Ratings - please read before commenting - Archived

[conan_the_cribber]

Hi Cobwebcat (cwc)

I apologise for my outburst before. I promise to take my medicine and treat your comments with the respect they deserve. I disagree with you strongly, but that's no excuse for being rude. Some replies to your points below.

Firstly, before I answer your points, I assume that you are talking about the performance rankings. I spent some time advocating the 'who beat who' ratings, because I wanted an alternative to the predictive rankings. The results that Martin achieved in such a short time were impressive, but upon further reflection, I knew that they could also not be what I was looking for. Much to my surprise, Martin read the comment about Johnny Nelson and almost immediately came up with a BRILLIANT algorithm for the 'performance' rankings overnight. OK, now that I've explained that, on with some answers.

Conan, if you are truly a statitician you should know that you are talking nonsense.

There are many sports that have one on one and none of them use the ridiclous system that if you win one fight/game/match whatever then you are the best...no matter who you beat.

cwc, I am a statistician and hopefully I dont talk nonsense too often. You are correct that there are many sports where two teams or individuals compete against each other. In the majority of these sports, you are not thought of as the best, because you beat the best. This would be a result of the 'who beat who' rankings. However, this is NOT an immediate result of the 'performance' rankings. There can be an upset in the performance rankings and a third boxer will overtake the lead, as a measure of his lifetime achievements. To give an example. After their epic fights, Barrera and Morales where leading their weight class. Then Pacman came along and defeated Barrera somewhat surprisingly. Pacman's win, although significant, probably didnt earn him enough points in the 'performance' rankings to overtake Morales at the time.

Just because boxing has belts and proclaims someone "The Champ" because he "beat the man" does not mean this has any basis in statistical analysis. James Buster Douglas beat Tyson when Tyson had the clap and hardly trained, but did that make him the best fighter in the world?

cwc, I think you are mixing up various points of my critisms here, which is understandable, given your justifiable anger at my previous response. I will try and explain. There are two things at play in rankings at boxrec, one is the way that boxers are ranked traditionally and the other, which has long been pushed here, that the best rankings are the rankings that have the best predictability. Tradtionally, in boxing, when a top fighter gets beat surprisingly, then like in war, all spoils go to the victor. The victor gets the titles, gets the number one ranking, regardless of what people believe will happen in the rematch. Boxing is the hardest of all sports. Not only do you have to be superbly trained, you take a beating in every fight. The logic behind putting Buster Douglas as number one is pretty natural, the same as animals sorting out who's the leader of the pack. Buster Douglas is the new leader of the pack, till Mike Tyson can lick his wounds and come back and beat the shit out of the upstart. There is always the big question mark after a loss, whether the loser can come back or not, or has he been physically or mentally damaged. In this specific example, we saw the beginning of a pattern that would appear later, that if you stood up to Tyson and took him into the late rounds, then Tyson got into trouble. I make no claim that it is statistically valid to put a rank outsider like Douglas ahead of Tyson, in the sense of who will win the next match. However, I do maintain that it is essential to the traditions of boxing to do it.

I thought you had some interesting points re the predictive ranking but all I see now is you getting Martin to produce a ranking system that is flawed beyond belief and you cant take the same criticism of yours you were so quick to give out to other peoples.

This new "ranking" is based on a ladder system used by club squash players for god sakes.

Well again, it all depends what we are debating here. Like I said in the long opening statement, I wanted a somewhat more traditional alternative to the predictive rankings. Martin implemented this initially with the 'who beat who' algorithm, which as you correctly pointed out resembled a squash ladder. If you read every post thoroughly (what a lot of work) then you would've seen that I wanted an accelator factor that is different from a squash ladder, but Martin wanted to keep it simple. The NEW implementation from Martin, the 'performance' rankings has the best of both worlds and the one that I now wish to pursue. This is the algorithm, that I now wholeheartadly support. It addresses your central complaint, that if number 30 beats number 1, then he shouldn't be number 1 and addresses my central wishes, that a) it is a performance based rating that reflects the achievements of a fighter throughout their career and b) it respects the traditional boxing rankings with such things as A beats B, then as long as A and B have not fought since, then A is ranked ahead of B.

Any system is open to scrutiny. Yours is no different and is badly flawed. Just because it might seem popular, mainly because people actually want to see a form guide, does not make it accurate.

If you look at the performance rankings, I think you will be pressed to find a flaw. I could not find anything to complain about, and I'm known as a complaing dickhead. They are certainly not flawed in terms of predictabiliy, the statistic is very close to the predictive level and it's only one weekend old!! Imagine if we can tune it a little more. I see you around on the boards a bit and so you must know, the groundswell of opinion about Sechew Powell and Calvin brock being number ones. It will be popular to print these rankings, and they are accurate, with the potential to get even more accurate. I dont see what the problem is to print two rankings.

Martin/ JCS: I'm disappointed in you guys! Didn't getting rid of this A beats B nonsense improve your predictive rankings? Shouldn't it be disregarded in any meaningful ranking traditional or otherwise?

I cant speak for Martin or JCS. Getting rid of A beating B may have improved the predictive rankings, but if you read the posts, then you can understand why this was the wrong goal. As 'emile' (not me) cleverly pointed out, predictability should not be the goal. You should not deem upsets as 'wrong', especially if that means the predictive algorithm produces such wild outliers as "Briggs beats Brian Minto and becomes the number one heavy".

Conan: Congratulations on being very persuasive but you are a hypocrite. If you cant take it, don't dish it out. I'm afraid you can see other people's elephants but are terrible at spotting your own.

Oh, I am not a hypocrite. I am sometimes too pissed off about things that I shouldn't be, but I dont say one thing and then the opposite. The whole time I've been arguing for rankings that reflect the traditional ways that boxers have been ranked and I'm very happy that Martin has found a good way that is also very close to the predictive rankings. I don't think you will find many people in boxing that say, Spinks should be behing Kamazan, Paul Briggs behind Stipe Drews, Barerra ahead of Pacman, immediately after their last fight.

