ADVERTISEMENT

WrestleStat Rankings | Week 9 2022

Those "some wins" prior to Sunday were 2 wins over the same guy who himself is now only #16. I don't see any way that ranking was possible based on the results. But after Sunday looks like it was a good prediction!

I do agree with some of the comments that the algorithm could likely be improved by not putting as much weight on results from 3+ years ago (for example, still having Desanto ahead of RBY). I also think NCAA tournament matches should have more weight (for example, still having Kemerer ahead of Starocci).

Obviously there is not much "usefulness" to ANY rankings, but in the conversation of rankings as a whole, having a set of rankings available that have literally no subjectivity is pretty cool. So I don't see the issue with making suggestions...obviously they don't have to take them. I would also say having the rankings be "more accurate" would be helpful considering how awesome the rest of the site is...then for example if you are already looking at the common opponents in a dual comparison there may not be as much need to flip back to Intermat or whatever to check out the rankings.
Here's the thing about his results. He has been perfect at (1) beating the kids he should beat, (2) losing to the kids he really should lose to, and (3) getting some wins over some kids that he should kinda lose to.
Based on the algorithm,
1 doesn't help much
2 doesn't hurt much
3 helps quite a bit

His two wins over Surtin, one over Smith at 34 and Witcraft at 60, all came when he was between 69th and 83rd. So, according to the math, those wins helped a lot more than other wins.

Also, I think another thing people fail to realize is the math is based on ELO score, not rank. So, as a 125 with 1200 ELO points, beating #33 with 1381 ELO points is dramatically different than the math at 133, where #33 has 1475 points.

The ELO points are also important because a bunch of kids can be bunched together, with only 5 or 10 points separating three or four wrestlers.
 
Observations RE Wrestlestat:
1. If you want to inject human interference to manually correct what seem like obvious errors......it is no longer an unbiased ranking.
2. If you change the algorithm to "correct" the perceived obvious errors.......it will change all calculations and will almost assuredly just move the "Obvious errors" to different matchups. Basically kicking the error(s) to a different location
3. After reading all these suggestions/complaints.......I am going to Bitch-Slap my toaster* until further notice :)

* YES.....that's what they are calling it now :)
Stop being so reasonable and please apologize to your toaster. :cool:
 
I would like to know how career matches applies. So if he lost all 150 what then. I'm not saying it shouldn't apply I just want to know how it applies to this f d up algorithm they use. Once again if your algorithm is broke. Fix it
Just saw that you had posted this kind of as a question.

Each wrestler starts out with 1200 ELO points. While losing every match would drop that score, I don't think it would drop by much, for two reasons. First, a wrestler with an ELO score of 1200 isn't expected to beat any top wrestlers, so a loss to them wouldn't hurt their score much. And, a wrestler with a 1200 ELO isn't expected to win or lose to another wrestler with an ELO score of 1200, so a loss doesn't hurt that much, but would drop the score a little.
 
Observations RE Wrestlestat:
1. If you want to inject human interference to manually correct what seem like obvious errors......it is no longer an unbiased ranking.
2. If you change the algorithm to "correct" the perceived obvious errors.......it will change all calculations and will almost assuredly just move the "Obvious errors" to different matchups. Basically kicking the error(s) to a different location
3. After reading all these suggestions/complaints.......I am going to Bitch-Slap my toaster* until further notice :)

* YES.....that's what they are calling it now :)
If only someone had thought of #2 before. 🙄
 
Just saw that you had posted this kind of as a question.

Each wrestler starts out with 1200 ELO points. While losing every match would drop that score, I don't think it would drop by much, for two reasons. First, a wrestler with an ELO score of 1200 isn't expected to beat any top wrestlers, so a loss to them wouldn't hurt their score much. And, a wrestler with a 1200 ELO isn't expected to win or lose to another wrestler with an ELO score of 1200, so a loss doesn't hurt that much, but would drop the score a little.
That makes sense. I guess there is a lot more that goes into than I thought. You math guys got it going on lol. In any event thanks for all the hard work in the end it is greatly appreciated
 
Here's the thing about his results. He has been perfect at (1) beating the kids he should beat, (2) losing to the kids he really should lose to, and (3) getting some wins over some kids that he should kinda lose to.
Based on the algorithm,
1 doesn't help much
2 doesn't hurt much
3 helps quite a bit

His two wins over Surtin, one over Smith at 34 and Witcraft at 60, all came when he was between 69th and 83rd. So, according to the math, those wins helped a lot more than other wins.

Also, I think another thing people fail to realize is the math is based on ELO score, not rank. So, as a 125 with 1200 ELO points, beating #33 with 1381 ELO points is dramatically different than the math at 133, where #33 has 1475 points.

