A look at March Madness: Pretenders and Contenders

smathis30

Ramblin' Wreck
Messages
732
Every year, a friend and I take a look at the coding competition at Kaggle and try to predict who will do well in March Madness each year by predicting the probability of each and every possible matchup. The data for this is pulled by bringing in information from various websites, and then using machine learning to assign weights to each factor to determine how much of an impact it has on score. Here are the factors, in order of importance they have on score, as well as some factors obtained that don't indicate relative success:

1. Ken Pom Offensive Efficiency ranking
2. Ken Pom Defensive Efficiency ranking
3. Turnover Ratio
4. # of players on team that average at least 10 ppg
5. Opponent Free Throw rate below 0.31
6. Ken Pom Team Ranking
7. Opponent 2pt FG% below 0.48
8. Did team win their conference tournament or regular season title
9. Has the coach made the Sweet 16 in the past
10. Does the team have a forward or center with at least 10ppg

Non-factors:
Seed, RPI ranking, 3pt%


With that, here is an overview of each division with teams in order of their rank in that division:
Midwest
1. Duke (+1)
2. Kansas (-1)
3. Michigan State
4. Auburn
5. Clemson
6. Arizona State/Syracuse (both fit here) (+5)
7. Seton Hall (+1)
8. TCU (-2)
9. NC State
10. Oklahoma
11. Rhode Island (-4)
12. New Mexico State
13. Iona (+2)
14. Penn (+2)
15. Charleston (-2)
16. Bucknell (-2)

IMO Duke and Kansas have the best shot. Michigan State looks promising, but this team has all the flaws that the 2016 team had on paper: The absolute worst turnover ratio of a top 4 seed in the tournament at 0.78. Other Teams in the past with similar turnover ratios at or below 0.80:
2017 #9 MSU-Round of 32
2016 #5 Purdue-lost first round
2016 #2 MSU-lost in first round
2016 #4 Cal- lost in first round
2015 #11 Texas- Round of 32
2013 #7 Creighton- Round of 32

So any team with a turnover ration below 0.80 has never made the sweet 16. It wouldn't be weird for them to break precedent on it, but its definietely a major red flag. They due have the best draw of any team as to who their matchup is against, with Bucknell being garbage.

Two outright upsets predicted, with Syrcause/Arizona state beating Rhode Island and Oklahoma beating Seton Hall

Onto the next division, the west:


1. Michigan (+2)
2. Gonzaga (+2)
3. Ohio State (+2)
4. UNC (-2)
5. Houston (+1)
6. Xavier (-5)
7. San Diego State (+4)
8. Texas A and M (-1)
9. Florida State
10. South Dakota State (+2)
11. Mizzou (-3)
12. Providence (-2)
13. UNC Greensboro
14. Montana
15. Texas Southern
16. Lipscomb

Notes: In the regular season, Gonzaga whomped Ohio State at Ohio State and UNC tore Michigan a new one at home right after getting whomped by Michigan State. This time Gonzaga gets to play Ohio State at more or less a nuetral site in Boise, Idaho. But thats assuming that Ohio state wins their glorified road game vs South Dakota State. I like Gonzaga over Xavier due to the homecourt advantadge they will pull on the west coast. UNC over Michigan due to season results, then pick your poison for the champ. I like the Zags again due to home court advantadge

Next Division, the South:

1. UVA
2. Cincinnati
3. Kentucky (+2)
4. Nevada (+3)
5. Tennessee (-2)
6. Arizona (-2)
7. Creighton (+1)
8. Loyola-Chi (+3)
9. Davidson (+3)
10. Texas
11. Miami FL (-5)
12. Kansas St (-3)
13. Buffallo
14. UMBC (+2)
15. Wright State (-1)
16. Ga state (-1)

UVA was the #1 overall team (to no ones surprise). Loyola-Chicago predicted with the outright upset as well. Mostly due to Miami having a bad defense and not having a lot of high scorers, with only 2 players with 10ppg.

For the sweet 16, Tennessee-Cincy should be a great matchup due to Tennessee having a homecourt advantage in Atlanta, and Kentucky-UVA for the same reason. UVA drew a really short straw with how many conference champs ended up in their division. Whoever wins between UVA and UK wins the tourney IMO

Last division, the east:

1. Nova
2. Purdue
3. Florida (+3)
4. West Virginia (+1)
5. Virginia Tech (+3)
6. Texas Tech (-3)
7. Butler (+3)
8. Wichita State (-4)
9. UCLA/St. Bona (+2)
10. Arkansas (-3)
11. Murray State (+1)
12. Bama (-3)
13. SF Austin (+1)
14. Marshall (-1)
15. Radford (+1)
16. Cal-Fullerton (-1)

Mostly Chalk, and Wichita State being overseeded instead of underseeded for the first time ever!
Villinova pulling Boston was a lot of luck, so i think they have the easiest walk to the final 4 out of every team
Texas Tech pulling Dallas was also a lot of luck

In short Teams that have a shot:
Virginia
UNC
Gonzaga
Xavier
Duke
Villinova
Purdue
 