Having said all that I hope the three of you hit on something and prove me wrong but just as Conan stood on the outside pointing out possible flaws in the predictive system, some of which were good points, I believe he cant see the wood for the trees either now but hey, he's got his own way so who cares right?

Like I said at the beginning, I apologise for ma outburst and I hope you continue to participate in the debate. It is certainly not my way, as you imply, to try and exclude people from the debate. If I am persuasive, then I hope it is with logic and effort, rather than abuse.

All the best

conan

[conan_the_cribber]

Hi,

from Martin's list, here is the p4p.

conan


1 M Middleweight          1929 Ronald Wright
2 M Middleweight          1877 Jermain Taylor
3 M Middleweight          1678 Bernard Hopkins
4 M Super Featherweight  1659 Marco Antonio Barrera
5 M Welterweight          1657 Floyd Mayweather Jr
6 M Welterweight          1479 Ricky Hatton
7 M Super Featherweight  1241 Manny Pacquiao
8 M Welterweight          1226 Carlos Manuel Baldomir
9 M Light Welterweight    1164 Jose Luis Castillo
10 M Super Middleweight    1133 Joe Calzaghe
11 M Light Heavyweight    1132 Antonio Tarver
12 M Light Middleweight    1117 Cory Spinks
13 M Welterweight          1114 Antonio Margarito
14 M Light Middleweight    1114 Shane Mosley
15 M Bantamweight          1096 Rafael Marquez
16 M Heavyweight          1065 Wladimir Klitschko
17 M Light Heavyweight    1043 Glen Johnson
18 M Cruiserweight        1034 O'Neil Bell
19 M Heavyweight           971 Nikolay Valuev
20 M Light Welterweight    958 Isaac Hlatshwayo
21 M Featherweight         951 Chris John

Don't want tot start anything, but didn't you s ay yourself earlier you don't deal with p4p rankings because they're too difficult?

Nothing started Matty. I breezed over the ranking and saw what was produced. I think it was interesting that an excellent pfp could be generated as a by-product of the performance rankings. I consider this list totally viable. I'm surprised by two things only, Margarito too high and Chris Johns too low.

The reason I didn't want to ever get started on a p4p discussion is because you can never win. Nobody ever agrees on the criteria and my view is just as valid as the next mans. Who can logically say that Valuev is better than or worse than Chris John. One possible way, see above, is to measure their lifetime acheivements. That's why I made the passing comment about it being a good attempt.

Another way, favoured by others, is to imagine if both fighters took their skills and physical advantages and faced each other at a standard weight. In a match against Chris Johns, Valuev would become a very tall feather with porous defence, not unlike Coralles at jnr lightweight. Many people would favour Chis Johns in such a fantasy bout and rnk him higher. This is just as vaild.

Anyhow, the comment still stands, I aint really interested in p4p, but if I get it as a by-product, then why not use it.

conan

[conan_the_cribber]

Hi Martin,

some questions about the algorithm below.

... now the last tweek of achievement ratings - looks nearly perfect from a traditional perspective, I think ...

I agree

even predicition rate now up to 76.9 %

Great.

Rules

R1 initialisation
- every boxer starts with 50 points

R2 points for win/loss
- winner gets part of defeated opponent's points
- 40% of the difference of the opponent's rating minus 25% of own rating

Can you explain that with a formula or example? I'm not sure who's getting ranking points and who the minus 25% is being applied to.

- no points if rating of loser is lower than 25% of own rating

I think this might have to be revisited. If you look at the table below, you would see that if Mayweather beat Margarito, then he would get no points for the win. This would be unfair to Mayweather, as the win is a very significant one, against someone who is in most peoples top 20 pfp. Perhaps there could be a slightly staggered scale, that you get X points if someone is within 25% of your points and Y points if they are withing 50% of your points (where Y is obviously much less than X, the closer you get to the 50% barrier).

- loser loses these points
- if winner is still lower than winner, then loser is set to mean value of ratings and winner 1 point above

Ha. I think that was the John Shep algorithm, way back when. It would be interesting to know how often this happens for top fighters, say over 800 points at time of loss. A lot of the discussions about rankings are these rare out of the blue total upsets, but if they are 1 in 10000, then the debate loses it's intensity.

R3 inactivity

- points reduced by 33% of points above init value of 50 points for every 18 months of inactivity

Perhaps an area for tuning. 33% does sound harsh. It would be interesting to see how fighters over 400 PTS perform after a lay off (reduced form 600 pts). Perhaps it is possible to tune this a bit in isolation.

R4 missing quality of opponents

- points reduced by 25% of points above init vlaue of 50 points for every 18 month, where best faced opponent was lower than 25% of own rating

As above, I think this needs some tuning. Elite fighters like Mayweather could get absolutely caned because there is no-one in their division that is withing 25% of their rating. Using the example above, Mayweather could fight Margarito and it wouldn't count. I think the principle is good, but the algorithm is wrong for the elite fighters. Maybe you could come up with some averages for various positions in divisions e.g. the top 45 fighters is on average those fighters with 450 rating pts or the top 15 is on average aroudn 600 rating pts. I'm not sure how to get around this.

R5 division change
- points are transformed between divisions in proportion to square of division weight limits
... and here we go - bouts until August 3, 2006 included:

What weight did you use for heavweight? 200? Is the proportion for light middle to middle 154*154/160*160 = 92% of your rating is transferred. That sounds reasonable.