The ELO points are also important because a bunch of kids can be bunched together, with only 5 or 10 points separating three or four wrestlers.

I like to consider myself a math guy, but this doesn’t add up. Beating #’s 16, 33, and 60 shouldn’t get you to #7 when those are literally your only notable wins. Not to mention #33 is a first year starter in the Socon without any notable wins of his own, and #60 is a backup (although he at least had a couple decent wins back in 2020), so calling those guys he “should kinda lose to” is a stretch IMO.

One other aspect that could likely be improved is I think bonus points are weighted too heavily. Seems like he was rewarded for bonusing a bunch of backups at 2 opens.
 
I like to consider myself a math guy, but this doesn’t add up. Beating #’s 16, 33, and 60 shouldn’t get you to #7 when those are literally your only notable wins. Not to mention #33 is a first year starter in the Socon without any notable wins of his own, and #60 is a backup (although he at least had a couple decent wins back in 2020), so calling those guys he “should kinda lose to” is a stretch IMO.

One other aspect that could likely be improved is I think bonus points are weighted too heavily. Seems like he was rewarded for bonusing a bunch of backups at 2 opens.
I think you're getting too hung up on the rank, and not considering the ELO scores of those he beat, and what's happening to the other wrestlers' scores.

And, "should kinda lose to" has nothing to do with whether he should or shouldn't lose to them. It has everything to do with the current ELO score at the time of the match. If a wrestler with an ELO score of 1200 wrestles a guy with an ELO score of 1600, the 1200 should lose, based strictly on that score. But, the beauty of ELO is that when 1200 beats 1600, he is rewarded much more significantly than if he beat 1205. You appear to be allowing your bias, based on what you know of the wrestlers histories/circumstances to guide you. Straight math doesn't allow for bias.

But, the bigger thing is this... you would absolutely be correct about , if WS used purely ELO math. But, get this...Because of these very types of complaints in the early WS years, WS adjusted the algorithm to get young wrestlers moving towards their eventual "correct" place more quickly. Let me see if I can think of a young phenom wrestler from 2013-14 that a certain fan base would be disgruntled about not moving up the board quickly enough. Hmm, how about Zain Retherford? This board was likely apoplectic when Zain beat Steiber, but was still 55th on WS. So, WS added a multiplier to the algorithm, which helped get wrestlers to their likely "sweet spot" more quickly.

Finally, you might be right about the bonus point thing, so let's add that to the list.
-Freshmen need to get to equilibrium more quickly
- Add a recency bias
- Adjust bonus win multiplier by opponent quality.

Do you truly believe that you can adjust for the three of those, and still not come up with examples of "wrongness?"
 
I’m curious, does the algorithm factor in the postseason with a multiplier or is every match treated as an equal data point?

I love, REALLY LOVE WS, and have also criticized some of its strange rankings. I personally would weight NCAA matches heavily, and conference tournament ones more than regular season matches.

The “flaws” that stand out are things like RBY being an undefeated defending Nat Champ, yet being ranked 3rd behind 2 guys he beat. There has to be some way to compensate for these situations.

Either way, keep up the good work. WS is an amazing asset.
 
  • Like
Reactions: GogglesPaizano
I think you're getting too hung up on the rank, and not considering the ELO scores of those he beat, and what's happening to the other wrestlers' scores.

And, "should kinda lose to" has nothing to do with whether he should or shouldn't lose to them. It has everything to do with the current ELO score at the time of the match. If a wrestler with an ELO score of 1200 wrestles a guy with an ELO score of 1600, the 1200 should lose, based strictly on that score. But, the beauty of ELO is that when 1200 beats 1600, he is rewarded much more significantly than if he beat 1205. You appear to be allowing your bias, based on what you know of the wrestlers histories/circumstances to guide you. Straight math doesn't allow for bias.

But, the bigger thing is this... you would absolutely be correct about , if WS used purely ELO math. But, get this...Because of these very types of complaints in the early WS years, WS adjusted the algorithm to get young wrestlers moving towards their eventual "correct" place more quickly. Let me see if I can think of a young phenom wrestler from 2013-14 that a certain fan base would be disgruntled about not moving up the board quickly enough. Hmm, how about Zain Retherford? This board was likely apoplectic when Zain beat Steiber, but was still 55th on WS. So, WS added a multiplier to the algorithm, which helped get wrestlers to their likely "sweet spot" more quickly.

Finally, you might be right about the bonus point thing, so let's add that to the list.
-Freshmen need to get to equilibrium more quickly
- Add a recency bias
- Adjust bonus win multiplier by opponent quality.

