Ratings - please read before commenting

[SportsRatings]

I think the recent change was a big improvement. The only issue I see remaining is some guys who have not had a decent performance in a VERY long time are pretty highly ranked, sometimes over top guys in the division:

Yes, I've been complaining about this for a while. During Covid it was expanded to 3 years and I thought that was ideal.

Dropping it all together though.... has some of these guys lingering near the top and is a different kind of strange.

I wonder how the pool of points for top fighters has changed over time and if that is a contributing factor?

Can only imagine Covid depleted points, like WWI and especially WWII would have. Though WHR is supposed to "skip over" that kind of, depending on returning form.

I know some ppl really don't like "diminishing" points slowly based on inactivity, but something like that might be a compromise between an all-or-nothing solution for poor schedule strength

[jujigatame]

luis ortiz at #6

This one is particularly weird with Wilder ranked below him despite the 2 KOs.

[JCS]

luis ortiz at #6

This one is particularly weird with Wilder ranked below him despite the 2 KOs.

Those KOs are almost ancient history at this point...

[jujigatame]

luis ortiz at #6

This one is particularly weird with Wilder ranked below him despite the 2 KOs.

Those KOs are almost ancient history at this point...

That's true but it's not like either guy has done all that much that would effect their rankings since then. At least not relative to each other. Like Ortiz doesn't have some huge win that should have vaulted him up and over.

[Umbico]

I'm Benoit huber

https://boxrec.com/en/box-pro/831735

Yesterday I was 86 in cruiserweight ranking.
Today I'm 158 ?????

I beat juergen uldedaj less than 13 months ago.
Juergen uldedaj is now 45.

I'm not okay with the algorithm.
The people who know how it works has to explain me.

Thank you.

[computerrank]

I'm Benoit huber

https://boxrec.com/en/box-pro/831735

Yesterday I was 86 in cruiserweight ranking.
Today I'm 158 ?????

I beat juergen uldedaj less than 13 months ago.
Juergen uldedaj is now 45.

I'm not okay with the algorithm.
The people who know how it works has to explain me.

Thank you.

You already got your answer to your private email, where you sent the same complaint before.

[joshj909]

Ortiz reach a career high 10.85 points with his win against Martin in 2022 and lost a bit with his loss against Ruiz.

That also makes no sense. Why would a 2 year old win over Martin vault him up that high. Since losing to Kownacki 6 years ago, Martin beat 2 cans and Gerald Washington following his loss to Kownacki. I assume Washington's points came from beating Helenius, who beat Teper but Teper also had not beaten anyone of note. Prior to that, they all had recent losses.

Removing the points from the ratings really removes any transparency.

[JCS]

Oh my.. just noticed Nicholas Walters is #8 at 135..... This seems a bit far fetched..

[computerrank]

Oh my.. just noticed Nicholas Walters is #8 at 135..... This seems a bit far fetched..

Since the Donaire win he dropped from 31 to 20 points - but yes, his rating is very uncertain.

[Grey]

Why is a win he had a decade ago still pulling so much weight on his rating?

[computerrank]

Why is a win he had a decade ago still pulling so much weight on his rating?

Walter's current rating would be at 14 points with his win against Donaire lifted. The old super win against Donaire lets the system be uncertain about his real current rating. Maybe he still is much better and maybe he meanwhile is much worse. It is a 50 percent assumption by the system, backed by historical result evaluation of all bouts in the system. That could change, if he would lose against an opponent at his present rating level.

[JCS]

Hmm.. we all know where Walters realistically belongs.. and that's lower ranked.

So the 18 month rule was lifted... and I agree it was too harsh... The rule was 3 years during Covid.. that felt OK to me.. but if that is still impactful on the predictability in a negative way... surely there could be some protection against a fighter like Walters being in the Top 10.. What would a 5 year or even 7 year rule look like.. and how would that impact prediction rates?

