ChatGPT picks the USA MAG Olympic Team
What would the US MAG Olympic Team look like if you just used the first three days of data?
I realize the selection process will be more complex than this, using data from both a set of all scores from 4 days and a set with the low dropped. I didn’t take into account an individual gymnast could be auto-selected by finishing top 3 in 3 events. Because so many contenders had low scores on key events, I predict the teams using 4 scores and 1 dropped score will not be the same, so the selection committee will have a role in the final selection.
I entered some numbers into my computer and pressed some buttons. Unless they are computer scientists or statisticians, this is likely the way the Men’s Program Committee will get their initial teams. This is how they can come up with teams within 20 minutes. It takes seconds to test multiple scenarios.
I used the average of the scores from the first three days, just to see what would happen.
For Round 1 gave ChatGPT the explanation of the format and asked for the top 5 teams.
Here are the top 5 highest scoring teams of gymnasts based on their total scores:
Team 1: Juda, Malone, Moldauer, Nedoroscik, Richard
Total Score: 258.2375
Team 2: Malone, Moldauer, Nedoroscik, Richard, Wiskus
Total Score: 258.1250
Team 3: Hoopes, Juda, Malone, Moldauer, Richard
Total Score: 258.0860
Team 4: Hoopes, Malone, Moldauer, Richard, Wiskus
Total Score: 257.9735
Team 5: Hong, Juda, Malone, Nedoroscik, Richard
Total Score: 257.9500
Then I asked ChatGPT to do a Monte Carlo Analysis. ChatGPT states, “A Monte Carlo analysis is a computational technique that uses random sampling and statistical modeling to estimate the probability distribution of possible outcomes in a complex system or process.” The first method I tested is just arithmetic, which is why the results are not the same.
It went all Monte Carlo on me, meaning this thing likes to take risks.
Take a look at team #1. It probably will never happen because it would result in 3 up 3 count in qualifications on every event except horse and parallel bars (Hoopes competes PB).
Top 5 Teams from Monte Carlo Analysis
Team: Hoopes, Moldauer, Nedoroscik, Richard, Malone
Total Score: 257.8110
Team: Nedoroscik, Moldauer, Bock, Richard, Malone
Total Score: 257.7625
Team: Nedoroscik, Hong, Richard, Malone, Wiskus
Total Score: 257.7500
Team: Malone, Wiskus, Hong, Richard, Hoopes
Total Score: 257.5985
Team: Nedoroscik, Hong, Malone, Bock, Richard
Total Score: 257.5875
Team # 1 isn’t totally insane, but I’m sure if USAG picks something like this there will be a riot.
Thing the computer is probably not getting wrong:
Malone and Richard are on every team.
This could change if Young or Moldauer go lights out on pommels on the final day and/or Hoopes and Nedoroscik have major errors.
Notice that every single team has a pommel specialist on it. Pommels will be a key event in determining the makeup of the team. The above scenarios don’t include Khoi Young, as he underperformed on that event on two days. Most of the teams created from the first two days had Young on them. The committee may end up with the dilemma of deciding whether or not to put someone on the team for an event on which he only hit 50 % of routines. The computer didn’t think his 5.6 vault alone was enough. Without knowing the code USAG is using, I don’t know what their computer will think of it.
Anyone who thinks the above teams are dumb can complain to ChatGPT and Python. I just pushed the buttons.
Hope no one accidentally enters a wrong score and that whatever code they use works properly.
All the decisions should have some numbers behind them, even if the selection goes to the committee.