Jump to content

I simulated the State Tournament 1,000,000 Times


Recommended Posts

I got bored and I wanted to avoid the actual work that I have to do, so I coded a Monte Carlo simulation of the IHSAA state wrestling tournament in Matlab.

 

To determine win probabilities of wrestlers, I used a rating system comprising rank, record, and grade (~67% rank, 20% record, ~13% grade). With this, an undefeated senior ranked #1 has about a 90% win percentage against the worst wrestler meaning that the worst wrestler has about a 95% chance of winning one match if about 30 matches are wrestled against the best wrestler.

 

I set the simulation to run 1,000,000 times (admittedly this was probably ridiculous, but oh well). The amount of times each wrestler got each place was tabulated and used to make the observations below (Some Many of which are trivial).

 

106:

  •  Lowery is top 4 58.63% of the time (73% of the times he places)
  • Miller wins 18.36% of the time (the most despite being on the same side as (2) and (3)
  •  Miller only makes it to the finals 32.16% of the time, but wins 57% of the finals matches he’s in
    • Chundi has very similar results
  • Triana wins the 2nd least of any wrestler at any weight (0.14%)

 

113:

  • Cottey wins the 2nd most of any other wrestler at any weight (27.93%)
  • Cottey and Ross are only ones at 113 that win more than 10% of the time

 

120:

  • Seltzer is top 3 more than half of the time (58.44%)
  • Diaz actually ends up better than Bettich on average

 

126:

  • Dalton and Roberson have very similar chances at 1st 2nd and 3rd
  • Rioux wins 20.45% with 365802 finals appearances

 

132: Nothing too significant

 

138:

  • Campbell makes the finals more than Bailey, but loses more often when he gets there, leading to similar 1st place finishes

 

145: Pretty good pairings here for top 4, no real surprises

 

152:

  • Boe, Kervin, and Hall all win more than Coleman
  • Coleman wins almost 60% of the 254,538 finals matches

 

160:

  • Levitz has the best shot at reaching the finals here (32.76%)

 

170:

  • On average, Deters is able to do better than Calhoun

 

182:

  • Walker has the highest finals win percentage of anyone
  • Warren makes his way to the finals 273,963 times, but has a 50-50 shot at winning once he’s there
    • Lone is very similar, just less finals appearances

 

195:

  • Silas runs away with 319,526 times on top of the podium
  • If kemper makes the finals, he wins more often than not

 

220: Not too much interesting here

 

285:

  • Dancy and Irick win more than Keys,
  • Swallow makes a solid run most of the time placing top 5 about 52.80% of the time

 

Summary:

Most of these actual numbers are completely meaningless as they are only as good as the data I had at my disposal. The overall trends however are at least somewhat representative of what is likely. I thought this would be an interesting look at the state tournament and it was (for me at least) and really highlights how effective a good or bad draw can be. That’s all from me for now, but ill answer just about any questions either about the simulation or how certain people finished. Overall, it was fun and I hope to try this again with better ratings and such, potentially using old state tournaments as training data for a machine learning rating system.

Monte Carlo Results.pdf

Edited by Tony Ramos Gin Fizz
Link to comment
Share on other sites

All right...I'll do it...

 

I forced the Indianamat.com bot who has been posting all season to read the brackets and message board 1,000 times, and here are the results.

 

An Indiana Wrestling fan laughs at an ankle band. A referee hits him and the ankle band for double stalling. 

 

800 lbs. Gorilla: This guy is tough as nails.His opponent is also a nail.

 

Referee 1: You are stalling with extra passivity. 

 

Region Wrestling fan 1: Booooooooo. Angery my fingers extend. Twooooooooooo.

 

Referee 2: Also twoooooooooooo, but also stalling. 

 

Indy Wrestling Fan 1: It's not fair that you are calling twoooooooooooo. Booooooooo.

 

800 lbs. Gorilla: Nails for sale by the dozen. 

 

Region Wrestling Fan 2: The region will have 46 state champs this season. Rats wrestle as tough as a rat tooth. Nails are soft like children.

