McCray (#7 ATH)

CB: Jourdan Lewis (#13 CB) Gareon Conley (#20 CB) Ross Douglas (#28 CB)

S: Dymonte Thomas (#7 S) Channing Stribling (#111 ATH)



LS: Scott Sypniewski (#3 LS)





I tried to put guys into the position their projected to play at UM (Gedeon, McCray, Stribling, Fox).  It was my guess that Stribling would move to S, possibly as a backup Boundary-CB/FS as an upperclassmen.

LSAClassOf2000

July 30th, 2012 at 1:12 PM ^

...and, of course, without knowing their algorithm, it is hard to tell whether or not I am onto anything here, but I took some of the data from our commits and correlated the composite rankings to positional rankings, and the correlation is actually negative (-0.63, to be precise), whereas the correlation between the old 247 rankings and the composite score is 0.93.