I think you need SOMETHING to predict against extreme outliers. If Riddick Bowe or Ike Ibeabuchi fought again, I fear how high they'd be ranked in this latest version. HEY, how high WOULD they be ranked? :)

[computerrank]

Hmm.. we all know where Walters realistically belongs.. and that's lower ranked.

So the 18 month rule was lifted... and I agree it was too harsh... The rule was 3 years during Covid.. that felt OK to me.. but if that is still impactful on the predictability in a negative way... surely there could be some protection against a fighter like Walters being in the Top 10.. What would a 5 year or even 7 year rule look like.. and how would that impact prediction rates?

I think you need SOMETHING to predict against extreme outliers. If Riddick Bowe or Ike Ibeabuchi fought again, I fear how high they'd be ranked in this latest version. HEY, how high WOULD they be ranked? :)

Well, a cap for the ratings depending on the best recent achievement resulted in inconsistencies and so was dropped.
A depreciation of older results is not consistent with the WHR ratings, as all results are coupled from past and future by random uncertainty.
But WHR uses prior dummy results to bind a boxer's ratings to a mean start value. I now added additional dummy results to a boxers profile. For each real result a dummy result bound to the start mean with a weight in proportion to the time gone since the last bout. So the ratings may decrease in case of missing result quality and with ongoing time.
That improves the winner prediction ratio from 86.37 to a new high of 86.66 percent - and produces interesting rankings:

- Nicholas Walters would drop to #96
- Riddick Bowe and Ike Ibeabuchi would be #269 and #274 after a draw

- Fury would be #11

Code: Select all

  1 Heavyweight           659461 Anthony Joshua                 GB 7.012   2.7 1
  2 Heavyweight           659772 Oleksandr Usyk                 UA 6.850   2.9 1
  3 Heavyweight           861332 Jared Anderson                 US 6.121   2.8 1
  4 Heavyweight           613846 Joseph Parker                  NZ 5.260   2.1 1
  5 Heavyweight           813758 Filip Hrgovic                  HR 5.075   2.8 1
  6 Heavyweight           814335 Frank Sanchez                  CU 4.793   3.2 1
  7 Heavyweight           575332 Agit Kabayel                   DE 4.774   3.0 1
  8 Heavyweight           787430 Daniel Dubois                  GB 4.486   2.3 1
  9 Heavyweight           804221 Efe Ajagba                     NG 4.284   2.5 1
 10 Heavyweight           785893 Fabio Wardley                  GB 4.219   2.5 1
 11 Heavyweight           479205 Tyson Fury                     GB 3.941   3.4 1
 12 Heavyweight           681602 Martin Bakole                  CD 3.853   3.0 1
 13 Heavyweight           863084 Justis Huni                    AU 3.750   2.8 1
 14 Heavyweight           692149 Zhilei Zhang                   CN 3.666   2.5 1
 15 Heavyweight           811429 Joe Joyce                      GB 3.627   2.7 1

[JCS]

Fury at #11 will not win any new fans for the rankings... but logically, it makes some sense. Though I'm not sure Jared Anderson at #3 makes sense.

This update seems to over-reward "can crushing". Have a feeling it'd result in strange movement as well..

[margaret thatcher]

joshua #1, anderson #3, fury #11.......................kinda bizarre rankings tbh

[computerrank]

joshua #1, anderson #3, fury #11.......................kinda bizarre rankings tbh

I do not think, it is that bizarre.
- Fury dropped against Ngannou and did nothing recently
- Joshua recovered well from his close loss to Usyk with good wins
- Usyk did nothing remarkable recently
- Anderson was busy and successfull

[margaret thatcher]

i mean since usyk beat josh twice usyk beat one of the top 10 ranked fighters in those rankings, while joshua beat none of them

anderson's best career win is charles martin

fury was awful vs franny bbut he still got the W and has beaten several champs/contenders. now he's 8 spots below a guy whose best wins are martin, merhy, and i guess george arias?

