Want to share this, particularly since the results are looking correct other than 1 major factor (elites).

I am a data scientist between jobs, filling time by analyzing this game until my new job starts in a week. Today, I built a model to analyze which teams were the best from a recruiting perspective, using 3 seasons of D1A recruit Data (Wilk 201, Rockne 200/201). I calculated the distance of two things:
1) Recruits to each school.
2) Each school to each other.

I then created a metric that looks at recruits per school, given some custom weighting. I tried to weight schools based on how close they were to each other (a school 5 miles away weighs more than a school 300 miles away) up to 500 miles. I did not do this weighting with recruits, but instead looked at all recruits within 500 miles. I also weighted power conferences (by my math they will have 17% more money per recruit assuming each automatic non-power champion loses in the 1st round), and elites. I also weighted down non-power schools by 80% to represent the fact that those are more likely sim teams. Where did the 80% come from? made it up :)

The results look correct, except some of the well held beliefs (like Penn State being 1 or 2 with USC) don't hold. Im curious if anyone thinks its due to another reason, or if its simply that I'm not properly weighting being an elite school high enough.

Below is an excerpt of my results.

1. USC
2. BC
3. UCONN
4. LSU
5. Miami (FL)
6. UCLA
7. Texas
8. Ohio State
9. Michigan
10 (Tie) Rice & Houston (Houston)
12. Syracuse
13. Army
14. Florida
15. Nebraska
16. Oklahoma
17. Rutgers
18. Alabama
19. Florida State
20. Notre Dame
21. Tennessee
21. Temple
22. Penn State
23. Arizona
24. East Carolina
25. Arizona State
26. Buffalo
27. Maryland
28. Navy
29. Nevada
30. Virginia
31. NC State
32. UNC
33. Duke
34. Pittsburgh
35. UCF
36. San Diego State
37 (Tie). Air Force & Colorado State (Colorado Springs, TX)
39. UNLV
40. Fresno State
41. South Florida
42. Kent State
43. Cal
44. Stanford
...

116. Colorado
117. Washington
118. Wyoming
119. Idaho
120. Hawaii

A few things jump out at me:
1. Some schools do seem to be really good schools (like Rutgers), but they don't have human coaches as often as real life good schools, so their reputation seems down among members of this game. This works similarly true for a school like Washington, just in the other direction.
2. Being Elite is valuable (i forced this), but theres something about elites competing against each other that Im missing.
3. I may be underweighting the extra cash per recruit at 17% since you can pool it together into only a few battles.

Curious other's thoughts as I try to refine this.

Choosing Wilk could cause a bit of an outlier for you when it comes to Penn St- It isnt common that Syracuse, Pitt, BC, UCONN, Temple, and Army are all filled in other worlds.

Also in Wilk, Penn St is neutralized with their Elite bump (still exists) by all those Big East Schools cash bump. I think your 17% bump is low (although only my hunch) - I think its 25% on the low end based on my recruiting at USC/Tennessee/FSU vs Iowa/Colorado St/Arkansas

Also I have never had Penn St, but the largest availability/concentration of good players to me was when I coached Rockne UCONN - There were so many good players in the Northeast (based on how this game distributes recruits) that Penn St, Temple, UCONN, BC are all perennial Top 10 programs in that world and have minimal fighting over recruits.

I am very interested to see how this goes if you extend beyond these 2 worlds and build on your methodology.

I agree with John. Wilkinson/Penn State is a massive outlier compared with Penn State's position in other worlds. The number of cash-rich Big East schools in Wilkinson is crazy, and in other worlds the Big East is mostly either sparsely populated or not much of a factor, and Penn State rules over northeast recruiting.

Posted by johnalanharr on 12/28/2023 5:11:00 PM (view original):
Choosing Wilk could cause a bit of an outlier for you when it comes to Penn St- It isnt common that Syracuse, Pitt, BC, UCONN, Temple, and Army are all filled in other worlds.

Also in Wilk, Penn St is neutralized with their Elite bump (still exists) by all those Big East Schools cash bump. I think your 17% bump is low (although only my hunch) - I think its 25% on the low end based on my recruiting at USC/Tennessee/FSU vs Iowa/Colorado St/Arkansas

Also I have never had Penn St, but the largest availability/concentration of good players to me was when I coached Rockne UCONN - There were so many good players in the Northeast (based on how this game distributes recruits) that Penn St, Temple, UCONN, BC are all perennial Top 10 programs in that world and have minimal fighting over recruits.

I am very interested to see how this goes if you extend beyond these 2 worlds and build on your methodology.

I only looked at Wilk players, not who signed whom. I did some world specific analyses, but the one Im sharing here was just an all up analysis based on player generation from WIS.