Code: Select all

M Welterweight         1657    0 Floyd Mayweather Jr
M Welterweight         1479    0 Ricky Hatton
M Welterweight         1226    0 Carlos Manuel Baldomir
M Welterweight         1114    0 Antonio Margarito
M Welterweight          643    0 Zab Judah
M Welterweight          632    0 Vivian Harris
M Welterweight          551    0 Arturo Gatti
M Welterweight          527    0 Sharmba Mitchell
M Welterweight          521    0 Juan Lazcano
M Welterweight          482    0 Kermit Cintron
M Welterweight          470    0 Oktay Urkal
M Welterweight          444    0 Carlos Quintana
M Welterweight          429    0 Paul Williams
M Welterweight          426    0 Cosme Rivera
M Welterweight          402    0 Joshua Clottey
M Welterweight          387    0 Antonio Pitalua
M Welterweight          381    0 Kevin Anderson
M Welterweight          380    0 Luis Collazo
M Welterweight          356    0 Oscar Diaz
M Welterweight          349    0 Joseph Makaringe

Looking at the table above an the 25% rule. Mayweather would gain nothing except for beating Hatton and he is lucky that by a mere 13 pts (1%) a victory over Baldi would help him. None of the fighters below Margarito help him at all. If Zab took a tough comeback fight, then neither Cintro, Urkal or Williams would help his ranking, which seems a bit harsh. I think there needs to a sliding scale there somehow.

conan

[JCS]

I think figures like the inactivity one can be tweaked... Maybe by using 15 months instead of 18 and by using a slightly lower percentage. Who knows.

Conan,

- winner gets part of defeated opponent's points
- 40% of the difference of the opponent's rating minus 25% of own rating
- loser loses these points


From what I understand.. if Judah fought and beat Cintron, the following would happen


Zab New = 648 + (482*.4 - 648*.25) = 678.8 (+30.8)
Cintron New = 482 - 30.8 = 451.2


Oh BTW, the reference weight for Heavy is 205. And the formula moving from Super Middle to Middle would be as follows


(154/160)^0.3 = 98.86% of rating (1.14% deducted)

[conan_the_cribber]

I think figures like the inactivity one can be tweaked... Maybe by using 15 months instead of 18 and by using a slightly lower percentage. Who knows.

Conan,

- winner gets part of defeated opponent's points
- 40% of the difference of the opponent's rating minus 25% of own rating
- loser loses these points


From what I understand.. if Judah fought and beat Cintron, the following would happen


Zab New = 648 + (482*.4 - 648*.25) = 678.8 (+30.8)
Cintron New = 482 - 30.8 = 451.2


Oh BTW, the reference weight for Heavy is 205. And the formula moving from Super Middle to Middle would be as follows


(154/160)^0.3 = 98.86% of rating (1.14% deducted)

I think Martin needs to confirm this. 75% of 648 is 486, which means 482 is too low for Zab to get any plus points for beating Cintron.

did you mean yuper middle or light middle. Either way, Martin said the product of ratios, which has little to do with the 0.3 used. But perhaps the forumla has changed.

If you used 205 for heavy and 200 for cruiser, then this division change is the least difference of all, instead of being the hardest. The number one cruiser Bell would waltz in at number one or two at heavy, which as Jirov and Toney and Gomez and many others have illustrated, is not easy to do.

conan

[computerrank]

... as I already analysed in a recent message - the rankings in the lower rankings were absolutely mixed up in the achievement ratings (and I have even big problems with the launching process in the current performance ratings).

I now found a new way to launch the boxers, thus all boxers are excellently ranked in the top, middle and lows. This is an absolutely new achievement.

There is no launching zone with unrated boxers left... rated from the start

... but this costs rediction rate at the moment - we now have 76.2 for these following achievement ratings - where I now have also integrated all known basic features from the current ratings in analogy:

R1 Inititialisation

- all boxers start with 0 points

R2 Win/Loss/Draw

- the winner gets the 10 points anyway
- the winner gets a percentage of the difference of his losers own rating minus 25% of his own rating as long as the loser was inside 25% of his own rating
- percentage is 20% for close win (SD, MD, close win on points)
- percentage is 40% else
-- A (1000) KO B (500):
--- new_A=1000+0.4*(500-1000/4)=1100
--- new_B = 400
- the loser losing the same amount of points
- if the winner is lower than the loser, the winner is set to the mean of both plus 1 point (only for tradition)

- in case of a draw the higher rated loses 20% of the difference, the lower gets it

R3 Inactivity

- a boxer loses 33% of his rating in case of every period of inactivity of 18 months

R4 Missing Qualiity of Opponents

- a boxer loses 25% of his rating in case of every period of missing an opponent of at least 25% of his own rating within 18 months

R5 Division Change

- the ratings are multiplied by the reciprocal factor the division weight limits (240 pounds for heavyweight)

R6 Home Advantage

- the ratings for the calculation of win/loss/draw are corrected for the home advantage
- if a boxer fights more than 2 bouts in the same country, he is regarded to have a home advantage of factor 1.2
- the rating values for calculation are multiplied by the square root of the ratio of the home advantage factors
-- example:
--- boxer A (1000) with home advantage gets had_A=1.2 KO
--- boxer_B (500) with home advantage gets had_B=1.0
--- new_A = 1000 + 0.4*(500*sqr(had_B/had_A)-1000/4*sqr(had_A/had_B))

Results (on top undisturbed - down very fine - you cannot see this):

Code: Select all

M Welterweight         1591    0 Floyd Mayweather Jr
M Welterweight         1307    0 Ricky Hatton
M Welterweight         1076    0 Antonio Margarito
M Welterweight          988    0 Carlos Manuel Baldomir
M Welterweight          615    0 Zab Judah
M Welterweight          579    0 Vivian Harris
M Welterweight          503    0 Sharmba Mitchell
M Welterweight          502    0 Juan Lazcano
M Welterweight          479    0 Oktay Urkal
M Welterweight          478    0 Arturo Gatti
M Welterweight          464    0 Kermit Cintron
M Welterweight          414    0 Carlos Quintana
M Welterweight          387    0 Paul Williams
M Welterweight          365    0 Cosme Rivera
M Welterweight          355    0 Joshua Clottey
M Welterweight          345    0 Michel Trabant
M Welterweight          345    0 Frederic Klose
M Welterweight          337    0 Ted Bami
M Welterweight          333    0 Kevin Anderson

