WrestleStat Rankings - Week 11

[andegre]

A bug was found that affected last week's rankings where the ranking process was incorrectly factoring FOR, MFOR, and INJ win types. That bug has been resolved.

One other thing I want to note since I'm sure these will be asked...

125 - Daton Fix - he moved up, yet didn't have any new matches because he was directly affected by the bug mentioned above
141 - AJ Jaffe - by-product of the bug mentioned above, still investigating
165 - Joseph at #3 while Massa is #2 - still investigating, the algorithm should have switched these two
184 - Bowman - still investigating, he should not have moved down
197 - Shakur Rasheed - this is a direct reflection of depth of the 197 weight class. If he were to be in the 184 pound weight class, he would actually be ranked #8, which would be closer to what he probably should be. Based on the algorithm though, he should probably be in the 15/16 range - still investigating
285 - Sam Stoll - should not have moved down, should be in the #3 position based on the algorithm - still investigating

Events missing from this weeks rankings:
- Purple Raider Open

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

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

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

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

Hodge Watch: https://www.wrestlestat.com/rankings/hodge

(Create your own rankings) Tournament Projection Tool: https://www.wrestlestat.com/tourney/projection

 

[cbond]

A bug was found that affected last week's rankings where the ranking process was incorrectly factoring FOR, MFOR, and INJ win types. That bug has been resolved.

One other thing I want to note since I'm sure these will be asked...

125 - Daton Fix - he moved up, yet didn't have any new matches because he was directly affected by the bug mentioned above
141 - AJ Jaffe - by-product of the bug mentioned above, still investigating
165 - Joseph at #3 while Massa is #2 - still investigating, the algorithm should have switched these two
184 - Bowman - still investigating, he should not have moved down
197 - Shakur Rasheed - this is a direct reflection of depth of the 197 weight class. If he were to be in the 184 pound weight class, he would actually be ranked #8, which would be closer to what he probably should be. Based on the algorithm though, he should probably be in the 15/16 range - still investigating
285 - Sam Stoll - should not have moved down, should be in the #3 position based on the algorithm - still investigating

Events missing from this weeks rankings:
- Purple Raider Open

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

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

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

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

Hodge Watch: https://www.wrestlestat.com/rankings/hodge

(Create your own rankings) Tournament Projection Tool: https://www.wrestlestat.com/tourney/projection

White loses to the #3 guy in 2 SV periods and drops 12 spots. That is one harsh algorithm.

 

[ionel]

A bug was found that affected last week's rankings where the ranking process was incorrectly factoring FOR, MFOR, and INJ win types. That bug has been resolved.

One other thing I want to note since I'm sure these will be asked...

125 - Daton Fix - he moved up, yet didn't have any new matches because he was directly affected by the bug mentioned above
141 - AJ Jaffe - by-product of the bug mentioned above, still investigating
165 - Joseph at #3 while Massa is #2 - still investigating, the algorithm should have switched these two
184 - Bowman - still investigating, he should not have moved down
197 - Shakur Rasheed - this is a direct reflection of depth of the 197 weight class. If he were to be in the 184 pound weight class, he would actually be ranked #8, which would be closer to what he probably should be. Based on the algorithm though, he should probably be in the 15/16 range - still investigating
285 - Sam Stoll - should not have moved down, should be in the #3 position based on the algorithm - still investigating

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

Again do like your web site, but you already admitted these are career points, thus not ranking, thus little meaning for current season. Rest of web site is great.

 

[Wrassler17]

Andegre, love the website and I use it often. I've seen you answer some questions about the rankings/predictions on the Iowa and PSU boards and I have a couple more if you can stand it. And I'm not looking to start an argument, just so you know.

On the rankings, you've said they are based on the ELO rating algorithm originally created for chess. Are there any modifications made for when a wrestler changes weight classes (kind of like in Derek White's case)? So if a wrestler shows marked improvement from changing weights, is that factored or weighted any differently?

On the dual predictions, I realize it's not based strictly on the wrestler's ranking or ELO points, but is something like home/away/neutral factored in as well. For example, let's say Tony Ramos never lost in Carver, would the prediction algorithm factor that in?

 

[oberebo]

Andegre, love the website and I use it often. I've seen you answer some questions about the rankings/predictions on the Iowa and PSU boards and I have a couple more if you can stand it. And I'm not looking to start an argument, just so you know.