 

Evansville Wrestling Fan 1: I like everyone. All are good who wrestle. 

 

All other wrestling fans yell at gentle Evansville grandparents.

 

Referee 1: Continuous stalling on all. I wish I had more fists.

 

Ankle band: I do not like being strapped to the shoulder of large boys.

 

The tournament is over, but the buzzers continue to alarm the spectators who only yell that it isn't fair. Fade to black. 

 

Link to comment
Share on other sites

2 hours ago, Tony Ramos Gin Fizz said:

I got bored and I wanted to avoid the actual work that I have to do, so I coded a Monte Carlo simulation of the IHSAA state wrestling tournament in Matlab.

 

To determine win probabilities of wrestlers, I used a rating system comprising rank, record, and grade (~67% rank, 20% record, ~13% grade). With this, an undefeated senior ranked #1 has about a 90% win percentage against the worst wrestler meaning that the worst wrestler has about a 95% chance of winning one match if about 30 matches are wrestled against the best wrestler.

 

I set the simulation to run 1,000,000 times (admittedly this was probably ridiculous, but oh well). The amount of times each wrestler got each place was tabulated and used to make the observations below (Some Many of which are trivial).

 

106:

  •  Lowery is top 4 58.63% of the time (73% of the times he places)
  • Miller wins 18.36% of the time (the most despite being on the same side as (2) and (3)
  •  Miller only makes it to the finals 32.16% of the time, but wins 57% of the finals matches he’s in
    • Chundi has very similar results
  • Triana wins the 2nd least of any wrestler at any weight (0.14%)

 

113:

  • Cottey wins the 2nd most of any other wrestler at any weight (27.93%)
  • Cottey and Ross are only ones at 113 that win more than 10% of the time

 

120:

  • Seltzer is top 3 more than half of the time (58.44%)
  • Diaz actually ends up better than Bettich on average

 

126:

  • Dalton and Roberson have very similar chances at 1st 2nd and 3rd
  • Rioux wins 20.45% with 365802 finals appearances

 

132: Nothing too significant

 

138:

  • Campbell makes the finals more than Bailey, but loses more often when he gets there, leading to similar 1st place finishes

 

145: Pretty good pairings here for top 4, no real surprises

 

152:

  • Boe, Kervin, and Hall all win more than Coleman
  • Coleman wins almost 60% of the 254,538 finals matches

 

160:

  • Levitz has the best shot at reaching the finals here (32.76%)

 

170:

  • On average, Deters is able to do better than Calhoun

 

182:

  • Walker has the highest finals win percentage of anyone
  • Warren makes his way to the finals 273,963 times, but has a 50-50 shot at winning once he’s there
    • Lone is very similar, just less finals appearances

 

195:

  • Silas runs away with 319,526 times on top of the podium
  • If kemper makes the finals, he wins more often than not

 

220: Not too much interesting here

 

285:

  • Dancy and Irick win more than Keys,
  • Swallow makes a solid run most of the time placing top 5 about 52.80% of the time

 

Summary:

Most of these actual numbers are completely meaningless as they are only as good as the data I had at my disposal. The overall trends however are at least somewhat representative of what is likely. I thought this would be an interesting look at the state tournament and it was (for me at least) and really highlights how effective a good or bad draw can be. That’s all from me for now, but ill answer just about any questions either about the simulation or how certain people finished. Overall, it was fun and I hope to try this again with better ratings and such, potentially using old state tournaments as training data for a machine learning rating system.

feel better in love GIF 
 

this is what we need more of on this Site . Facts, numbers, stats that don’t lie . LOVE IT ! 

Link to comment
Share on other sites

2 hours ago, base said:

Can we please get a 1 million times simulation of General HeavyHandz vs FCFighter?!?  Amiright?

confused math GIF by CBC

YES ! It’s in that equation

(Ā) = 1- P (A+)

right there . It’s says DDQ for bitting . Guess one of them liked to bite. 

