1. Welcome to Georgia Tech Swarm! JOIN US and be a part of the SWARM! GO JACKETS! THWg!

How 'Should' Tech Do in Recruiting Rank - Analysis

Discussion in 'Georgia Tech Football' started by BCJacket, Dec 1, 2018.

  1. BCJacket

    BCJacket Helluva Engineer

    Messages:
    223
    I keep seeing a debate on this board (and others) over Tech's recruiting rankings. One side argues that Tech should be able to recruit top talent and the staff and scheme were holding us back. The other side says that Tech recruiting is what it is and the scheme let us work within our limitations. Both sides selectively point to periods in history to support their point, but don't have much in the way of facts.

    I personally leaned towards CPJ's argument that programs largely determine recruiting ranking and individual coaches have a limited impact. But I recognized I had no data to support that- which irked me as a data driven nerd.

    So, I made a predictive model of recruiting rankings based on attributes of P5 schools to see what schools over perform or under perform in recruiting, and how GT does.

    I tried to come up with factors that I expected (partly based on the narratives on this board) would correlate with recruiting rankings. School size, Facilities, Academics, Money, fanbase etc. I found the best data or proxy data for these factors. I analyzed the data sets I collected using a Pearson correlation coefficient vs the actual 5 year average of recruiting rankings per Rivals.

    I built a model using each schools ranking for each factor, weighted by the strength of the correlation for that factor. (The model back-tested very strong against actual results.) Then, I compared the predicted rank per my model to the actual 5 year average of recruiting rank per Rivals' rankings.

    FWIW, I started this over a week ago, before CPJ announced he was leaving. But it feels particularly prescient in the debate over whether to get a great recruiter and conventional offense. Or focus on getting a HC with an innovative offense.

    Partial results below. Recognize, these are relative to the predicted rank for the individual school, not absolute. (eg Maryland recruits much better than expected for Maryland, but it's still mid-tier overall ranked#37 in actual 5 year recruiting vs #63 predicted.)

    Schools that recruited much better than predicted:
    Maryland 26 spots higher on actual rankings than predicted rankings
    NC State 23
    Florida State 12
    Ohio State 11
    Miami 11
    Oklahoma State 11
    Northwestern 10

    Schools that recruited much worse than predicted:
    Texas -10 spots lower on actual rankings than predicted rankings
    Washington -10
    Wisconsin -10
    Duke -11
    Wake Forest -13
    Colorado -17
    Iowa -18
    Boston College -26

    Others of interest:
    Clemson 8
    North Carolina 7
    Virginia Tech 4
    Georgia 2
    Georgia Tech 0
    Virginia -4

    A lot of these make some sense. Texas has every resource and advantage (except academics and conference) but they've been down as a program for a while. Thus their recruiting rank has under performed (Still top 20, just not the top 10 you'd expect from their resources.) Duke, BC, Wake have resources, but they don't put them towards Football. NC State is kind of the "football school" in NC these days. The other NC schools care relatively more about B-ball.

    I honestly expected GT would be an under performer. It's taken for granted on the board and in the media that we under perform in recruiting. But, counter to the narrative, GT recruited exactly as would be predicted based on our attributes. We don't under perform at all (or over perform).

    As for what this means for GT. First, I think we need to cool it on the narratives that we recruit badly and our fan base is crappy. GT has done about what you'd expect relative to most other schools. We're in a really tough neighborhood, but comparing ourselves to Clemson, uGag and Bama really isn't fair. Our staff, players, and fans are average, not awful. We should recognize that.

    Moving forward, control what we can control. We should strive to be the outlier. The things that appear to really move the needle are facilities and revenue. TStan is working hard on upgrading our facilities. Growing our fan base would help, attendance drives revenue. A head coach who was a great marketer for the program would help a lot in our market. We have a ton of local media, let's get them on our side as much as possible.

    I'd say I'm optimistic that Tech can over perform under our next coaching staff. But, we need to be realistic- over performing would probably only get us to 40-45th in recruiting rankings. I don't think a 'great recruiter' for a HC, all else equal, could make us competitive running a conventional system. Long term, if we want to recruit better, we need to improve our attributes.

    Further details of the methodology and results posted below.
     
  2. BCJacket

    BCJacket Helluva Engineer

    Messages:
    223
    The nittier grittier:

    I did not make this GT-centric in any way. I ran the numbers and tried not to look at the data until I was done, but to be fair, I am obviously biased. This is not peer-review quality research. Just a project for fun. So, the methodology, data, etc are what they are. There may well be mistakes in the data transcription and analysis. But the model results do correlate strongly with the real world observations. Take it for what it's worth.