M Super Middleweight   1238    0 Joe Calzaghe
M Super Middleweight    637    0 Markus Beyer
M Super Middleweight    613    0 Jeff Lacy
M Super Middleweight    581    0 Mikkel Kessler
M Super Middleweight    505    0 Carl Froch
M Super Middleweight    478    0 Librado Andrade
M Super Middleweight    393    0 Lucian Bute
M Super Middleweight    389    0 Anthony Mundine
M Super Middleweight    385    0 Robin Reid
M Super Middleweight    382    0 Rudy Markussen
M Super Middleweight    375    0 Mario Veit
M Super Middleweight    374    0 Scott Dann
M Super Middleweight    364    0 Mariano Natalio Carrera
M Super Middleweight    321    0 Jean Paul Mendy
M Super Middleweight    314    0 Denis Inkin
M Super Middleweight    291    0 Alejandro Berrio
M Super Middleweight    287    0 Robert Stieglitz
M Super Middleweight    287    0 Allan Green
M Super Middleweight    283    0 Sakio Bika
M Super Middleweight    274    0 Peter Manfredo Jr
M Super Middleweight    271    0 Jose Luis Herrera
M Super Middleweight    270    0 Francisco Antonio Mora

M Super Flyweight       566    0 Masamori Tokuyama
M Super Flyweight       466    0 Nobuo Nashiro
M Super Flyweight       391    0 Martin Castillo
M Super Flyweight       359    0 Luis Alberto Perez
M Super Flyweight       317    0 Simon Ramoni
M Super Flyweight       310    0 Masibulele Makepula
M Super Flyweight       309    0 Cristian Mijares
M Super Flyweight       288    0 Pramuansak Posuwan
M Super Flyweight       281    0 Alexander Munoz
M Super Flyweight       216    0 Takefumi Sakata
M Super Flyweight       214    0 Z Gorres
M Super Flyweight       214    0 Oscar Andrade
M Super Flyweight       201    0 Teppei Kikui
M Super Flyweight       181    0 Trash Nakanuma
M Super Flyweight       180    0 Mbwana Matumla
M Super Flyweight       179    0 Rosendo Alvarez
M Super Flyweight       155    0 Hiroyuki Hisataka

M Super Featherweight  1282    0 Marco Antonio Barrera
M Super Featherweight  1117    0 Manny Pacquiao
M Super Featherweight   694    0 Jorge Rodrigo Barrios
M Super Featherweight   549    0 Alex Arthur
M Super Featherweight   491    0 Humberto Soto
M Super Featherweight   483    0 Erik Morales
M Super Featherweight   482    0 Vicente Mosquera
M Super Featherweight   418    0 Rocky Juarez
M Super Featherweight   404    0 Oscar Larios
M Super Featherweight   401    0 Gairy St Clair
M Super Featherweight   400    0 Cassius Baloyi
M Super Featherweight   396    0 Bobby Pacquiao
M Super Featherweight   386    0 Robbie Peden
M Super Featherweight   356    0 Lakva Sim
M Super Featherweight   350    0 Mzonke Fana
M Super Featherweight   309    0 Janos Nagy
M Super Featherweight   288    0 Ali Funeka
M Super Featherweight   284    0 Edwin Valero

M Super Bantamweight    729    0 Israel Vazquez
M Super Bantamweight    410    0 Daniel Ponce De Leon
M Super Bantamweight    398    0 Celestino Caballero
M Super Bantamweight    369    0 Somsak Sithchatchawal
M Super Bantamweight    368    0 Mahyar Monshipour
M Super Bantamweight    344    0 Takalani Ndlovu
M Super Bantamweight    309    0 Wethya Sakmuangklang
M Super Bantamweight    309    0 Ricardo Castillo
M Super Bantamweight    275    0 Katsushige Kawashima
M Super Bantamweight    270    0 Michael Hunter
M Super Bantamweight    263    0 Saenghiran Lookbanyai
M Super Bantamweight    260    0 Agapito Sanchez
M Super Bantamweight    253    0 Julio Zarate
M Super Bantamweight    228    0 Sergio Manuel Medina
M Super Bantamweight    224    0 Esham Pickering
M Super Bantamweight    223    0 Ratanachai Sor Vorapin
M Super Bantamweight    211    0 Steve Molitor
M Super Bantamweight    210    0 Jose Nieves

M Minimumweight         392    0 Eagle Kyowa
M Minimumweight         362    0 Ivan Calderon
M Minimumweight         268    0 Satoshi Kogumazaka
M Minimumweight         226    0 Muhammad Rachman
M Minimumweight         187    0 Yutaka Niida
M Minimumweight         185    0 Rodel Mayol
M Minimumweight         184    0 Oleydong Sithsamerchai
M Minimumweight         172    0 Isaac Bustos
M Minimumweight         160    0 Katsunari Takayama
M Minimumweight         152    0 Pornsawan Kratingdaenggym
M Minimumweight         139    0 Nkosinathi Joyi
M Minimumweight         137    0 Marti Polii
M Minimumweight         135    0 Florante Condes
M Minimumweight         122    0 Teruo Misawa
M Minimumweight         110    0 Benjie Sorolla
M Minimumweight         108    0 Kenichi Horikawa
M Minimumweight         106    0 Gabriel Pumar
M Minimumweight         105    0 Toshiki Ogawa
M Minimumweight         103    0 Kazuma Kumada
M Minimumweight         100    0 Lorenzo Trejo
M Minimumweight          99    0 Pigmy Muangchaiyaphum
M Minimumweight          99    0 Daniel Reyes
M Minimumweight          96    0 Omar Soto