Do you truly believe that you can adjust for the three of those, and still not come up with examples of "wrongness?"

I think I am considering the ELO scores of those he beat. That’s my point. Prior to Sunday, he had only beaten one guy with a decent ELO score (twice). I guess one thing in his favor is that 125 has a ton of parity so a lot of guys probably have losses to guys they “shouldn’t” have lost to, and I get that you’re saying Ayala wasn’t really punished for losing 3x since he “shouldn’t” have won anyway. But I still don’t see how that catapults him ahead of guys like Cardinale (AA last year and undefeated this year) or Mastrogiovanni (undefeated this year beating McKee).

In terms of a “recency bias,” I’d just say I think the current season should be worth more than prior seasons, and they shouldn’t really count results from 3+ years ago. I also think NCAA tournament matches should be weighted more than others (in the same season).

I’m sure it won’t be perfect, but these seem to be overall trends so yes I do think making those tweaks could improve the rankings.
 
… 2. If you change the algorithm to "correct" the perceived obvious errors.......it will change all calculations and will almost assuredly just move the "Obvious errors" to different matchups. Basically kicking the error(s) to a different location …
Changing the algorithm will assuredly change the number and severity of “obvious errors”. Andegre can change it 20 different ways and keep the best way. That would be called learning. If a machine does it, it would be called machine learning. :) #beoptimistic!
 
I think I am considering the ELO scores of those he beat. That’s my point. Prior to Sunday, he had only beaten one guy with a decent ELO score (twice). I guess one thing in his favor is that 125 has a ton of parity so a lot of guys probably have losses to guys they “shouldn’t” have lost to, and I get that you’re saying Ayala wasn’t really punished for losing 3x since he “shouldn’t” have won anyway. But I still don’t see how that catapults him ahead of guys like Cardinale (AA last year and undefeated this year) or Mastrogiovanni (undefeated this year beating McKee).

In terms of a “recency bias,” I’d just say I think the current season should be worth more than prior seasons, and they shouldn’t really count results from 3+ years ago. I also think NCAA tournament matches should be weighted more than others (in the same season).

I’m sure it won’t be perfect, but these seem to be overall trends so yes I do think making those tweaks could improve the rankings.
I think it has to do with the new guy multiplier. And, whether the multiplier is "right" or "wrong" is going to take a lot of time to determine rightness or wrongness.
 
So, this morning I learned that @Sportfan2017 is a plagiarizer.

Every morning I yell at my toaster: “why can’t you be a coffee brewer!”

Interesting reading from 4 years ago...
 
I’d love to hear the outcome of pawrestler yelling at his dishwasher.
 
What do you think its usefulness is? Serious question. What is the usefulness of, say, Intermat's rankings? Betting? Playing KYPSW? As a conversation piece? As a means of thoroughly ticking off aggrieved wrestling fans because their guy was snubbed by a computer?

This really is not that big of a deal and it is something nearly every business faces daily. I work in IT and virtually any model used for analytics is adjusted empirically as results are achieved and studied.
One continually reviews the results and adjusts the model, rinse and repeat, eventually progressing towards achieving optimization. It aint rocket science.
 
This really is not that big of a deal and it is something nearly every business faces daily. I work in IT and virtually any model used for analytics is adjusted empirically as results are achieved and studied.
One continually reviews the results and adjusts the model, rinse and repeat, eventually progressing towards achieving optimization. It aint rocket science.
I'm not sure you answered what usefulness these rankings have to you (or anyone else). Do models and algorithms affect business decisions that affect the bottom line? Yes. Do WS rankings?

And, I'm not sure the boys at WS are making a living at this. If they're living off the tip jar, they're likely going hungry.
 
I’d buy you that beer in front of her while you tell her that..
I’m looking forward to the explanation of one of your dishwashers.
 
  • Like
Reactions: pawrestlersintn
I think it has to do with the new guy multiplier. And, whether the multiplier is "right" or "wrong" is going to take a lot of time to determine rightness or wrongness.

Yeah I can give them a pass on the "new guy" thing, although I think it's to be expected that freshmen would be LOWER than they probably "should" be due to a lack of results (especially in cases like in the old thread of RBY who didn't have much in the way of notable wins early in his freshman season) - this same thing usually happens with "subjective" rankings as well of course. The Olivieri example seems to show that the Ayala situation was not a fluke - I guess this is an example of them trying to "improve" the algorithm but it probably went too far in the opposite direction. I would think there should be some way to put freshmen equal/above guys they have beaten, but I don't see how having freshmen jump undefeated returning AA's who were already ranked above the guy the freshman beat makes any sense.