MikeJackets1967

Helluva Engineer
Messages
14,844
Location
Lovely Ducktown,Tennessee
Every year, a friend and I take a look at the coding competition at Kaggle and try to predict who will do well in March Madness each year by predicting the probability of each and every possible matchup. The data for this is pulled by bringing in information from various websites, and then using machine learning to assign weights to each factor to determine how much of an impact it has on score. Here are the factors, in order of importance they have on score, as well as some factors obtained that don't indicate relative success:

1. Ken Pom Offensive Efficiency ranking
2. Ken Pom Defensive Efficiency ranking
3. Turnover Ratio
4. # of players on team that average at least 10 ppg
5. Opponent Free Throw rate below 0.31
6. Ken Pom Team Ranking
7. Opponent 2pt FG% below 0.48
8. Did team win their conference tournament or regular season title
9. Has the coach made the Sweet 16 in the past
10. Does the team have a forward or center with at least 10ppg

Non-factors:
Seed, RPI ranking, 3pt%


With that, here is an overview of each division with teams in order of their rank in that division:
Midwest
1. Duke (+1)
2. Kansas (-1)
3. Michigan State
4. Auburn
5. Clemson
6. Arizona State/Syracuse (both fit here) (+5)
7. Seton Hall (+1)
8. TCU (-2)
9. NC State
10. Oklahoma
11. Rhode Island (-4)
12. New Mexico State
13. Iona (+2)
14. Penn (+2)
15. Charleston (-2)
16. Bucknell (-2)

IMO Duke and Kansas have the best shot. Michigan State looks promising, but this team has all the flaws that the 2016 team had on paper: The absolute worst turnover ratio of a top 4 seed in the tournament at 0.78. Other Teams in the past with similar turnover ratios at or below 0.80:
2017 #9 MSU-Round of 32
2016 #5 Purdue-lost first round
2016 #2 MSU-lost in first round
2016 #4 Cal- lost in first round
2015 #11 Texas- Round of 32
2013 #7 Creighton- Round of 32

So any team with a turnover ration below 0.80 has never made the sweet 16. It wouldn't be weird for them to break precedent on it, but its definietely a major red flag. They due have the best draw of any team as to who their matchup is against, with Bucknell being garbage.

Two outright upsets predicted, with Syrcause/Arizona state beating Rhode Island and Oklahoma beating Seton Hall

Onto the next division, the west:


1. Michigan (+2)
2. Gonzaga (+2)
3. Ohio State (+2)
4. UNC (-2)
5. Houston (+1)
6. Xavier (-5)
7. San Diego State (+4)
8. Texas A and M (-1)
9. Florida State
10. South Dakota State (+2)
11. Mizzou (-3)
12. Providence (-2)
13. UNC Greensboro
14. Montana
15. Texas Southern
16. Lipscomb

Notes: In the regular season, Gonzaga whomped Ohio State at Ohio State and UNC tore Michigan a new one at home right after getting whomped by Michigan State. This time Gonzaga gets to play Ohio State at more or less a nuetral site in Boise, Idaho. But thats assuming that Ohio state wins their glorified road game vs South Dakota State. I like Gonzaga over Xavier due to the homecourt advantadge they will pull on the west coast. UNC over Michigan due to season results, then pick your poison for the champ. I like the Zags again due to home court advantadge

Next Division, the South:

1. UVA
2. Cincinnati
3. Kentucky (+2)
4. Nevada (+3)
5. Tennessee (-2)
6. Arizona (-2)
7. Creighton (+1)
8. Loyola-Chi (+3)
9. Davidson (+3)
10. Texas
11. Miami FL (-5)
12. Kansas St (-3)
13. Buffallo
14. UMBC (+2)
15. Wright State (-1)
16. Ga state (-1)

UVA was the #1 overall team (to no ones surprise). Loyola-Chicago predicted with the outright upset as well. Mostly due to Miami having a bad defense and not having a lot of high scorers, with only 2 players with 10ppg.

For the sweet 16, Tennessee-Cincy should be a great matchup due to Tennessee having a homecourt advantage in Atlanta, and Kentucky-UVA for the same reason. UVA drew a really short straw with how many conference champs ended up in their division. Whoever wins between UVA and UK wins the tourney IMO

Last division, the east:

1. Nova
2. Purdue
3. Florida (+3)
4. West Virginia (+1)
5. Virginia Tech (+3)
6. Texas Tech (-3)
7. Butler (+3)
8. Wichita State (-4)
9. UCLA/St. Bona (+2)
10. Arkansas (-3)
11. Murray State (+1)
12. Bama (-3)
13. SF Austin (+1)
14. Marshall (-1)
15. Radford (+1)
16. Cal-Fullerton (-1)

Mostly Chalk, and Wichita State being overseeded instead of underseeded for the first time ever!
Villinova pulling Boston was a lot of luck, so i think they have the easiest walk to the final 4 out of every team
Texas Tech pulling Dallas was also a lot of luck

In short Teams that have a shot:
Virginia
UNC
Gonzaga
Xavier
Duke
Villinova
Purdue
I think Tennessee beats Wright State,Miami,Cincinnati and Virginia and makes it to the Final 4 playing North Carolina in the first game of the Final 4 and Villanova in the Championship Game. I think Tennessee avenges losses to North Carolina and Villanova winning the National Championship sort of like NCSU in 1983;)
 
Last edited:
Top