M Middleweight         1892    0 Bernard Hopkins
M Middleweight         1808    0 Ronald Wright
M Middleweight         1645    0 Jermain Taylor
M Middleweight          754    0 Arthur Abraham
M Middleweight          543    0 Sam Soliman
M Middleweight          526    0 Javier Castillejo
M Middleweight          504    0 Sergio Mora
M Middleweight          394    0 Raymond Joval
M Middleweight          394    0 Kelly Pavlik
M Middleweight          377    0 Felix Sturm
M Middleweight          364    0 David Alfonso Lopez
M Middleweight          353    0 Edison Miranda
M Middleweight          350    0 Evans Ashira
M Middleweight          341    0 Kingsley Ikeke
M Middleweight          333    0 Verno Phillips
M Middleweight          292    0 Enrique Ornelas
M Middleweight          291    0 Willie Gibbs
M Middleweight          288    0 Randy Griffin
M Middleweight          285    0 Amin Asikainen
M Middleweight          284    0 Sebastian Sylvester
M Middleweight          281    0 Howard Eastman

M Lightweight           817    0 Diego Corrales
M Lightweight           700    0 Juan Diaz
M Lightweight           649    0 Acelino Freitas
M Lightweight           616    0 Julio Diaz
M Lightweight           522    0 Jesus Chavez
M Lightweight           517    0 Zahir Raheem
M Lightweight           502    0 Joel Casamayor
M Lightweight           477    0 Joan Guzman
M Lightweight           460    0 Aldo Nazareno Rios
M Lightweight           348    0 Jose Armando Santa Cruz
M Lightweight           340    0 Graham Earl
M Lightweight           319    0 Ricky Quiles
M Lightweight           310    0 Fernando David Saucedo
M Lightweight           272    0 Matt Zegan
M Lightweight           267    0 Nate Campbell
M Lightweight           258    0 Mike Anchondo
M Lightweight           257    0 Michael Katsidis
M Lightweight           255    0 Koba Gogoladze
M Lightweight           255    0 Jairo Rafael Ramirez
M Lightweight           254    0 Almazbek Raiymkulov

M Light Welterweight   1126    0 Jose Luis Castillo
M Light Welterweight    828    0 Miguel Angel Cotto
M Light Welterweight    821    0 Isaac Hlatshwayo
M Light Welterweight    704    0 Junior Witter
M Light Welterweight    513    0 Carlos Maussa
M Light Welterweight    470    0 Demetrius Hopkins
M Light Welterweight    382    0 Juan Urango
M Light Welterweight    375    0 Herman Ngoudjo
M Light Welterweight    372    0 Lenny Daws
M Light Welterweight    370    0 Andreas Kotelnik
M Light Welterweight    327    0 Rolando Reyes
M Light Welterweight    323    0 Yodsanan Sor Nanthachai
M Light Welterweight    316    0 Cesar Rene Cuenca
M Light Welterweight    312    0 Juan Carlos Rodriguez
M Light Welterweight    300    0 Eleazar Contreras Jr
M Light Welterweight    296    0 Arturo Morua
M Light Welterweight    294    0 Naoufel Ben Rabah
M Light Welterweight    290    0 Fernando Angulo
M Light Welterweight    285    0 Sirimongkol Singwangcha
M Light Welterweight    279    0 Mike Arnaoutis
M Light Welterweight    279    0 Jonathan Thaxton
M Light Welterweight    278    0 DeMarcus Corley
M Light Welterweight    267    0 Paul Malignaggi
M Light Welterweight    266    0 Norio Kimura
M Light Welterweight    255    0 Lamont Peterson

M Light Middleweight   1130    0 Shane Mosley
M Light Middleweight    833    0 Cory Spinks
M Light Middleweight    831    0 Oscar De La Hoya
M Light Middleweight    782    0 Kassim Ouma
M Light Middleweight    755    0 Roman Karmazin
M Light Middleweight    617    0 Sergiy Dzinziruk
M Light Middleweight    604    0 Sergio Gabriel Martinez
M Light Middleweight    529    0 Rodney Jones
M Light Middleweight    481    0 Ricardo Mayorga
M Light Middleweight    477    0 Sechew Powell
M Light Middleweight    465    0 Joachim Alcine
M Light Middleweight    421    0 Fernando Vargas
M Light Middleweight    381    0 Vernon Forrest
M Light Middleweight    381    0 Michele Piccirillo
M Light Middleweight    359    0 Ike Quartey
M Light Middleweight    356    0 Reda Zam Zam
M Light Middleweight    355    0 Attila Kovacs
M Light Middleweight    334    0 Marco Antonio Rubio
M Light Middleweight    324    0 Wayne Alexander
M Light Middleweight    300    0 Ishe Smith
M Light Middleweight    288    0 Jose Antonio Rivera
M Light Middleweight    283    0 Crazy Kim
M Light Middleweight    275    0 Ossie Duran
M Light Middleweight    268    0 Javier Alberto Mamani
M Light Middleweight    268    0 Christophe Canclaux
M Light Middleweight    265    0 Jamie Moore