    The data I analyzed:
    • Total athletics revenue 2017*; Moderate/strong positive correlation with recruiting ranking, PCC = .6722
    • School enrollment; weak/moderate positive correlation with recruiting ranking, PCC =0.2596
    • US News College Rank** (Academic proxy) weak negative correlation with recruiting ranking, PCC =-0.1638
    • APR Ranking; no correlation with recruiting ranking, PCC = 0.076315507 (Excluded)
    • Football Home Attendance; strong positive correlation with recruiting ranking, PCC = 0.77037528
    • Stadium Capacity; strong positive correlation with recruiting ranking, PCC = 0.771508046
    • "Fandom" Ratio***; moderate positive correlation with recruiting ranking, PCC =0.39206026
    • Public vs Private****
    • Conference*****
    The model results have a strong positive correlation with recruiting ranking, PCC = 0.8312. ie my model outperforms the underlying data to predict what a schools recruiting rank "should be".

    The Results Ordered by differential (Over performers at top, under performers at bottom)

    upload_2018-12-1_13-59-1.png

    *Athletics department revenue from publicly reported info or the best estimate I could find for private Schools. Total athletic department revenue, so Football expenditures would probably be a better data point, but I could not find a good source for that data.

    **Both the USnews and APR data sets are a very imperfect measure of the impact of academics on the Football team. Lots of really great schools also have strong Football programs, but the players aren't majoring in the intensive high rated majors. Big schools, like ND, Michigan, UF are able to game the rankings, team GPA, APR, etc, by having the players make great progress towards graduating... in 'recreational studies'. Couldn't find a source for better data, maybe football team average SAT score? I expected academics to be more of a negative, but it seems Academics are a plus for many recruits (maybe not the 3-and-done crowd, but they're a tiny minority and not coming here to play school.) Plenty of smart 4*s out there for us to go get.

    *** "Fandom"- Ratio of student body to football attendance. Proxy for schools that have large/small fanbases relative to alumni. The correlation isn't that strong, but including it in the model helped correct for some outliers on attendance #s- ND, TCU, and Clemson (surprisingly to me) are outliers to the high side. Minnesota, Cal and Illinois are outlying under performers. Private schools and big Southern State U's tend to over perform. Big Northern and Western public schools under perform. GT is smack dab average on attendance relative to our enrollment size. On par with FSU, UF, UT and tOSU. That isn't what you'd think from the narrative that we have a crappy fan base. Our big Southeast State U neighbors skew the perception. You could argue we should outperform, not be average, given that we are a public Southern school in Atlanta. But we also have a disproportionate grad student population relative to total enrollment. Many of those will stay loyal to their undergrad teams (or are international and don't care about gridiron).

    **** Public vs Private - Can't really quantify this, for a correlation as it's a binary facotr. But Private schools averaged 9 places higher than predicted all else equal. Adjusting for this bias made the model more accurate.

    ***** Conference adjustment, didn't include this at first, but the Big Ten schools way under perform (...except Maryland o_O?) and the Pac 12 Overperform. Adjusting for this bias made the model more accurate.

    This is also not considering whether recruiting rankings are an accurate assessment of talent. They do strongly correlate with on field success. But I personally don't believe the 'rankings' are causative. I think both correlate to a third factor, which is overall program strength.
     
  3. BCJacket

    BCJacket Helluva Engineer

    Messages:
    223
    Also, I know one P5 school is missing from the data. I made a mistake partway through my data transcription. It was easier to leave that school out than fix it...
     
  4. TheTechGuy

    TheTechGuy Helluva Engineer

    Messages:
    460
    Great work! Unfortunately, you're missing some factors that are difficult to quantify. Facilities, booster connections, and campus atmosphere just to name a few. While Tech certainly has better academics than "factory" programs, many factories are able to connect their student-athletes to excellent careers because their alums want to hire well-known athletes. This factor does mitigate Tech's advantage in academics.
     
    BCJacket likes this.
  5. 4shotB

    4shotB Helluva Engineer

    Messages:
    1,384
    I second this. Great work. Another key piece of the puzzle as i scan your list is talent nearby. As a general rule, of the schools you have highlighted in pink, the underachievers, are not in areas noted for their great local FB talent. That includes Michigan and ND who therefore must depend on national recruiting. The two notable exceptions - Texas and Texas A & M - compete with a ton of in and out of state schools for their talent including Baylor which is also in the pink category.
     