I gave each elite team a 30% bump, then for the elite teams themselves (since they are elite), I haircut the number of teams in their area as well to bump elites up in my rankings. Just interesting that PSU is so low. It could be because I created a hard cutoff mileage wise, but I need to investigate further.

If you have files of players from other worlds, I'm happy to include them in my analysis. These are just the only 2 worlds where I have D1 teams.

Posted by sportsdouche on 12/28/2023 5:29:00 PM (view original):
I agree with John. Wilkinson/Penn State is a massive outlier compared with Penn State's position in other worlds. The number of cash-rich Big East schools in Wilkinson is crazy, and in other worlds the Big East is mostly either sparsely populated or not much of a factor, and Penn State rules over northeast recruiting.

This is what I'm curious about. I wonder if PSU is regarded as the top team because they ARE, or because the schools that would compete with them are generally Sim teams and so they run the show in the northeast. My analysis is suggesting the latter.

You can't use Wilkenson present day in any recruiting analysis due to the Big East being a Super Conference of All Star coaches. Go back 20 or 30 seasons, before the super conference was organized, and then you can do an analysis. But otherwise, you're wasting your time analyzing Wilkenson D-1A recruiting without understanding the presence of a Super Conference there.

Posted by WingMan474 on 12/28/2023 7:20:00 PM (view original):
You can't use Wilkenson present day in any recruiting analysis due to the Big East being a Super Conference of All Star coaches. Go back 20 or 30 seasons, before the super conference was organized, and then you can do an analysis. But otherwise, you're wasting your time analyzing Wilkenson D-1A recruiting without understanding the presence of a Super Conference there.

I only look at recruits generated, not which teams get which recruits. As you say, theres too much variance around quality of coach, etc. to make that analysis worthwhile.

Is there something about the generation of players in Wilkinson that I should be weary of?

Posted by greener10 on 12/28/2023 7:37:00 PM (view original):

Posted by WingMan474 on 12/28/2023 7:20:00 PM (view original):
You can't use Wilkenson present day in any recruiting analysis due to the Big East being a Super Conference of All Star coaches. Go back 20 or 30 seasons, before the super conference was organized, and then you can do an analysis. But otherwise, you're wasting your time analyzing Wilkenson D-1A recruiting without understanding the presence of a Super Conference there.

I only look at recruits generated, not which teams get which recruits. As you say, theres too much variance around quality of coach, etc. to make that analysis worthwhile.

Is there something about the generation of players in Wilkinson that I should be weary of?

No and I think I understand what you were doing better now

Can you tune your parameters to 360 and 180 miles since those are the biggest jump in recruiting costs?

So this measures recruit availability - Does recruit quality factor here?

Posted by johnalanharr on 12/28/2023 8:30:00 PM (view original):

Posted by greener10 on 12/28/2023 7:37:00 PM (view original):

Posted by WingMan474 on 12/28/2023 7:20:00 PM (view original):
You can't use Wilkenson present day in any recruiting analysis due to the Big East being a Super Conference of All Star coaches. Go back 20 or 30 seasons, before the super conference was organized, and then you can do an analysis. But otherwise, you're wasting your time analyzing Wilkenson D-1A recruiting without understanding the presence of a Super Conference there.

I only look at recruits generated, not which teams get which recruits. As you say, theres too much variance around quality of coach, etc. to make that analysis worthwhile.

Is there something about the generation of players in Wilkinson that I should be weary of?

No and I think I understand what you were doing better now

Can you tune your parameters to 360 and 180 miles since those are the biggest jump in recruiting costs?

So this measures recruit availability - Does recruit quality factor here?

I certainly can tune parameters to 180 and 360. Send me a sitemail and we can hammer out the details and report back here.

Recruit quality does not factor in. Because I'm trying to create a generalized analysis, I assumed each recruit is normally distributed around some mean quality value. Worried that looking at quality with only 3 seasons of data would overfit to how the system generated THESE specific recruits. If we know that certain areas produce worse players we could adjust for that, but my assumption is that its pretty random in terms of which recruits are good/bad on a year-to-year basis. Or if you want to send me more seasons of data, I could try to do that too, but it will greatly slow down my algo as I compress recruits into locations so I can iterate distances faster.

Elites are even money on any recruit inside their 360 vs any non-elite BCS schools 180. If you re-run the data for all elites 360 being equal to every other non-elite BCS schools 180 you will find out why PSU is the best job in the game.

Add 60-70th percentile. If there’s 100 recruits only 30-40 of them will make the percentile for a top 25 class.

Fascinating. I updated my algo based on conversations with John and Rice. I now calculate 3 ratios, the ratio of recruits to teams at 180, 360, and 600 miles. I weight up recruits by a full category in recruits (IE the 180 value for recruits at elite is the actual number of recruits at 360 miles). I then score it where 50% of the weight is 180, 30% is 360, and 20% is 600. Im still seeing that Penn State is the worst elite and that the best non-elites are in the North East. I wonder if what is happening in Wilkinson is actually representitve of the game engine and we have just never had a ton of population in the Big East to challenge Penn state. Said another way, if the big east has a normal population, Penn State is not overpowered, but if people don’t populate the big East then Penn State gets the very populated NE.