M Light Heavyweight    1137    0 Antonio Tarver
M Light Heavyweight     903    0 Glen Johnson
M Light Heavyweight     727    0 Roy Jones Jr
M Light Heavyweight     704    0 Zsolt Erdei
M Light Heavyweight     689    0 Fabrice Tiozzo
M Light Heavyweight     631    0 Tomasz Adamek
M Light Heavyweight     606    0 Clinton Woods
M Light Heavyweight     433    0 Stipe Drews
M Light Heavyweight     428    0 Paul Briggs
M Light Heavyweight     412    0 Chad Dawson
M Light Heavyweight     396    0 Julio Gonzalez
M Light Heavyweight     380    0 Montell Griffin
M Light Heavyweight     314    0 Hugo Hernan Garay
M Light Heavyweight     289    0 Adrian Diaconu
M Light Heavyweight     283    0 Prince Badi Ajamu
M Light Heavyweight     278    0 George Khalid Jones
M Light Heavyweight     264    0 Thomas Ulrich
M Light Heavyweight     250    0 Silvio Branco
M Light Heavyweight     239    0 Peter Haymer
M Light Heavyweight     230    0 Berry Butler
M Light Heavyweight     225    0 Reggie Johnson
M Light Heavyweight     203    0 Paul Murdoch
M Light Heavyweight     196    0 Mehdi Sahnoune
M Light Heavyweight     189    0 Eric Harding

M Light Flyweight       309    0 Roberto Vasquez
M Light Flyweight       301    0 Brian Viloria
M Light Flyweight       297    0 Ulises Solis
M Light Flyweight       262    0 Muvhuso Nedzanani
M Light Flyweight       214    0 Hugo Fidel Cazares
M Light Flyweight       212    0 Koki Kameda
M Light Flyweight       207    0 Edgar Sosa
M Light Flyweight       161    0 Kaichon Sor Vorapin
M Light Flyweight       161    0 Angky Angkota
M Light Flyweight       155    0 Wandee Singwangcha
M Light Flyweight       155    0 Nobuaki Masuda
M Light Flyweight       149    0 Omar Salado
M Light Flyweight       141    0 Nethra Sasiprapa
M Light Flyweight       141    0 Junichi Ebisuoka
M Light Flyweight       130    0 Along Denoy
M Light Flyweight       129    0 Will Grigsby
M Light Flyweight       125    0 Francisco Rosas
M Light Flyweight       122    0 Munetsugu Kayo

M Heavyweight           936    0 Wladimir Klitschko
M Heavyweight           729    0 Nikolay Valuev
M Heavyweight           728    0 John Ruiz
M Heavyweight           657    0 Chris Byrd
M Heavyweight           606    0 Serguei Lyakhovich
M Heavyweight           566    0 Calvin Brock
M Heavyweight           536    0 James Toney
M Heavyweight           527    0 Hasim Rahman
M Heavyweight           467    0 Lamon Brewster
M Heavyweight           466    0 Matt Skelton
M Heavyweight           410    0 Samuel Peter
M Heavyweight           370    0 Oleg Maskaev
M Heavyweight           355    0 Vladimir Virchis
M Heavyweight           354    0 DaVarryl Williamson
M Heavyweight           353    0 Ruslan Chagaev
M Heavyweight           350    0 Tony Thompson
M Heavyweight           326    0 Sinan Samil Sam
M Heavyweight           324    0 Danny Williams
M Heavyweight           305    0 Jameel McCline
M Heavyweight           303    0 Oliver McCall
M Heavyweight           301    0 Sultan Ibragimov
M Heavyweight           290    0 Ray Austin
M Heavyweight           290    0 Luan Krasniqi
M Heavyweight           288    0 Brian Minto
M Heavyweight           285    0 Monte Barrett
M Heavyweight           281    0 Tye Fields
M Heavyweight           279    0 Larry Donald

M Flyweight             645    0 Lorenzo Parra
M Flyweight             560    0 Vic Darchinyan
M Flyweight             549    0 Jorge Armando Arce
M Flyweight             445    0 Omar  Andres Narvaez
M Flyweight             423    0 Pongsaklek Wonjongkam
M Flyweight             261    0 Daisuke Naito
M Flyweight             230    0 Brahim Asloum
M Flyweight             218    0 Luis Alberto Lazarte
M Flyweight             216    0 Luis Maldonado
M Flyweight             216    0 Jose Lopez
M Flyweight             211    0 Denkaosan Kaovichit
M Flyweight             208    0 Juan Alberto Rosas
M Flyweight             191    0 Monelisi Mhikiza Myekeni
M Flyweight             187    0 Mark Johnson
M Flyweight             160    0 Hussein Hussein
M Flyweight             151    0 Shingo Yamaguchi
M Flyweight             151    0 Alejandro Hernandez
M Flyweight             144    0 Nelson Dieppa

M Featherweight         796    0 Chris John
M Featherweight         661    0 Scott Harrison
M Featherweight         646    0 Juan Manuel Marquez
M Featherweight         516    0 Thomas Mashaba
M Featherweight         434    0 Nicky Cook
M Featherweight         366    0 Orlando Salido
M Featherweight         360    0 Robert Guerrero
M Featherweight         342    0 Jorge Solis
M Featherweight         325    0 Steven Luevano
M Featherweight         316    0 In Jin Chi
M Featherweight         311    0 Toshiaki Nishioka
M Featherweight         270    0 Martin Honorio
M Featherweight         249    0 Eric Aiken
M Featherweight         247    0 Choi Tseveenpurev
M Featherweight         237    0 Hiroyuki Enoki
M Featherweight         231    0 Kazuhisa Watanabe
M Featherweight         230    0 Bernard Dunne
M Featherweight         226    0 Nedal Hussein

M Cruiserweight         944    0 O'Neil Bell
M Cruiserweight         665    0 Johnny Nelson
M Cruiserweight         531    0 Jean Marc Mormeck
M Cruiserweight         431    0 Enzo Maccarinelli
M Cruiserweight         418    0 Guillermo Jones
M Cruiserweight         405    0 Vadim Tokarev
M Cruiserweight         401    0 David Haye
M Cruiserweight         345    0 Steve Cunningham
M Cruiserweight         345    0 Grigory Drozd
M Cruiserweight         321    0 Krzysztof Wlodarczyk
M Cruiserweight         317    0 Emmanuel Nwodo
M Cruiserweight         279    0 Marco Huck
M Cruiserweight         262    0 Luis Andres Pineda
M Cruiserweight         259    0 Carl Thompson
M Cruiserweight         256    0 Rico Hoye
M Cruiserweight         253    0 Virgil Hill
M Cruiserweight         247    0 Matt Godfrey
M Cruiserweight         241    0 Mark Hobson
M Cruiserweight         235    0 Marcelo Fabian Dominguez
M Cruiserweight         235    0 Dale Brown