On the rankings, you've said they are based on the ELO rating algorithm originally created for chess. Are there any modifications made for when a wrestler changes weight classes (kind of like in Derek White's case)? So if a wrestler shows marked improvement from changing weights, is that factored or weighted any differently?

On the dual predictions, I realize it's not based strictly on the wrestler's ranking or ELO points, but is something like home/away/neutral factored in as well. For example, let's say Tony Ramos never lost in Carver, would the prediction algorithm factor that in?

I too like and use this website but I would not know an algorithim if it bit me on the ass. I try not to get upset by rankings because as far as I can tell this site is the worst for rankings. It has a lot of great features however and you can tell a lot of work has been done by Andegre.

 

[Wrassler17]

I too like and use this website but I would not know an algorithim if it bit me on the ass. I try not to get upset by rankings because as far as I can tell this site is the worst for rankings. It has a lot of great features however and you can tell a lot of work has been done by Andegre.

Ha, got to watch out for those nasty algorithms. I also don't get upset by any rankings, you can't win anything on paper or a blackboard.

The wrestling sites are all good until wrestler A beats wrestler B, wrestler B beats wrestler C, and then wrestler C beats wrestler A. I also think they sometimes overreact to a win. You'll see an unranked guy jump into the top 10. Or they don't consider a weight change - the famous one being Flo ranking Gulibon #1 at 141 before he wrestled a match there. But there's no perfect way.

I'm not a programmer, but I find the wrestlestat rankings kind of interesting. I just think maybe the more recent results should be prioritized - would anyone bet against Jordan Burroughs his last 2 years at Nebraska? And also head-to-head should maybe outweigh scoring dominance - David Taylor might tech everyone while Kyle Dake beats them 6-0, but after some head to head matchups you probably want to pick Dake.

Anyway, kind of interesting.

 

[andegre]

Good questions...

1) Weight change does not change the calculations/algorithms at all. It's assumed that the quality of wrestlers [and their ELO values] are basically "equal" across the different weights. So if a wrestler moves up or down, doesn't matter, when the rankings run, it's still just wrestler A [ELO value] vs wrestler B [ELO value].

2) The prediction engine actually only uses the ELO rankings as a baseline. From their, it uses actual match results, from previous matches vs wrestlers in the closest weight range. So, if wrestler A [ELO 1300] and wrestler B [ELO 1200] are run the through the prediction engine, then it will analyze, for wrestler A, all of the previous matches where the opponents are in a [ex] 1250 to 1350 ELO range. From their, it averages out the points scored [for], and points scored against (opponents points). The reverse this process (wrestler B vs wrestler A), then average those 2 sets of results. A little more math involved here, but that's the gyst of it.

3) Locations home/away/neutral are not factored in at all.

Note: I was planning on improving this by tweaking the previous head-to-head results weighting, but I did not get around to it this offseason. As it stands today, the Prediction Engine has not changed since last season.

4) Weighting more recent results - in theory, I believe this is technically taken care of the way the algorithm is written, since it's an accumulation/career. Obviously some fans would like to see that factored/weighted more heavily though...

 

[Wrassler17]

Good questions...

1) Weight change does not change the calculations/algorithms at all. It's assumed that the quality of wrestlers [and their ELO values] are basically "equal" across the different weights. So if a wrestler moves up or down, doesn't matter, when the rankings run, it's still just wrestler A [ELO value] vs wrestler B [ELO value].

2) The prediction engine actually only uses the ELO rankings as a baseline. From their, it uses actual match results, from previous matches vs wrestlers in the closest weight range. So, if wrestler A [ELO 1300] and wrestler B [ELO 1200] are run the through the prediction engine, then it will analyze, for wrestler A, all of the previous matches where the opponents are in a [ex] 1250 to 1350 ELO range. From their, it averages out the points scored [for], and points scored against (opponents points). The reverse this process (wrestler B vs wrestler A), then average those 2 sets of results. A little more math involved here, but that's the gyst of it.

3) Locations home/away/neutral are not factored in at all.

Note: I was planning on improving this by tweaking the previous head-to-head results weighting, but I did not get around to it this offseason. As it stands today, the Prediction Engine has not changed since last season.

4) Weighting more recent results - in theory, I believe this is technically taken care of the way the algorithm is written, since it's an accumulation/career. Obviously some fans would like to see that factored/weighted more heavily though...

Thanks for the answers. A lot going on behind the curtain obviously.