    Jacket in Dairyland and BCJacket like this.
  6. LibertyTurns

    LibertyTurns Helluva Engineer

    Messages:
    2,443
    Like the analysis, but one comment.

    We were expected to be ranked 55 of 64 & ended up 55th so we didn’t underachieve because we ended up where we expected. I get that. With 64 P5 teams, we’ve got to figure a way to get to #32. At some point we won’t be able to overachieve on the field nearly every single year and our starting point needs to be at least average to give whatever coaching staff/scheme we end up with a fighting chance.
     
    Jacket in Dairyland and BCJacket like this.
  7. BCJacket

    BCJacket Helluva Engineer

    Messages:
    223
    100% correct.

    I'm open to ideas for other data points. I used stadium size in addition to attendance as a proxy for facilities. But it's a very poor proxy. The attendance numbers and "Fan ratio" probably capture some of the intangible 'atmosphere' factors. As you recognized a model like this is limited by what can be quantified and is readily available data.

    Honestly, I started this thinking it was going to be a null result. That I was actually able to use independent data to make a model with a fairly strong correlation to the real world results for something like recruiting rankings is fascinating in itself.

    I do think some of the geographic difference gets smoothed out by adjusting for the conference differences. The SEC, PAC12 and ACC get a boost because they're largely in sunny football loving areas. The northern and midwestern Big 10 and Big 12 schools not so much. If someone wanted to do the work, they could probably rate each state by the quality of recruits in-state and adjacent and include this in the model for each school based on their state. I'm not that someone.
     
  8. 4shotB

    4shotB Helluva Engineer

    Messages:
    1,384
    Nor am I BC. I appreciate your work and I am sure that you understand that my comment was more of an observation than a criticism. Again, thanks for taking the time. I always love those who interject facts, data and analysis into the mix here. Reminds me of a guy who was the plant engineer and maintenance manager in my previous life - his go to line was 'In God I trust - all others must bring data." ;)
     
    BCJacket likes this.
  9. BCJacket

    BCJacket Helluva Engineer

    Messages:
    223
    Fair point, and one that needs to be recognized. All the results say is that our recruiting didn't underachieve relative to our attributes. But as an Institute, it's fair to say we're underachieving on our attributes.

    No way should Tech be 55th out of 64. But I think it's important to recognize that's where we are. I think some folks think with the right coach we could jump into the top 25 in the recruiting rankings. I don't think we have the infrastructure for that. We need to build the program, because a strong program is what brings in the recruits.
     
    GTNavyNuke likes this.
  10. 4shotB

    4shotB Helluva Engineer

    Messages:
    1,384
    The 'build it and they will come" approach. Quite the conundrum for any FB coach outside of the top 10 - 15 FB schools. I think the best coaches have the mindset that they can overcome the hurdles. I think CPJ brought that mindset to the table. But, over time, it takes quite a toll as evidenced by his "need a break" comment. As others have said, it is a shame that Tstan didn't arrive at the beginning of his tenure. Another big "what if" in GT FB history.
     
  11. LibertyTurns

    LibertyTurns Helluva Engineer

    Messages:
    2,443
    @BCJacket Yes, I’d like to see a strategy. Something like, we’re going to invest in or change kthe following:

    Infrastructure: $200M in facilities over the next 5 yrs. Upgrade xxx, xxx and xxx. Funded by special requests in 2018, 2021, etc

    Staffing: $15M per year increase by 2020. Immediate: $2M for assistants, $3M for administrative staff, $3M for recruiting; Coach TBD

    Technology roadmap: How plan to leverage Institute’s advantages

    Game day experience: How to fill up the stands, improve atmosphere, etc

    Institute support: Making required classes available that don’t conflict with practices, flexibility in degree programs, etc. How to continue making student athletes successful.

    Media/advertising: Plans to improve

    I see some stuff being generated by Stansbury, but it’s fairly disconnected/ stovepiped.
     
    biggtfan likes this.
  12. Boaty1

    Boaty1 Helluva Engineer

    Messages:
    484

    Great stuff. Thanks for this.
     
  13. BCJacket

    BCJacket Helluva Engineer

    Messages:
    223
    Good thoughts, I see TStan doing some good things, particularly in facilities and staff. But I'd like it if there was more of a defined strategic direction. Maybe there is one and it's not public.

