For fun, I used GPT-4 to fill out a 2023 March Madness Bracket (NCAA Men’s Basketball Tournament) – and compared it to my own.
Result? Both my bracket and the GPT-4 bracket had the same total number of games predicted correctly (33 total), but it beat me in total points 48 vs. 44 (only because Gonzaga beat UCLA).
I don’t follow college basketball closely, but enjoy the March Madness tourney… my final 4 predictions were: Alabama, Tennessee, Houston, and Kansas – with Alabama vs. Houston in the championship, and Alabama as champions.
GPT-4 had Alabama, Purdue, Houston, and Gonzaga in its final 4 – with Houston vs. Gonzaga in the championship, and Gonzaga as champions.
For the final 16 teams, I filled out a “second chance” bracket on ESPN – and had GPT-4 do the same… neither bracket did well, but it beat me… a lot of luck involved here.
For the final 4 teams, both GPT-4 and myself predicted: UConn beating Miami, SDSU beating FAU, and UConn winning the championship… but this is probably what most predicted.
Update: GPT-4 vs. My Prediction (UConn vs. SDSU)
Both GPT-4 and myself predicted that UConn would defeat SDSU.
GPT-4: Predicted that UConn would win in a relatively close game 78-72.
Drew: Predicted that UConn would win by double digits (10+) and cover the spread… no exact score prediction though.
Both GPT-4 and myself were correct in predicting the winner… but my “eye test”-based prediction prevailed over GPT-4’s analytical prediction in getting closer to the margin of victory (probably random luck).
The opening minutes of the game had me thinking SDSU may be able to hang with UConn… but at halftime I figured we were en route to a blowout.
SDSU surprised me late in the game – getting to within 6 points or something, but this was followed by an SDSU brick-laying clinic & UConn regaining composure.
Entertaining game, but UConn was too much for SDSU to handle… continuing their tournament domination through the championship.
Note: The only expert I’m aware of who picked UConn to win the championship prior to the tournament is Jay Bilas (he had UConn defeating Duke).
GPT-4 predicts UConn vs. SDSU (NCAA Men’s D1 Championship 2023)
I could ask GPT-4 to make a random prediction between UConn & SDSU for the national championship game based on just odds or something, but I want to give it some quality data to work with so that it can make an informed prediction.
Feeding GPT-4 data
Initially I fed GPT-4 the following: KenPom data (updated), tournament scores for each team, roster attributes, and statistical information (each player).
I asked if any additional information would be helpful, and it requested: head-to-head record from this season (if available), injury reports & player availability, and coaching strategies.
- KenPom: Updated KenPom data for UConn & SDSU (through April 3, 2023).
- Tournament scores: Scores for every UConn & SDSU game throughout the NCAA tournament.
- Rosters: Entire lineup, heights & weights for each player, etc.
- Statistics: Statistical data for each of the rosters (FT %, FG %, 3-PT %, RBs, etc.). It knew how each team shoots, who takes the most shots, etc.
- Coaching: Information about the coaches for each team – including records over the past several seasons.
- Injury info.: GPT-4 requested updated information about injuries for each team. (No major injuries – just one player probable who had food poisoning).
- Intangibles: Commentary on intangibles of each team from experts (relatively subjective).
- Location: Of the championship game (probably didn’t matter).
GPT-4 Matchup Analysis (UConn vs. SDSU)
Based on the provided KenPom data, tournament scores, betting odds, and roster data, here’s an analysis of the upcoming NCAA Championship game between UConn and SDSU, and a prediction of the winner.
UConn:
- Ranked 5th in KenPom adjusted offensive efficiency and 8th in adjusted defensive efficiency.
- Has a balanced team with good depth, with several players contributing in various aspects of the game.
- Top scorer Adama Sanogo averages 17.2 points per game and has a strong field goal percentage of 60.7%.
- Donovan Clingan is a significant defensive presence with 1.8 blocks per game.
SDSU:
- Ranked 9th in KenPom adjusted offensive efficiency and 6th in adjusted defensive efficiency.
- Strong defensive team with an emphasis on team play and ball movement.
- Top scorer Matt Bradley averages 18.7 points per game and has a solid field goal percentage of 43.3%.
- Nathan Mensah provides a defensive anchor with 2.1 blocks per game.
- Betting odds slightly favor UConn.
Considering the provided information, the game is expected to be a closely contested matchup, with both teams boasting strong defenses and efficient offenses.
UConn has an advantage in offensive efficiency, while SDSU has a slight edge in defensive efficiency.
UConn has a more balanced roster, with multiple players capable of contributing on both ends of the floor.
Additionally, the presence of Donovan Clingan, a 7’2″ center, could potentially cause problems for SDSU’s interior offense.
SDSU, on the other hand, has a strong defensive core and relies on teamwork and ball movement to create scoring opportunities.
Taking all these factors into account, I predict that UConn will win the NCAA Championship game.
Their balanced roster and strong performances throughout the tournament make them a formidable opponent.
