Analysis - NCAA teams with the best yearly improvements

Men's 1500m from 2016 - 2017:

improvement(seconds) team

8.31 South Dakota St.

6.10 New Mexico

6.00 UCLA

5.49 Notre Dame

3.78 Southern Utah

3.73 Florida State

3.64 Georgia Tech

3.52 Northern Arizona

3.44 Colorado St.

3.28 Morehead State

3.20 Wisconsin

2.92 Georgia

2.92 Furman

2.88 Maryland-Eastern Shore

2.82 Iowa State

2.78 Illinois

2.77 Portland

Men's 1500m from 2015 - 2016:

improvement(seconds) team

8.65 Arkansas

7.15 Navy

6.94 UTSA

6.64 St. Mary's (Cal.)

5.91 UMKC

5.59 McNeese State

5.42 Northeastern

5.13 Boston U.

4.91 Memphis

4.82 William and Mary

4.57 Auburn

4.27 SE Missouri

3.98 Princeton

3.84 Cornell

3.80 North Florida

3.75 Yale

3.36 South Alabama

3.20 Texas A&M

3.15 Iona

3.06 Connecticut

2.92 Dartmouth

2.87 Oakland

2.77 Kentucky

2.75 Oklahoma State

2.72 Wichita State

2.42 Miami (Ohio)

Men's 1500m from 2014 - 2015:

improvement(seconds) team

6.97 Louisville

6.60 Dartmouth

6.44 Yale

6.40 Houston

6.35 UC Riverside

5.61 California

5.43 George Mason

5.39 Montana State

5.21 Illinois

5.19 Cal Poly

4.50 Kent State

4.37 Virginia

4.30 UTSA

4.26 Southern Utah

4.15 UMass Amherst

4.00 Va. Military Institute

3.98 Arkansas

3.89 Kentucky

3.88 Clemson

3.86 UMass Lowell

3.71 Bucknell

3.51 Penn

Teams that were in the top 50 improvements all 4 years:

Yale

Auburn

Georgia

Oklahoma State

This information is interesting to me as someone who is deciding where to run in college next year as I think I will eventually be a 1500 runner. My better event right now is the 800 but I think I have more potential in the 1500. Thanks.

Thanks for taking the time to do this. Any chance you can do the same for the women?

women? wrote:

Thanks for taking the time to do this. Any chance you can do the same for the women?

Sorry for the wait. I can only do this between other work.

Women's 1500m - schools that are in the top 75 of improvement all 4 years:

Penn State

Wake Forest

Oklahoma State

South Dakota

Arkansas

Kansas

SMU

Stanford

Dartmouth

UC Santa Barbara

Oregon

Michigan

Bradlet

Penn

Vanderbilt

Women's 1500m from 2016 - 2017:

improvement team

1 21.41999984 New Hampshire

2 8.920000076 Furman

3 8.570000172 UC Santa Barbara

4 8.365000129 Eastern Kentucky

5 7.840000153 Texas

6 7.690000057 Northern Illinois

7 7.139999986 Samford

8 6.559999943 Georgia Tech

9 6.49000001 Notre Dame

10 6.309999943 South Dakota

11 6.150000095 Marquette

12 6.110000134 Delaware

13 5.625 Princeton

14 5.279999971 Quinnipiac

15 5.140000105 North Carolina

16 4.840000153 Maryland

17 4.670000076 Loyola Marymount

18 4.569999933 Cal Poly

19 4.440000057 Mid. Tenn. State

20 4.393333356 San Francisco

goo.gl/r2y3zV

Women's 1500m from 2015 - 2016:

improvement team

1 10.73000002 North Carolina St.

2 9.960000038 Rice

3 9.439999819 BYU

4 9.200000048 Akron

5 8.980000019 Samford

6 8.879999876 Missouri

7 8.609999895 Bucknell

8 8.5 San Jose St.

9 8.440000057 Drake

10 7.869999886 Harvard

11 6.679999828 Duke

12 6.255000114 Penn State

13 5.625 St. John's

14 5.443333387 Miss State

15 5.439999819 Fresno State

16 4.470000029 Coastal Carolina

17 4.399999857 Cornell

18 4.325000048 UC Santa Barbara

19 4.119999886 Pittsburgh

20 4.089999914 Indiana State

goo.gl/IljUMn

Women's 1500m from 2014 - 2015:

improvement team

1 12.83000016 Boise State

2 12.56999993 New Hampshire

3 10.46000004 Colorado

4 8.870000124 UC Riverside

5 8.50999999 Indiana

6 8.129999955 Yale

7 7.910000086 Clemson

8 7.100000143 Elon

9 7.039999962 Penn State

10 7 Montana State

11 6.75999999 LSU

12 6.660000086 UMBC

13 6.649999857 Notre Dame

14 6.630000114 Eastern Kentucky

15 6.579999924 Northeastern

16 6.360000054 Virginia Tech

17 6.24000001 UC Davis

18 6.089999914 Wake Forest

19 5.720000029 West Virginia

20 5.529999971 William and Mary

goo.gl/t0XxZl

No worries! Thanks again.

There are a lot of "flukes" in individual years, so I think the most legitimate result is the schools that can maintain improvement for all 4 years (for example, above I showed the top few guy's 1500m schools and the top ~15 women's 1500m schools). I think I'll just post those 4-year results instead of the individual year numbers (plus, it will take up less room in the posts).