    FWIW, under the model, if 'a school' our size were to sell out 'a 55,000 seat stadium' and have a budget of $100,000,000 their predicted rank would be 40. Which would be 6th in the ACC, instead of dead last, right in line with VT. (I'm excluding BC, their predicted rank is silly because they're a private ACC school.) Clemson, FSU, UL, and Miami are on a different level. But we could/should be in the leaders of the second tier. Not pulling up the rear behind Wake and Syracuse. Then we over perform from there by 12 spots and we're within arms reach of top 25 talent, which would be 4th-5th in the ACC. Then we need a system and scheme to maximize that talent... I'm not saying that would be easy, but it seems attainable in a medium-term future.

    As I said above, I think GT is very well positioned to out perform. Our budget is smaller, but we have fewer programs. So, our Football budget could be disproportionate. (Title IX issues, I know). We have an outsized postgraduate enrollment. We have, arguably, the best geographic location of any school for recruiting.

    We need to be very smart and "GT" in our resource usage. Leverage technology and data to make the maximum usage within limits. The big boys can afford to throw spaghetti at the wall and see what sticks. We can't afford to do that. How many thousands (millions?) of dollars (even assuming no malfeasance) do you think uGA spent on recruiting Jake Eason? Regular private flightsto Washington, hotels, meals, etc. Staff to baby him constantly for 3 years. All-out spending on his visits. Then they ran him off after one year. We need to be identifying players who fit the Tech mold early on and intelligently recruiting them.

    I believe the biggest thing Tech is failing to take advantage of is 'Atlanta'. We are the P5 program in Atlanta; the capital of college football. Atlanta should be a core part of our identity. Our marketing should highlight Atlanta on everything. Every single piece of clothing we sell should say 'Atlanta' somewhere. This is anecdotal, but I have a theory that a huge proportion of the folks wearing 'Georgia' gear bought it because they live in Georgia and are proud of the state, couldn't name a single uGa player. Let's sell a ton of white, gold and [sorry] blue 'Atlanta' shirts. Also, 'The Institute'. Get a bunch of shirts that look like this into local Costco's:

    The Institute

    Atlanta
    GT

    If people buy the gear because they like it and it says 'Atlanta'. That opens the door to getting them interested in GT athletics if we deliver an interesting product.

    The other thing I believe we should make part of our identity is the success of our black athletes. There was a study out a couple years back that Tech is near the top in graduating it's black athletes. Being in Atlanta, one of the 2-3 centers of black culture in America, it seems like we could market ourselves as the school for the talented, ambitious, intelligent young black athlete. Sell people on the Pat Swilling model: make your first million in the NFL and your second 10-20 in your own business... And if you don't make the NFL, make your first million with Google and then your second 10-20 in your own business.

    That's not going to get the 3-and-done IMG 5*s here. But they haven't come here anyway. We need a lot more JeTs, AJ Grays, and Shamire's. I feel like, over time, we could get the black community in Atlanta to rally around our program, instead of those rednecks out east. We should market our successful young [black] men to the community. JeT, Calvin Johnson, Shaq, etc should be on billboards all across the city and beyond. There's probably tons of smart, ambitious and talented black student athletes across the country that would love to come be a P5 football star in Atlanta. I'd like an entire team of Malik Riveras. Hiring a personable black coach would really help this. It's one thing I like about Tony Elliot.

    I'd also like to see Roddy Jones as the 'voice of the Yellow Jackets'. He's got a great personality, he's from Dekalb county... imagine if we had him out in the community, visiting high schools, etc... pitching GT to Atlanta. (Demetra doesn't seem to bring any value to the table to me.)

    I also believe our brand is better outside Georgia than within. Tech has a kind of split personality in our state. When I tell people I got my degrees from Tech, their eyes get wide and they want to know how I got in; because they wish their kids could go there. (I'm young enough for it to be relevant) When I tell people I'm a Tech football fan, they recoil like I just said "Hail Satan". But to private school kids up North, eg a 4* athlete with a high GPA, Atlanta, ACC football and a Tech degree is highly desirable. We have to expand our recruiting footprint, IMHO. But that takes money.

    Here's the model results ordered by predicted rank. With a 'School 1' of where I think GT could be:

    upload_2018-12-2_9-51-41.png
     
  14. LibertyTurns

    LibertyTurns Helluva Engineer

    Messages:
    2,443
    @BCJacket Exactly. We’re not in the middle of no where. There’s other entertainment options in Atlanta and we need to compete.