[computerrank]

i mean since usyk beat josh twice usyk beat one of the top 10 ranked fighters in those rankings, while joshua beat none of them

anderson's best career win is charles martin

fury was awful vs franny bbut he still got the W and has beaten several champs/contenders. now he's 8 spots below a guy whose best wins are martin, merhy, and i guess george arias?

Usyk's win against Joshua is 20 months ago now(the winner above loser rule is valid 18 months).
Nobody knows, where Fury really is today. Let's see, what the bout against Usyk will bring.

[computerrank]

Hmm.. we all know where Walters realistically belongs.. and that's lower ranked.

So the 18 month rule was lifted... and I agree it was too harsh... The rule was 3 years during Covid.. that felt OK to me.. but if that is still impactful on the predictability in a negative way... surely there could be some protection against a fighter like Walters being in the Top 10.. What would a 5 year or even 7 year rule look like.. and how would that impact prediction rates?

I think you need SOMETHING to predict against extreme outliers. If Riddick Bowe or Ike Ibeabuchi fought again, I fear how high they'd be ranked in this latest version. HEY, how high WOULD they be ranked? :)

A WHR compatible prediction optimised damping of older results may produce a bit unexpected ratings. But a reduced damp factor seems to produce more expectable ratings in the same direction - and it still improves the prediction ratio from 86.37 to 86.62 percent:

- Nicholas Walters now at Lightweight #45 (before #9)
- after a draw Riddick Bowe #175 and Ike Ibeabuchi #191

Code: Select all

  1 Heavyweight           659772 Oleksandr Usyk                 UA 15.25   3.6 1
  2 Heavyweight           659461 Anthony Joshua                 GB 14.03   3.4 1
  3 Heavyweight           479205 Tyson Fury                     GB 8.547   4.6 1
  4 Heavyweight           613846 Joseph Parker                  NZ 7.664   2.2 1
  5 Heavyweight           861332 Jared Anderson                 US 7.153   3.3 1
  6 Heavyweight           787430 Daniel Dubois                  GB 6.661   2.6 1
  7 Heavyweight           575332 Agit Kabayel                   DE 6.504   3.9 1
  8 Heavyweight           813758 Filip Hrgovic                  HR 6.267   3.5 1
  9 Heavyweight           814335 Frank Sanchez                  CU 5.489   4.2 1
 10 Heavyweight           692149 Zhilei Zhang                   CN 5.154   2.8 1

Code: Select all

  1 Cruiserweight         811275 Chris Billam-Smith             GB 16.79   2.8 1
  2 Cruiserweight         509666 Gilberto Ramirez               MX 15.18   3.2 1
  3 Cruiserweight         719823 Jai Opetaia                    AU 12.84   3.6 1
  4 Cruiserweight         769396 Richard Riakporhe              GB 10.69   3.8 1
  5 Cruiserweight         789060 Lawrence Okolie                GB 10.67   2.5 1
  6 Cruiserweight         773151 Ryan Rozicki                   CA 8.781   2.7 1
  7 Cruiserweight         650083 Michal Cieslak                 PL 7.907   3.7 1
  8 Cruiserweight         571652 Norair Mikaeljan               AM 6.354   4.4 1
  9 Cruiserweight         358968 Mateusz Masternak              PL 6.236   3.3 1
 10 Cruiserweight         438253 Mike Perez                     CU 5.821   5.1 1

Code: Select all

  1 Super Middleweight    348759 Saul Alvarez                   MX 65.89   3.1 1
  2 Super Middleweight    672259 David Benavidez                US 39.96   3.3 1
  3 Super Middleweight    659924 Jaime Munguia                  MX 32.55   3.2 1
  4 Super Middleweight    785657 Christian Mbilli               FR 23.53   3.4 1
  5 Super Middleweight    852450 David Morrell                  CU 19.83   3.3 1
  6 Super Middleweight    760625 Edgar Berlanga                 PR 17.97   3.4 1
  7 Super Middleweight    802856 Kevin Lele Sadjo               FR 17.57   3.5 1
  8 Super Middleweight    865105 Diego Pacheco                  US 16.92   3.1 1
  9 Super Middleweight    850339 Osleys Iglesias                CU 16.47   3.2 1
 10 Super Middleweight    663953 Erik Bazinyan                  CA 14.72   3.9 1