Women's 5000m

Teams with the top improvements all 4 years

Air Force

Appalachian State

Vanderbilt

Arizona

Stanford

Indiana

Bucknell

Auburn

Penn

Providence

Eastern Kentucky

Minnesota

Boise State

Eastern Washington

Wyoming

Oklahoma

Coastal Carolina

Oops, that should be...

Air Force

Vanderbilt

Arkansas

Alabama

Central Michigan

Indiana

William and Mary

Auburn

Miss State

Eastern Kentucky

Minnesota

Eastern Washington

Oklahoma

Coastal Carolina

Loyola Marymount

Men's 5000m

Teams with the top improvements all 4 years

Arkansas

BYU

Cal Poly

Drake

Duquesne

Eastern Illinois

Georgia Tech

Gonzaga

Illinois

Lehigh

Loyola Marymount

Miami (Ohio)

Michigan State

North Carolina State

Northern Arizona

Oakland

stats.gangsta_the_real_1 wrote:

For each school with times in the NCAA top ~500, I calculated how many seconds their runners improved each year from Spring 2014 through Spring 2017.

I'm not sure about the best way to display and format this information, so I'm open to suggestions.

I'll post the calculations throughout the rest of the day.

So, is this how you did the math:

1) grouped all of the schools who had runners listed in the top X of the year end list

2) find the average time for each school

3) compared the averages from Year A to Year B to determine net improvement

I think it would also be interesting to note the number of athletes per school that are listed among the top X of a given year and see how the number of athletes who are within the top X changes between the given years. If the number of athletes that are within the top X rises as the performance average changes, then you could argue that the overall program is improving. Does that make sense?

This is really interesting information. Can you tell us more about your methods?

I'd be interested to know:

1. Source for the times

2. Number of athletes from each of the schools

3. I assume you are looking at improvement curve for individual runners?

4. Is there a way to work out which of these runners were peaking in an event? So, if I run a 1500 in early season, and another mid-season, I may not improve that much - but that might be fine if I'm really a 5000m runner and will run the 5000 in championships.

5. Across 4 year comparisons, do you know if coaches remained constant?

There is so much noise in this data that this analysis is useless. Please don't use this to judge your college choices. This is just random.

asdfagaga wrote:

There is so much noise in this data that this analysis is useless. Please don't use this to judge your college choices. This is just random.

I agree. It reminds me of an analysis of the top hedge fund managers.

Given the number of schools analyzed, what is the expected value for the number of teams in the top-x for 4 straight years, assuming random data. You could replace the real data with random data and run a monte carlo. I suspect you'll find that there's little "skill" in there.

As a data dude myself, it would be cool if you can do this and maybe post on a blog somewhere (plenty of free blog sites. I use wordpress). Maybe clean up the format in your spreadsheet. Of if you didn't use excel, paste into excel and clean it up.

As your work load permits, you can go back further in time and develop a useful database for high school seniors and other interested parties to view.

Definitely interesting but also a LOT of bad information in this.

A school could be on that list for so many reasons outside of "improvement." If you're looking for which schools have improved in those events, regardless of the source or reason, then it's sorta useful. If you're looking at which school develops their runners the best in those events then you will most likely be completely misled by utilizing these data.

If one could track every single athlete and their improvement/drop out/injury rate for every program, across all of their events, you would have a serious ranking list that I would LOVE to see. Probably possible but probably extremely time intensive.

OP, how are you calculating the improvement? Each runner? Team as a whole? Top 5?

If a female improves 5 seconds a year in the 1500 going from 4:40 to 4:25 that's not as impressive as improving 2.5 seconds a year going from 4:20 to 4:12.5

Using the IAAF points table might be more helpful.

Some challenges you face with this approach:

1) Heteroskedasticity: Middle and lower-tier teams can have much greater inter-annual swings with the gain or loss of a top runner. Using absolute seconds, you're going to definitely see the lower tier teams turning into middle tier teams but it will be harder to see sub-elite teams turning elite.

2) Interannual variability: Some years are faster than others, especially when the top times are set at "fast" courses or when runners are not able to run the same course under the same conditions in subsequent years. So you end up comparing different courses possibly at different times of the season.

3) Regional variability: Weather can impact regions differently. Favorable weather during fast racing times in the west may coincide with poor weather in the northeast or midwest, perhaps, when PRs would be set under better weather conditions. It's a subtle variation related to #2.

#3 is hard to overcome but #1 and #2 are more straightforward.

Calculate a z-score for every runner's times for each given year. That is, a runner's 2016 score would be their best time minus the mean of all runners' best times for 2016 divided by the standard deviation for those times. Then perform your analysis on the z-scores. Then you can rank them based on the Z-score improvements/declines. It's harder to interpret, but you can back the math out to real values in terms of percentages relative to the group mean. But at least the rankings of team improvements will be more honest.

might add that if you do this, you should really publish the dataset as a spreadsheet so the number crunchers can really dig into it and see why the results turned out like they did. I'm not even getting into other issues with this approach related to personnel and event changes. In track the runner pool for doing these stats by event is actually really small. For some reason (because I don't read) I had it in my mind that this was for Cross Country. I think doing it by event for track could be misleading.

I think it would be interesting to compare improvement of individual places at XC Regionals year to year. Consistent competition level and not affected by running conditions