BoxRec

computerrank wrote: ↑12 Oct 2019, 15:38 http://151.boxrec.com/~martin/ratings_at.php

While I appreciate Henry Armstrong's long overdue recognition as an all time great, too many unknown fighters from the past are clogging up top spots... While actual undeniable greats are way too low...

-Jack Carroll is the all time #2 best Welterweight
-Marvelous Marvin Hagler #19 best Middleweight
-Bob Pastor is a top 10 all time Heavyweight
-Pernell Whitaker is... Just, no.

I could go on, but basically there are too many indefensible rankings

It needs more tinkering, it'll get there bro, keep up the good work.

computerrank wrote: ↑12 Oct 2019, 15:38
http://151.boxrec.com/~martin/ratings.php

Much like before, nothing egregiously out of place. Sure there's some rankings I disagree with but nothing worth arguing over... Accept maybe Caleb Plant... but overall, looking good champ

Manrae wrote: ↑12 Oct 2019, 18:12

computerrank wrote: ↑12 Oct 2019, 15:38 http://151.boxrec.com/~martin/ratings_at.php

While I appreciate Henry Armstrong's long overdue recognition as an all time great, too many unknown fighters from the past are clogging up top spots... While actual undeniable greats are way too low...

-Jack Carroll is the all time #2 best Welterweight
-Marvelous Marvin Hagler #19 best Middleweight
-Bob Pastor is a top 10 all time Heavyweight
-Pernell Whitaker is... Just, no.

I could go on, but basically there are too many indefensible rankings

It needs more tinkering, it'll get there bro, keep up the good work.

The plain elevation of ancient and former boxers is too simple. I will look into statistics behind.

computerrank wrote: ↑13 Oct 2019, 02:52

Manrae wrote: ↑12 Oct 2019, 18:12

While I appreciate Henry Armstrong's long overdue recognition as an all time great, too many unknown fighters from the past are clogging up top spots... While actual undeniable greats are way too low...

-Jack Carroll is the all time #2 best Welterweight
-Marvelous Marvin Hagler #19 best Middleweight
-Bob Pastor is a top 10 all time Heavyweight
-Pernell Whitaker is... Just, no.

I could go on, but basically there are too many indefensible rankings

It needs more tinkering, it'll get there bro, keep up the good work.

The plain elevation of ancient and former boxers is too simple. I will look into statistics behind.

Hey, that Carroll guy wasn't bad.. had never heard of him before.

Notice he and Pastor had records like current top boxers.. except they fought more and point totals were tremendous.

JCS wrote: ↑14 Oct 2019, 10:44 Hey, that Carroll guy wasn't bad.. had never heard of him before.

Notice he and Pastor had records like current top boxers.. except they fought more and point totals were tremendous.

I assume Carroll is a legend in Australia, I've never heard of him. Upon the little that I have read of him, he seems worthy of some type of recognition, but #2 Welterweight of all time is beyond explanation.
I have heard of Pastor because of his fights with Joe Louis on YouTube. Again, worthy of recognition, but it's too high.

I know rankings are very subjective but, I can't see anyone arguing for their rankings over more noteworthy fighters.

However, one thing I really appreciate about seeing names I've never heard of is learning about said fighters.

The points level of the Whole-History Ratings depends on:
- the number of bouts per year
- the number of active boxers
- the conversion factors between the weight divisions

I compensated for this effect by;
- reducing the weight of a bout with higher number of bouts per year
- reducing the start rating of a boxer with higher number of active boxers
- reducing the conversion factors between the weight divisions

http://151.boxrec.com/~martin/ratings_at.php
http://151.boxrec.com/~martin/ratings.php

computerrank wrote: ↑15 Oct 2019, 19:11 http://151.boxrec.com/~martin/ratings.php

Usyk #1 at Heavyweight?


JESSIE VARGAS #1 at 154???????

I was open minded with Brook though I disagreed, but JV is bottom 10 AT BEST, no question

computerrank wrote: ↑15 Oct 2019, 19:11

http://151.boxrec.com/~martin/ratings_at.php

James J Jeffries #3 at Heavyweight
Bob Fitzsimmons #1 at Light Heavyweight
Hell, anyone not named Archie Moore #1 at Light Heavyweight
Bruno Arcari (who?) #1 at 140... over Julio Cesar Chavez no less
#2 Super Featherweight of all time is Nkosana Mgxaji...........WHO? lol


Let's go champ!

I now implemented a rating points conversion for boxers moving the division - directly in the core maximum likelihood algorithm.

You can inspect the results for 3 levels: w1, w2 and w3. please use the test databases:
- test_w1
- test_w2
_test_w3

with the known procedures:

http://151.boxrec.com/~martin/ratings.php
http://151.boxrec.com/~martin/ratings_at.php

computerrank wrote: ↑16 Oct 2019, 18:16

http://151.boxrec.com/~martin/ratings.php

Edit: I didn't do it correctly, sorry

computerrank wrote: ↑16 Oct 2019, 18:16
http://151.boxrec.com/~martin/ratings.php

Ok, let's see what I can see...

1. Tom Schwarz's high ranking. Not only is it too high, it's over Otto Wallin. The former was embarrassed by while the latter competitive with Tyson Fury, a top 3 heavyweight at worst. (All 3)

2. Super Middleweight. Juergen Braehmer beat a future world champion here in Rob Brant, 2 years ago. Nothing notable since and nothing really notable before either to be honest (at 175). Perhaps he's a top 20-30 guy out of respect for a former champion.

3. Light Welterweight. Terry Flanagan was an OK lightweight, and has done absolutely nothing to deserve a top 10 spot at 140. Jorge Linares is a bit high for someone defeated in 1 round by a middle tier gatekeeper. No disrespect to Cano, but Linares can't maintain such a high ranking. He's 11th in the 3rd setting, but it's still a bit high.