Code: Select all

  1 Middleweight          774417 Zhanibek Alimkhanuly           KZ 32.07   3.1 1
  2 Middleweight          471634 Jermall Charlo                 US 28.28   6.0 1
  3 Middleweight          588468 Chris Eubank Jr                GB 26.07   3.8 1
  4 Middleweight          778622 Meiirim Nursultanov            KZ 25.07   3.2 1
  5 Middleweight          808683 Hamzah Sheeraz                 GB 23.56   3.4 1
  6 Middleweight          457231 Erislandy Lara                 US 22.35   4.5 1
  7 Middleweight          466535 Liam Smith                     GB 17.75   3.0 1
  8 Middleweight          723287 Carlos Adames                  DO 17.32   3.5 1
  9 Middleweight          861505 Austin Williams                US 15.25   3.4 1
 10 Middleweight          927452 Elijah Garcia                  US 13.09   2.9 1

Code: Select all

  1 Light Middleweight    766727 Vergil Ortiz Jr                US 32.67   4.5 1
  2 Light Middleweight    851173 Israil Madrimov                UZ 26.16   3.8 1
  3 Light Middleweight    773847 Sebastian Fundora              US 25.74   2.8 1
  4 Light Middleweight    783149 Tim Tszyu                      AU 24.18   2.6 1
  5 Light Middleweight    788164 Serhii Bohachuk                UA 18.76   3.2 1
  6 Light Middleweight    433135 Jermell Charlo                 US 18.42   4.3 1
  7 Light Middleweight    689777 Bakhram Murtazaliev            RU 17.86   4.9 1
  8 Light Middleweight    670934 Erickson Lubin                 US 16.95   2.9 1
  9 Light Middleweight    835962 Slawa Spomer                   DE 16.74   3.3 1
 10 Light Middleweight    791994 Charles Conwell                US 15.73   4.6 1

[JCS]

Those rankings seem pretty reasonable...

[computerrank]

BoxRec Ratings WHR.1.3.2 are live now.

There is a new damping factor for old results. Winner prediction ratio improved from 86.37 to 86.58 percent.

[Zielas147]

Hello,
I purchased new boxrec+ for fans (basic).
It says that I should be able to "See the win-probabliity on bouts". How can I access this feature?

[computerrank]

Hello,
I purchased new boxrec+ for fans (basic).
It says that I should be able to "See the win-probabliity on bouts". How can I access this feature?

You can either click on predict on the red tool bar and then select a boxer and an opponent.

Or you can click on an event in the schedule and there will be shown the win probabilities of each bout.

[Zielas147]

Thank you, you are correct. It seems like a very nice function for stat geeks, like myself.
Glad to see some numbers, that describe fights, return.
It would be amazing, if we could see pre-fight probabilities for past fights. I imagine it would take a lot of work to make that happen.
My dream is to be able to sort all fights in Your register. For example: biggest upset in terms of pre-fight odds, biggest mismatch, etc.
Good luck with crunching numbers for algorithm.

Hello,
I purchased new boxrec+ for fans (basic).
It says that I should be able to "See the win-probabliity on bouts". How can I access this feature?

You can either click on predict on the red tool bar and then select a boxer and an opponent.

Or you can click on an event in the schedule and there will be shown the win probabilities of each bout.

[gregregegg]

What changed in the algorithm that absolutely killed the heavyweights?

-wilder vs Hellenise is a 1 star fight..
-parker vs wilder, zhang, AJ, Whyte, chisora all 2 star fights.

but,

-Connor ben vs algeri, dobs, van heerden 3 star
-kongo vs mckinson or ekow 3 star.


There are legit world world world class heavyweight fights that the algorithim is evaluating as less than brittish levle welterweight fights.