    Attendance to me = engagement. I’d rather have 55k engaged at $35 than 37k engaged at $50.

    Minority athlete graduation rates- right on!

    I guess I differ on the recruitment of higher level athlete. We do nothing to leverage our technological excellence to our advantage. We should be at the cutting edge of athletic gear technological development, analytics, sports performance engineering in conjunction with Emory for example.

    There’s lot of things that can be done. We need the desire to do it.
     
  15. takethepoints

    takethepoints Helluva Engineer

    Messages:
    2,776
    Three comments:

    1. Great work. And, I might add, the third post (mine, yours, and Augusta's) based on what passes for reliable recruiting data that has come to the same conclusion: Tech does ok in recruiting, given our limitations and the resources we put into it, and it will take awhile and a lot of money to move up the stalk. Some methodological points, however.

    2. It's ok to calculate a Pearson correlation using a dummy variable. They are continuous (sorta). If you use them as dependent variables in regressions there can be trouble and they are hard to interpret, but …

    3. I would have put my indicators into a principle components analysis and formed an index using the loadings. That would weight the contribution of each indicator to explained variance more evenly. Then you could correlate the index with the recruiting rankings.

    4. Augusta and I used average "stars" instead of rankings since they aren't subject to variance due to the number of slots you have open reach year. But … given that the star ratings are so squirrelly, your choice is probably just as good.

    And, if you are looking for a terrific free data analysis application, try Jamovi (https://www.jamovi.org). It is based on R, surprisingly complete, completely menu driven. I still use R, but Jamovi is growing on me.
     
    GTNavyNuke, slugboy and BCJacket like this.
  16. ibeattetris

    ibeattetris Helluva Engineer

    Messages:
    1,624
  17. Animal02

    Animal02 Helluva Engineer

    Messages:
    3,392
    Disagree about Atlanta .....I just don't think being in the city is that big of an issue/ importance. These kids are not going to the symphony or art museums. They are looking at the shiny objects on campus .......what the dorms are like, party atmosphere, oh and women
     
    AlphaBuzz, Scubapro and smathis30 like this.
  18. AlabamaBuzz

    AlabamaBuzz Helluva Engineer

    Messages:
    1,799

    Agreed, Animal. Also, I believe a lot of parents still think of the ATL as a dangerous place to attend college, and that probably has impacted us over the years.
     
    Animal02 likes this.
  19. BCJacket

    BCJacket Helluva Engineer

    Messages:
    223
    Great find! That was exactly the data I was looking for. Much more strongly correlated (r=.7997) with the results than total athletics revenue (r = .6722)

    Adding that data in lieu of revenue brought the total model correlation up to .8861. Which is ridiculously accurate. Changes things a bit. I also showed average star rating in addition to average ranking since @takethepoints asked about it:

    upload_2018-12-3_23-1-23.png
     
    presjacket, GTRX7 and ibeattetris like this.
  20. BCJacket

    BCJacket Helluva Engineer

    Messages:
    223
    1. Thanks! Do you have a link to the other posts? Or a tip on how to search for them. Don't recall seeing them, might have been before my time on this board. Be very interested to see another approach.

    2. If I follow what you're saying, I think you're talking about the two non-quantitative indicators- public/private and Conference? I didn't use a Pearson correlation for those, which I'm not sure if that was clear- Not sure if you're saying it's ok, because you thought I did do that. Or you're saying that I could have but you followed that I didn't. I analyzed each value separately and corrected for the average impact of each quality. It was a more qualitative approach, but it helped the model correlate more closely with the observations.

    3. I'll look into a principal component analysis when I have a chance. Very interesting. I just used Pearson correlation because that's what I'm familiar with.

    4. I used average stars as well. But the total rankings more closely matched my results, so that's what I used for the 'report' I posted on here. FWIW, I think insofar as the rankings are accurate, they're more accurate than average stars. The schools that are over-signing, grey shirting, and 'processing' derive a benefit from that. The additional total talent pool they're pulling their depth chart from (it's potentially like having 10-15 more scholarships) is captured in the rankings.

    Also, pretty sure your single post had more intelligence than the sum total of some entire uGa boards. (And maybe all Miami boards in existence combined.)
     
    GTNavyNuke likes this.

Share This Page