The combination of Adama Sanogo’s scoring ability and Donovan Clingan’s defensive presence could be the key to overcoming SDSU’s strong defense.
However, the game is likely to be highly competitive, and SDSU should not be underestimated.
Predict a final score for fun…
Predicting an exact score total can be challenging, as it depends on various unpredictable factors such as player form on the game day, coaching decisions, and game tempo, among others.
- UConn: 78
- San Diego State: 72
Based on the data provided, both teams have strong defenses and efficient offenses.
Taking into account their per-game point averages in the tournament, UConn scores around 78.5 points per game, while SDSU scores approximately 76.6 points per game.
Considering their defensive prowess, we can expect a relatively lower-scoring game.
With these factors in mind, an estimated score total for the game could be in the range of 145 to 155 points, with UConn winning by a small margin.
However, this is merely an estimation and should be taken with caution, as the actual outcome may vary significantly based on the factors mentioned earlier.
Injuries: Connecticut starting guard Jordan Hawkins is listed as probable with food poisoning. Since he played well in the semifinal against Miami, it’s likely he will contribute significantly to the championship game. San Diego State reports no injuries.
Coaching: UConn’s Danny Hurley and SDSU’s Brian Dutcher are both excellent coaches. Dutcher has a better career coaching win percentage and two 30+ win seasons in the last four years. While Hurley is on the rise, Dutcher’s sustained success makes this aspect even between the two teams.
Intangibles: UConn has the edge in terms of confidence and swagger, embodying the attitude of their head coach. SDSU, on the other hand, has good karma and momentum on their side, thriving in their role as an overlooked underdog.
Given this additional information, the championship game appears to be even more closely contested. UConn has a strong roster, a probable key player returning from illness, and a confident attitude. SDSU has momentum, an experienced coach, and the underdog advantage.
However, considering UConn’s overall performance, the depth of their roster, and the probable contribution of Jordan Hawkins, I maintain my prediction that UConn will win the championship game by a small margin.
As previously estimated, the score total for the game might be in the range of 145 to 155 points. Keep in mind that this is only an estimation and the actual outcome may vary significantly.
For fun, I’ll make an exact score prediction for the NCAA championship game between UConn and SDSU. Please remember that this is purely speculative, and the actual outcome may vary significantly.
GPT-4 score explanation…
GPT-4’s prediction of UConn winning 78-72 over San Diego State is based on the information I provided and the following factors:
- UConn has a stronger roster with more depth and higher overall statistics.
- UConn’s tournament performance has been impressive, with convincing wins against strong opponents.
- Jordan Hawkins, UConn’s starting guard, is listed as probable despite having food poisoning. His performance in the semifinal game against Miami indicates that he should still be a significant contributor in the championship game.
- While both teams have experienced and successful coaches, the intangibles favor UConn, with the team displaying a high level of confidence and determination.
- San Diego State has been an underdog throughout the tournament, and while they’ve performed admirably, UConn seems to have a stronger overall team.
- I arrived at the specific score prediction by considering the average points scored by each team during the tournament. UConn has been averaging around 78 points per game, and San Diego State has been scoring slightly less.
- Additionally, I took into account the intensity and pressure of a championship game, which could lead to a more closely contested match.
- Again, this is a speculative prediction, and the actual outcome may vary significantly from the numbers presented here.
Drew’s prediction: UConn double digit win (10+ points)
My prediction is that UConn will win the 2023 NCAA Men’s basketball championship (over SDSU) by at least double-digits (10+ points).
This prediction isn’t based on any science or statistical analysis… mostly just the “eye test” (watching both UConn & SDSU play throughout the tournament).
I also made my prediction prior to using GPT-4… just wanted to see how my “eye test” or “gut-shot” type pick compared to GPT-4’s analytical pick.
UConn: Deep roster (strategically created), strong offense & defense (balanced), excellent coaching, Clingan (height), dominant (blow out wins). (Eye test: Can anyone compete?)
SDSU: Elite defense, garbage shooting (streaky at times), solid coaching, not dominant (close games). (Eye test: Beatable – could’ve lost to Creighton & FAU).
Because it’s a single-game format and both teams are well-coached, anything is possible – SDSU might surprise me and eke out a close win or something if: (1) they’re hitting shots & (2) their elite defense somehow thwarts UConn’s attack.
That said, UConn absolutely punished every team they’ve faced thus far in the tournament:
- 87-63 (Iona): Expected first round beat down.
- 70-55 (St. Mary’s): Continued destruction.
- 88-65 (Arkansas): I expected a close game here… wrong.
- 82-54 (Gonzaga): I expected a close game here… wrong again.
- 72-59 (Miami): Figured Miami would lose… continued dominance.
I don’t expect SDSU (of all teams) to perform better than Miami or Gonzaga against UConn… and think UConn should cover the -7.5-point spread with ease unless something goes haywire.
What’s your prediction for UConn vs. SDSU?
Have you been following the March Madness 2023 tournament?
Who do you think will win the UConn vs. SDSU game? (Why?)