Ultimately, I think that goes back to my main recommendation which is that there doesn't seem to be enough weight on NCAA tournament results. To start the year, there were 6 guys who were the returning champ at the same weight that were not ranked #1, 4 of them were behind the guys they beat at NCAA's, 3 of them were ranked #3 or lower (Starocci was and remains #4! and I"m not even a PSU fan), and one (Carr) is still #3 despite one of the guys ahead of him taking a loss this year. I don't think it's rude to say that's an indication that the algorithm can likely be improved.

Edit: I'm not a big tech guy so not sure I'd be much help on algorithms, but I do do some data analysis as part of my job, would certainly be willing to volunteer some time to help wrestlestat in any way I could, I think it's the best wrestling site going today.
 
This really is not that big of a deal and it is something nearly every business faces daily. I work in IT and virtually any model used for analytics is adjusted empirically as results are achieved and studied.
One continually reviews the results and adjusts the model, rinse and repeat, eventually progressing towards achieving optimization. It aint rocket science.
So this is how my computer knows i could use penis growth pills? Somebody has some good analytics. Maybe @andegre should hire that dude.
 
  • Haha
Reactions: Dogwelder
So, this morning I learned that @Sportfan2017 is a plagiarizer.
So, this morning we all learned that you have absolutely no idea what plagiarism is:

"I am going to Bitch-Slap my toaster"
“why can’t you be a coffee brewer!”

Despite every single word being different and one being a question......they are identical in every way. Keep up the awesome work..... :)
 
So, this morning we all learned that you have absolutely no idea what plagiarism is:

"I am going to Bitch-Slap my toaster"
“why can’t you be a coffee brewer!”

Despite every single word being different and one being a question......they are identical in every way. Keep up the awesome work..... :)
Looks to me like you used his idea.

pla·gia·rism
/ˈplājəˌrizəm/
Learn to pronounce
noun
the practice of taking someone else's work or ideas and passing them off as one's own.
"there were accusations of plagiarism"
 
Ha ha. If you were riffing on my old post, Sportsfan, thank you for remembering it! I had forgotten all about it. Otherwise, it’s a coincidence, and there is nothing genius about the word “toaster”, and who is going to seek glory with it? :)
 
  • Like
Reactions: Sportfan2017
Looks to me like you used his idea.

pla·gia·rism
/ˈplājəˌrizəm/
Learn to pronounce
noun
the practice of taking someone else's work or ideas and passing them off as one's own.
"there were accusations of plagiarism"
Yes.....threatening harm to something
Questioning/wishing why that one thing wasn't something else are the same idea in your world.
"I am going to bitch slap my brother"
"Why can't my older brother be my younger sister"?
SAME IDEA??
I believe we have identified the problem Dr. :)
 
Last edited:
  • Like
Reactions: Dogwelder
Looks to me like you used his idea.

pla·gia·rism
/ˈplājəˌrizəm/
Learn to pronounce
noun
the practice of taking someone else's work or ideas and passing them off as one's own.
"there were accusations of plagiarism"
Let’s please leave the play-dumb-to-win-a-contrived-point-as-if-it-would-be-a-huge-victory tactic to the COVID thread and any (other) let’s-destroy-democracy-and-rationality threads. Thank you.
 
Last edited:
  • Like
Reactions: Sportfan2017
Dang, walk away for a day, and the thread triples in size. pawrestlersintn is still fighting the good fight, the thread has stayed between the ditches (i.e. out of the gutter!), and District four has offered half an apology (j/k Df). All of that is impressive.

I love WrestleStat too. Like pa, I got wrapped up in the ELO modeling early, so I mostly understand what comes out of the black box when all the data is entered. It doesn't take a rocket scientist to figure the main guy's name behind the model, though andegre and obrats are screen names here and other places. I'll call him Dan, and he tries to keep up on twitter and other places when questions are asked. As mentioned earlier he also shows up here for reasonable discussion.

I don't look at the rankings as much as the wrestler info. Other than having PSU wrestler data available on the PSWC website, it's the only place with a phenomenal amount of current and past data of ALL D1 wrestlers since and including the 2014 season.
 
Ha ha. If you were riffing on my old post, Sportsfan, thank you for remembering it! I had forgotten all about it. Otherwise, it’s a coincidence, and there is nothing genius about the word “toaster”, and who is going to seek glory with it? :)
I certainly didn't remember it. I barely remember yesterday. You should be proud to know that someone rifles through your old posts to find something to use in an absurd comparison in an attempt to insult others. YOU ARE
 
  • Haha
Reactions: Dogwelder
giphy.gif
 
ADVERTISEMENT