Edited by TeamGarcia
Link to comment
Share on other sites

3 hours ago, NW Indiana Wrestling Fan said:

I’d like to see who ended in each placement the most to see how accurate this really may be🤷🏻‍♀️

I'll post the results in a bit. To answer how "accurate" it is, it is only as accurate as looking at the indianamat brackets with rankings and guessing how much more likely each wrestler is to beat all the others based on that and then determining placement probability based on that. The goal this time around was not accuracy, but to have a start at a way to model the tournament.

 

I chose a monte carlo because it was a lot easier than markov chains. Also, this allows me to change some parameters in the future to describe things in a way that markov chains cannot because in reality the probability of winning is not based solely on your location in the state space. 

Link to comment
Share on other sites

6 hours ago, Jcjcjc said:

Can someone please make a "I made my AI machine watch 1000 hours of Olive Garden commericals, here are the results" meme of this please? 

You had me at Olive Garden, but then I got lost again with all this wrestling data people keep bringing up.

tenor.gif?itemid=14031479

Edited by MattM
Link to comment
Share on other sites

7 hours ago, Tony Ramos Gin Fizz said:

I got bored and I wanted to avoid the actual work that I have to do, so I coded a Monte Carlo simulation of the IHSAA state wrestling tournament in Matlab.

 

To determine win probabilities of wrestlers, I used a rating system comprising rank, record, and grade (~67% rank, 20% record, ~13% grade). With this, an undefeated senior ranked #1 has about a 90% win percentage against the worst wrestler meaning that the worst wrestler has about a 95% chance of winning one match if about 30 matches are wrestled against the best wrestler.

 

I set the simulation to run 1,000,000 times (admittedly this was probably ridiculous, but oh well). The amount of times each wrestler got each place was tabulated and used to make the observations below (Some Many of which are trivial).

 

106:

  •  Lowery is top 4 58.63% of the time (73% of the times he places)
  • Miller wins 18.36% of the time (the most despite being on the same side as (2) and (3)
  •  Miller only makes it to the finals 32.16% of the time, but wins 57% of the finals matches he’s in
    • Chundi has very similar results
  • Triana wins the 2nd least of any wrestler at any weight (0.14%)

 

113:

  • Cottey wins the 2nd most of any other wrestler at any weight (27.93%)
  • Cottey and Ross are only ones at 113 that win more than 10% of the time

 

120:

  • Seltzer is top 3 more than half of the time (58.44%)
  • Diaz actually ends up better than Bettich on average

 

126:

  • Dalton and Roberson have very similar chances at 1st 2nd and 3rd
  • Rioux wins 20.45% with 365802 finals appearances

 

132: Nothing too significant

 

138:

  • Campbell makes the finals more than Bailey, but loses more often when he gets there, leading to similar 1st place finishes

 

145: Pretty good pairings here for top 4, no real surprises

 

152:

  • Boe, Kervin, and Hall all win more than Coleman
  • Coleman wins almost 60% of the 254,538 finals matches

 

160:

  • Levitz has the best shot at reaching the finals here (32.76%)

 

170:

  • On average, Deters is able to do better than Calhoun

 

182:

  • Walker has the highest finals win percentage of anyone
  • Warren makes his way to the finals 273,963 times, but has a 50-50 shot at winning once he’s there
    • Lone is very similar, just less finals appearances

 

195:

  • Silas runs away with 319,526 times on top of the podium
  • If kemper makes the finals, he wins more often than not

 

220: Not too much interesting here

 

285:

  • Dancy and Irick win more than Keys,
  • Swallow makes a solid run most of the time placing top 5 about 52.80% of the time

 

Summary:

Most of these actual numbers are completely meaningless as they are only as good as the data I had at my disposal. The overall trends however are at least somewhat representative of what is likely. I thought this would be an interesting look at the state tournament and it was (for me at least) and really highlights how effective a good or bad draw can be. That’s all from me for now, but ill answer just about any questions either about the simulation or how certain people finished. Overall, it was fun and I hope to try this again with better ratings and such, potentially using old state tournaments as training data for a machine learning rating system.

Monte Carlo Results.pdf 260.97 kB · 20 downloads

Sheldon Cooper

Link to comment
Share on other sites

×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.