Toggle sidebar

One For The Nerds

BaccaFarmer

BaccaFarmer

Senior
Aug 20, 2018
194
498
63
Wrestleknownothing said:
You have been warned.

This one is not for everybody.

But if you want to read about the expected points model - have at it.

Expected Points - Friend or Foe? | Wrestling.Guru

The first thing I did when playing around with historical wrestling data back in 2020 was figure out what the average outcome was by seed. A #1 seed doesn&# ...
wrestling.guru wrestling.guru
Click to expand...
Nice work. Itā€™s always great to know that there are others who are more sick than I am about wrestling. Seriously, I read a fair amount of economic and investment papers and am able to wade through thick stuff and come away with a generalized idea of rational conclusions. Just the same, the eye test serves long term wrestling fans well (assuming that they attend and watch a large number of matches). This works very well for prognosticating results WITHIN A CONFERENCE. Where your analysis shines is at Nationals ā€¦. and so we thank you for the effort.
 
  • Like
Reactions: Cali_Nittany1, Psalm 1 guy and Wrestleknownothing
W

Wrestleknownothing

All-Conference
Oct 30, 2021
1,671
4,412
113
BaccaFarmer said:
Nice work. Itā€™s always great to know that there are others who are more sick than I am about wrestling. Seriously, I read a fair amount of economic and investment papers and am able to wade through thick stuff and come away with a generalized idea of rational conclusions. Just the same, the eye test serves long term wrestling fans well (assuming that they attend and watch a large number of matches). This works very well for prognosticating results WITHIN A CONFERENCE. Where your analysis shines is at Nationals ā€¦. and so we thank you for the effort.
Click to expand...
Yes, to be clear this data is only good for making statements about the NCAA tournament. It is not even applicable to the B1G tournament (smaller size, many forfeits, makes a big difference) never mind a dual format.
 
  • Like
Reactions: Psalm 1 guy
Col

Col

Sophomore
Oct 17, 2021
71
133
33
Assuming all results are not independent variables ( I submit that bonus points tend to have a weak dependency and wins have a stronger dependency on previous wins against the same wrestler) could you take your model and design a simulation (run lets say 1000 times) to get a even better range of outcome with a pretty good idea of the likelihood of each outcome? To be fair not sure you can establish a good dependency coefficient and being off just a little would probably invalidate the simulations.
 
W

Wrestleknownothing

All-Conference
Oct 30, 2021
1,671
4,412
113
Col said:
Assuming all results are not independent variables ( I submit that bonus points tend to have a weak dependency and wins have a stronger dependency on previous wins against the same wrestler) could you take your model and design a simulation (run lets say 1000 times) to get a even better range of outcome with a pretty good idea of the likelihood of each outcome? To be fair not sure you can establish a good dependency coefficient and being off just a little would probably invalidate the simulations.
Click to expand...
If I were attempting to predict the outcome of dual meets or other narrow results, like individual bouts, I think a Monte Carlo process would work better than my current process, but given the highly structured repeating nature of the NCAA tournament data there is no accuracy benefit from a simulation process and it would come at a huge computational cost.

There is a guy on Intermat (Pablo) who says he has done this for volleyball (match predictions) and now is attempting to do it for wrestling (bout and dual predictions). In addition to complaining about the cost of doing these calcs he admits that they do not work very well until at or near the end of the season.
 
  • Like
Reactions: Col
You must log in or register to reply here.
Top Bottom