M Bantamweight          846    0 Rafael Marquez
M Bantamweight          593    0 Hozumi Hasegawa
M Bantamweight          577    0 Jhonny Gonzalez
M Bantamweight          395    0 Silence Mabuza
M Bantamweight          366    0 Wladimir Sidorenko
M Bantamweight          356    0 Simone Maludrottu
M Bantamweight          346    0 Sasha Bakhtin
M Bantamweight          320    0 Fernando Montiel
M Bantamweight          292    0 Irene Pacheco
M Bantamweight          289    0 Vusi Malinga
M Bantamweight          282    0 Veeraphol Sahaprom
M Bantamweight          251    0 Johnny Bredahl
M Bantamweight          245    0 Mauricio Martinez
M Bantamweight          242    0 Genaro Garcia
M Bantamweight          236    0 Michael Domingo
M Bantamweight          227    0 Ricardo Cordoba
M Bantamweight          219    0 Masayuki Arinaga
M Bantamweight          209    0 Simpiwe Vetyeka
M Bantamweight          207    0 Poonsawat Kratingdaenggym
M Bantamweight          207    0 Malcolm Tunacao
M Bantamweight          197    0 Samuel Lopez
M Bantamweight          195    0 Jose Navarro
M Bantamweight          192    0 Alejandro Valdez

[JCS]

Where is Klitschko in that latest set?

Also Martin, can you utilize this launching procedure for the current predictive ratings or would that not make sense?

[computerrank]

@ conan. cobwebcat, JCS, all interested

Boxrec Ratings are the Current Performance Ratings and will be for the future - optimized for prediciton rate - will be improved further.

Boxrec decided to publish a tradional (achievement) rating in addition for public convienience.

This Traditional/Achievement Rating will no be optimized for predicition rate (although this is a factor to be regarded) - but should represent the exspectations of the public (simplicity and following traditional rules).

So you should cleary see:

- the winner will be above the loser
- the winner should only be rewarded for opponents close to the winner or above him
- a boxer should be set back for inactivity
- a boxer should be set back for missing quality opponents

I think, this is implemented now in 2 options:

1. Traditional Ranking

- just for ranks
- you can not get higher than defeated opponent was
- very simple structure
- no launching problem
- prediction rate 62% - weak

2. Achievement Rating

- for points
- simple structure
- you can get higher than defeated opponent - in limits
- no launching problem
- prediciton rate 76% - good to excellent


We will publish one of the options - not quite sure what flavour ...

[computerrank]

Where is Klitschko in that latest set?

Also Martin, can you utilize this launching procedure for the current predictive ratings or would that not make sense?

Jason,

1. thanks for Klitschko - corrected ...

2. I use the launching system only for comparision of prediciton rate - see my lates message

[conan_the_cribber]

- the winner gets the 10 points anyway
- the winner gets a percentage of the difference of his losers own rating minus 25% of his own rating as long as the loser was inside 25% of his own rating
- percentage is 20% for close win (SD, MD, close win on points)
- percentage is 40% else
-- A (1000) KO B (500):
--- new_A=1000+0.4*(500-1000/4)=1100
--- new_B = 400
- the loser losing the same amount of points


Still getting my head around this.

(500-1000*25%) = 250
40% of 250 = 100
new ranking = 1000 + 100 = 1100.

But if it was two fighters of a 1000, then it would be

(1000-1000*25%) = -250
40% of -250 = -100
new ranking = 1000 - 100 = 900.

So the winner moves backwards? I think some other formula is used.

I think my examples with Margarito is invalid. 25% of 1000 is 250, so a fighter with 1000 can fight anyone above 250. Am I right?

conan

[computerrank]

conan,

I cannot see the problem with not available opponents for Mayweather etc.

They all have more than 10 matching opponents in their division ... and even if they would lose some points - they would stay on top ...

[computerrank]

- the winner gets the 10 points anyway
- the winner gets a percentage of the difference of his losers own rating minus 25% of his own rating as long as the loser was inside 25% of his own rating
- percentage is 20% for close win (SD, MD, close win on points)
- percentage is 40% else
-- A (1000) KO B (500):
--- new_A=1000+0.4*(500-1000/4)=1100
--- new_B = 400
- the loser losing the same amount of points


Still getting my head around this.

(500-1000*25%) = 250
40% of 250 = 100
new ranking = 1000 + 100 = 1100.

But if it was two fighters of a 1000, then it would be

(1000-1000*25%) = -250
40% of -250 = -100
new ranking = 1000 - 100 = 900.

So the winner moves backwards? I think some other formula is used.

I think my examples with Margarito is invalid. 25% of 1000 is 250, so a fighter with 1000 can fight anyone above 250. Am I right?

conan

... try again ... find your fault ...

[conan_the_cribber]

@ conan. cobwebcat, JCS, all interested

Boxrec Ratings are the Current Performance Ratings and will be for the future - optimized for prediciton rate - will be improved further.

Boxrec decided to publish a tradional (achievement) rating in addition for public convienience.

This Traditional/Achievement Rating will no be optimized for predicition rate (although this is a factor to be regarded) - but should represent the exspectations of the public (simplicity and following traditional rules).

