WrestleStat Rankings - Week 3 2020

[andegre]

After our upset-packed weekend, the rankings have been re-processed.

Some notables:

Sebastian Rivera has 2 matches at 133, and his ranking at that weight is now #2, right behind Seth Gross.
Brent Moore of Virginia Tech moves up 12 spots to #4 after pinning #2 Sammy Sasso.
Ethan Smith of Ohio State moved up 10 spots to close at #12 at 165 pounds.
Dean Sherry also gets 14-place bump all the way to #12 after his upset win over Skatzka of Minnesota.
Kyle Conel slips down 2 spots to finish at #4 at 197.
There's a new #2 at 285, and he works for Michigan. Mason Parris trades places with Matt Stencel after his upset win last weekend. Don't worry though, Stencel has a chance at revenge THIS WEEK as Central Michigan and Michigan meet in a dual.


Wrestler Rankings: https://www.wrestlestat.com/rankings/starters

Dual Rankings: https://www.wrestlestat.com/rankings/dual

Tournament Rankings: https://www.wrestlestat.com/rankings/tournament

Redshirt Rankings: https://www.wrestlestat.com/rankings/redshirts

Statistical Rankings: https://www.wrestlestat.com/rankings/statistical



One reminder about the dual rankings; they do NOT reflect previous head-to-head dual results. They are a result of running each team through the prediction engine to see who would win in a dual based on the individual matchups.

 

[Abro1975]

RU starters within the Big 10 rankings (westlestat)

Wrestlestat rankings within the Big 10:

125- Aguilar #10

133- Alvarez #7

141- Aragona #14

149- Angelo #4

157- Van Brill- #11

165- Turley #8

174- Grello #6

184- Janzer #8

197- Pagano #9

285- Colucci #8

 

[MS-RU]

Wow, look at Angelo at #4 in the B1G

 

[bigmatt718]

RU starters within the Big 10 rankings (westlestat)

Wrestlestat rankings within the Big 10:

125- Aguilar #10

133- Alvarez #7

141- Aragona #14

149- Angelo #4

157- Van Brill- #11

165- Turley #8

174- Grello #6

184- Janzer #8

197- Pagano #9

285- Colucci #8

How absurdly deep is the B1G at 141 if Aragona is dead last?

 

[MS-RU]

How absurdly deep is the B1G at 141 if Aragona is dead last?

To be fair, he doesn't have enough data points for the algorithm to have any "confidence" he can beat higher ranked guys. If he didn't drop the decision during the Fresno dual, he would probably be significantly higher. With him being injured now and probably not wrestling for the next few weeks, he is going to have a very poor rankings for the time being.

He has only beaten #240, #72, #135 and two unranked guys since they aren't D1 opponents.

His two losses are to #12 (Chad Red from Nebraska) and #67 (Durbin from FSU)

With that being said, according to wrestlestat, the B1G makes up 5 (1,3,5,6,10) of the national top 10 and 7 of the top 20

 

[Abro1975]

Interesting that our first 4 wrestlers all have last names that start with the letter A (and 3 of the 4 containing 4 syllables ).

 

[Abro1975]

The retooling of the Wrestlestat ranking algorithm has been a big improvement.

I could be wrong but Aragona's low ranking could be that only 1 of his 5 wins appears to be against D1 starter, that being the LIU wrestler who has a very low ranking, approx 240. Aragona's 2 losses have been relatively close, and 1 (Chad Redd) to a pretty highly ranked wrestler. I'm not sure if the Wrestlestat algorithm factors in a close loss (quality loss ?) to a much higher ranked opponent.

Would like to hear from Andegre (or anyone) who might be in the know , about that part of their algorithm.

From Wrestlestat's FAQs :

How are the WrestleStat Rankings generated?
WrestleStat Rankings are 100% computer generated and is based on a custom variation of the popular Rating system called Elo (see Wikipedia for detailed information). The Elo rating system is used primarily in Chess, as well as Yahoo Games, and some other online gaming sites. There have been a few modifications to the algorithm to make this work well for wrestling. The algorithm factors in the margin of victory in every match. A FALL produces the largest margin of victory, followed by Tech Fall 5, then Tech Fall 4, etc. Other special weighting is applied for wrestlers that have fewer total matches. Any non-traditional results (forfeits, etc) do not have the margin of victory factored in when calculating the rating change for that particular match.

 

[MS-RU]

Typically In ELO rankings, winners essentially take points from the loser. If the winner had a higher score/ranking, they take less points from the loser since it was the expected outcome. If the winner was ranked lower, making it an upset, they will take more points from their opponent relative to the difference in their ranking plus other factors like margin of victory, etc.

With limited data points, models basically know opponent A is generally better than opponent B, but it will struggle to really know by "how much" until there are more results. There are sensitivities as well which determine how much the rankings should be impacted by new results so until JoJo starts winning matches where he is expected to lose (as per ELO) he really won't shoot up. Beating opponents close to your ranking that you are expected to beat doesn't really get you any where quickly.

 

[andegre]

The retooling of the Wrestlestat ranking algorithm has been a big improvement.

I could be wrong but Aragona's low ranking could be that only 1 of his 5 wins appears to be against D1 starter, that being the LIU wrestler who has a very low ranking, approx 240. Aragona's 2 losses have been relatively close, and 1 (Chad Redd) to a pretty highly ranked wrestler. I'm not sure if the Wrestlestat algorithm factors in a close loss (quality loss ?) to a much higher ranked opponent.

Would like to hear from Andegre (or anyone) who might be in the know , about that part of their algorithm.

From Wrestlestat's FAQs :

How are the WrestleStat Rankings generated?
WrestleStat Rankings are 100% computer generated and is based on a custom variation of the popular Rating system called Elo (see Wikipedia for detailed information). The Elo rating system is used primarily in Chess, as well as Yahoo Games, and some other online gaming sites. There have been a few modifications to the algorithm to make this work well for wrestling. The algorithm factors in the margin of victory in every match. A FALL produces the largest margin of victory, followed by Tech Fall 5, then Tech Fall 4, etc. Other special weighting is applied for wrestlers that have fewer total matches. Any non-traditional results (forfeits, etc) do not have the margin of victory factored in when calculating the rating change for that particular match.

You are 100% correct. A "close loss" will BARELY move a wrestler down, while getting demolished/pinned hurts badly.

Also, the opponents rating is factored in as well, just like you assumed. His close loss to Red probably did little to nothing on his rating, while a close loss to someone ranked #240 would drop him even more.

If I remember right, I believe the K value in the Elo algorithm is mapped to the margin of victory (since margin of victory isn't factored in the true Elo algorithm). The actual Elo algorithm I believe uses a k value of 32, whereas ours is anywhere from 1 to 20 (with MANY other factors added in).