Anyway, love more feedback and ideas. Here are the current rankings.

1. Team elite power recruit_ratio

2. 1 USC 1 1 282.416842

3. 16 Miami (FL) 1 1 233.257487

4. 32 LSU 1 1 230.245472

5. 28 Boston College 0 1 156.682887

6. 41 Syracuse 0 1 134.527586

7. 0 Temple 0 1 132.842771

8. 10 Connecticut 0 1 131.686818

9. 5 Oklahoma 1 1 123.212799

10. 8 Texas 1 1 121.561734

11. 11 Rutgers 0 1 119.197782

12. 84 Army 0 0 116.852793

13. 14 Florida 1 1 112.142542

14. 19 Nebraska 1 1 111.822068

15. 4 UCLA 0 1 108.445854

16. 45 Minnesota 0 1 104.733968

17. 22 Florida State 1 1 103.262571

18. 42 Ohio State 1 1 93.786602

19. 20 Michigan 1 1 90.080030

20. 52 Navy 0 0 88.504003

21. 36 Tennessee 1 1 83.475495

22. 7 Alabama 1 1 83.464475

23. 79 Maryland 0 1 80.679811

24. 6 Notre Dame 1 1 79.847233

25. 73 Rice 0 0 79.845310

26. 70 Houston 0 0 79.845310

27. 71 Houston 0 0 79.845310

28. 72 Rice 0 0 79.845310

29. 17 Penn State 1 1 71.260794

30. 39 Stanford 0 1 57.719262

31. 86 Iowa State 0 1 55.099853

32. 30 California 0 1 52.869522

33. 82 Arizona State 0 1 52.737779

34. 90 Buffalo 0 0 52.370219

35. 114 Central Michigan 0 0 51.159672

36. 9 Pittsburgh 0 1 51.126073

37. 35 Marquette 0 1 50.174418

38. 38 Washington 0 1 49.622291

39. 111 Fresno State 0 0 49.374128

40. 37 Texas A&M 0 1 48.799692

41. 49 Missouri 0 1 47.754308

42. 74 Nevada 0 0 47.648863

43. 105 Wisconsin 0 1 47.611602

44. 101 San Jose State 0 0 46.772500

45. 68 Virginia 0 1 45.849006

46. 53 West Virginia 0 1 45.499870

47. 56 Baylor 0 1 44.581570

48. 118 San Diego State 0 0 43.621810

49. 64 Central Florida 0 0 43.319347

50. 80 Arkansas State 0 0 42.710039

Well I found one issue: my API had the wrong location for Penn State. Once I corrected that they became 3rd on the list behind USC and LSU (I made some other tweaks that are better representative of population. I think this is getting close to being a good ranking of the best jobs in the game.

1. Team elite power recruit_ratio

2. 1 USC 1 1 451.393948

3. 32 LSU 1 1 309.865943

4. 17 Penn State 1 1 259.856389

5. 28 Boston College 0 1 259.356618

6. 16 Miami (FL) 1 1 241.658131

7. 0 Temple 0 1 228.573224

8. 10 Connecticut 0 1 226.625977

9. 11 Rutgers 0 1 216.013228

10. 84 Army 0 0 214.009804

11. 19 Nebraska 1 1 195.407071

12. 8 Texas 1 1 190.240541

13. 5 Oklahoma 1 1 190.098053

14. 41 Syracuse 0 1 184.147300

15. 52 Navy 0 0 157.340256

16. 79 Maryland 0 1 142.862529

17. 42 Ohio State 1 1 141.253721

18. 14 Florida 1 1 128.963369

19. 4 UCLA 0 1 128.566327

20. 22 Florida State 1 1 125.149583

21. 6 Notre Dame 1 1 124.880622

22. 36 Tennessee 1 1 124.487432

23. 20 Michigan 1 1 123.124309

24. 45 Minnesota 0 1 114.950183

25. 7 Alabama 1 1 111.521197

26. 73 Rice 0 0 99.235002

27. 71 Houston 0 0 99.235002

Getting better. I still would take either Arizona schools over Rice or Houston

One additional follow up is how you handle the Big East - In those 2 worlds it has earned its Power designation you give it but for the game as a whole it seems like it is viewed as lesser than the Power 5 but greater than the Sun Belt, MAC, WAC, and USA - But that's qualitative analysis in knocking of Sims in Recruiting

Just for questions sake - Whats the Standard Deviation for the dataset as a whole? Find it interesting the gap between 1 and 2 is the same as about 2 and 15