So you should cleary see:

- the winner will be above the loser
- the winner should only be rewarded for opponents close to the winner or above him
- a boxer should be set back for inactivity
- a boxer should be set back for missing quality opponents

I think, this is implemented now in 2 options:

1. Traditional Ranking

- just for ranks
- you can not get higher than defeated opponent was
- very simple structure
- no launching problem
- prediction rate 62% - weak

2. Achievement Rating

- for points
- simple structure
- you can get higher than defeated opponent - in limits
- no launching problem
- prediciton rate 76% - good to excellent


We will publish one of the options - not quite sure what flavour ...

The achievement rankings are far, far superior. There will be a lot less debate about them than the tradional rankings and they indicate much more than a simple rank. Also the predictability is higher. I am very glad that John Shep has agreed to this and thanks for any contribution you may have had to it all.

conan

[conan_the_cribber]

- the winner gets the 10 points anyway
- the winner gets a percentage of the difference of his losers own rating minus 25% of his own rating as long as the loser was inside 25% of his own rating
- percentage is 20% for close win (SD, MD, close win on points)
- percentage is 40% else
-- A (1000) KO B (500):
--- new_A=1000+0.4*(500-1000/4)=1100
--- new_B = 400
- the loser losing the same amount of points


Still getting my head around this.

(500-1000*25%) = 250
40% of 250 = 100
new ranking = 1000 + 100 = 1100.

But if it was two fighters of a 1000, then it would be

(0-1000*25%) = -250 <<<****fixed****
40% of -250 = -100
new ranking = 1000 - 100 = 900.

So the winner moves backwards? I think some other formula is used.

I think my examples with Margarito is invalid. 25% of 1000 is 250, so a fighter with 1000 can fight anyone above 250. Am I right?

conan

... try again ... find your fault ...

I found my fault, but not yours.

[computerrank]

... the new Boxrec Achievement Ratings are launched ...

[TerribleTim2]

... the new Boxrec Achievement Ratings are launched ...

Thanks computerrank
The already look a lot better. Your work is appreciated.

[alarmakool]

All time rankings....what's happened?!! Are they upside down?!

You mean you don't think 0-0 Jack Clark is the greatest Heavy of all time? =P

Something is goofed w/ the system or they haven't been run for All time ratings yet.

[computerrank]

All time rankings....what's happened?!! Are they upside down?!

You mean you don't think 0-0 Jack Clark is the greatest Heavy of all time? =P

Something is goofed w/ the system or they haven't been run for All time ratings yet.

... just missed the output of the new all time ratings ... will come soon ...

[Lennox]

Virtually the IBO ratings copied.

[computerrank]

Virtually the IBO ratings copied.

Lennox,

Searching for improvement of the Boxrec ratings I also contacted IBO - but they didn't disclose their rules for comparision.

What is known ... does IBO have:

- first of all - published rules - they are making mysteries of their rules - basically open to all manipulations as not controllable by public?

- dynamic rating distance adjustion?
- inactivity rules?
- coverage of complete scope of boxers with homogeneous rules - even women ratings?
- homogenous rules for missing qualtity of opponents over complete scope of boxers?
- differentiated rewards for results?
- regarding home advantage?
- regarding division change?

This is a sophisticated rating developed on the base of best expertise of it's own, not owned by a private owner - for public - not for private purpose.

Both ratings are based on points - as it seems - but that's it.

Best regards
Martin

[JCS]

I'm really confused!!

Performance rankings

Achievement rankings

Traditional rankings

Predictive rankings

Is one to be shown or two?

Which one is showing now?

Is the one which Conan is refering to here (below) the one now showing?

You are correct that there are many sports where two teams or individuals compete against each other. In the majority of these sports, you are not thought of as the best, because you beat the best. This would be a result of the 'who beat who' rankings. However, this is NOT an immediate result of the 'performance' rankings. The NEW implementation from Martin, the 'performance' rankings has the best of both worlds and the one that I now wish to pursue. This is the algorithm, that I now wholeheartadly support. It addresses your central complaint, that if number 30 beats number 1, then he shouldn't be number 1 and addresses my central wishes, that a) it is a performance based rating that reflects the achievements of a fighter throughout their career and b) it respects the traditional boxing rankings with such things as A beats B, then as long as A and B have not fought since, then A is ranked ahead of B.



If the performance ranking is the one showing now, and is to be improved on through predictability, and everyone is happy, why are we going to have another "traditional" one too?

All I know is these are achievement ratings which seem to be 3/4 traditional, 1/4 predictive. So, prediction rate is down from what was on.

Dunno where the predictive ones went, I was under the impression we were to have both. I guess the powers that be gave into the complaining and now we simply mock what everyone else does.

[emile]

I haven't had a chance to go through them too much, but the new rankings look much better. Thanks Martin for listening to suggestions and coming up with something that is more satisfactory to everyone.

I'll read through the methodology and see if I have any additional suggestions, but I think its a good idea to not change the ranking system too often. I wouldn't try to tweak it too much for an additional percentage point of past predictive accuracy.

Thanks again

[computerrank]

... the new ratings will be the Boxrec Current Ratings

... All Time Ratings and P4P Ratings derived from them

... no other ratings/rankings needed

... they will be object to further improvement

[JCS]

RECOMMENDATION

Ok well here's one

Most people would say Caballero should be ranked ahead of Ponce De Leon

So we'll use Ponce de Leon in this example


Limit Ponce De Leon to the post-fight rating of Caballero until De Leon beats a boxer who is that rating or higher.. or perhaps we can be liberal and say a fighter within 5% of that post-fight rating.

[JCS]

So do the current achievement rankings meet the criteria below or not?

"It addresses your central complaint, that if number 30 beats number 1, then he shouldn't be number 1"

I think it depends on the point differences involved with those two fighters and the fighters around them. Viloria and Romero are probably good examples. I'd venture to say Viloria was #2 or #3 before that fight, now they're back to back 7 and 8