4. My favorite fighter, Jhonny Gonzalez is a bit too high at 130..... but I'm not complaining

5. In conclusion, the 3rd test is a league above the 1st and 2nd in my honest opinion.

computerrank wrote: ↑16 Oct 2019, 18:16
http://151.boxrec.com/~martin/ratings_at.php

I'll just be direct. Jack Johnson not being in the top 50 all time heavyweights for all 3 tests is enough to say there's a major problem.

Nothing else really needs to be said.

computerrank wrote: ↑15 Oct 2019, 19:11 The points level of the Whole-History Ratings depends on:
- the number of bouts per year
- the number of active boxers
- the conversion factors between the weight divisions

I compensated for this effect by;
- reducing the weight of a bout with higher number of bouts per year
- reducing the start rating of a boxer with higher number of active boxers
- reducing the conversion factors between the weight divisions

http://151.boxrec.com/~martin/ratings_at.php
http://151.boxrec.com/~martin/ratings.php

As one who also makes computer rankings, I appreciate the work you're putting in to improve the system, and also sympathize with inevitably coming up with the "perfect rankings" only to find that one boxer has now moved to an 'unacceptable" position, in someone's eyes....

I'm curious about the ranking of Jack Dempsey, who almost always ranks very poorly in algorithmic systems but the boxing fandom wants to see in the top 10. Previously he was pretty low in the all-time rankings, now he's coming in at #17. What do you think is the basic mechanism of the WHR that allowed him to move up? Is it mostly due to using the peak ranking?

What is boxrec's policy on rating suspended drug users?

SportsRatings wrote: ↑17 Oct 2019, 08:58 As one who also makes computer rankings, I appreciate the work you're putting in to improve the system, and also sympathize with inevitably coming up with the "perfect rankings" only to find that one boxer has now moved to an 'unacceptable" position, in someone's eyes....

I'm curious about the ranking of Jack Dempsey, who almost always ranks very poorly in algorithmic systems but the boxing fandom wants to see in the top 10. Previously he was pretty low in the all-time rankings, now he's coming in at #17. What do you think is the basic mechanism of the WHR that allowed him to move up? Is it mostly due to using the peak ranking?

The current Boxrec version has him at #17. And this version works with points for defeating top ranked opponents.

The WHR version in test has him around #25. And this version works with career top rating.

tiny_acres wrote: ↑17 Oct 2019, 09:20 What is boxrec's policy on rating suspended drug users?

There is no Boxrec policy on this issue. as far as I know. You could ask this in the main forum.

- bouts before 1910 are weighted double now (see Jack Johnson)
- rating points conversion for weight divisions changed

- I guess, p4p rankings look better now and current ratings look better now too

http://151.boxrec.com/~martin/ratings_at.php
http://151.boxrec.com/~martin/ratings.php

computerrank wrote: ↑17 Oct 2019, 09:34

tiny_acres wrote: ↑17 Oct 2019, 09:20 What is boxrec's policy on rating suspended drug users?

There is no Boxrec policy on this issue. as far as I know. You could ask this in the main forum.

Thank you

tiny_acres wrote: ↑17 Oct 2019, 12:40

computerrank wrote: ↑17 Oct 2019, 09:34
There is no Boxrec policy on this issue. as far as I know. You could ask this in the main forum.

Thank you

Oh well.

Roy was great but how high does his resume realistically take him on a rating that goes by results?

Cobwebcat wrote: ↑17 Oct 2019, 14:24 Current fine.

All time not fine. At his peak Roy Jones was the best I’ve seen. Not a mention on p4p and low on LHW. There’s lots of examples but I also sympathise that no rating will ever be perfect for everyone. I still prefer your current system for all time strangely.

More imprtantly I’m personally never a fan of using random numbers that aren’t calculated for accuracy prediction. If you double weight fights before 1910 don’t you start to lose the integrity of the algorithm? Why not 1890 and triple? To me there has to be a mathematical reason for using a figure like optimising the K value in an elo rating for predictive accuracy. If it’s not objective it becomes subjective and then it’s only a hop and a skip from a personal list.

Your are right with this ad-hoc double weights before 1910. It may look better. But - all objective factors based on bouts/year and active boxers were priced in before.

So I will spare this factor.

Maybe the current system for all-time ratings with rewarding the balance of annual top wins and annual bad losses is the more acceptable. I will try to adapt it with the WHR.

At WHR all-time ratings now Jones is #3 at Light Heavyweight and #57 at P4P - at the Boxrec all-time ratings Jones is #2 at Light Heavyweight and #20 at P4P - not an eminent difference, I guess.

computerrank wrote: ↑17 Oct 2019, 11:48
http://151.boxrec.com/~martin/ratings.php

1. Same problem with Tom Schwarz.

2. Same problem with Juergen Braehmer (sp?)

3. I must've missed this, but how is Brian Castano not a top 10 154 fighter? He beat Michel Soro and drew with Erislandy Lara.

4. At welterweight, there's a fighter that jumped from #342 to #30. Raul Rodriguez. I searched for him on boxrec and here's what I found...

https://boxrec.com/en/proboxer/901073

...I'm just going to stop right here, no need to go any further

Manrae wrote: ↑17 Oct 2019, 19:55

computerrank wrote: ↑17 Oct 2019, 11:48
http://151.boxrec.com/~martin/ratings.php

1. Same problem with Tom Schwarz.

2. Same problem with Juergen Braehmer (sp?)

3. I must've missed this, but how is Brian Castano not a top 10 154 fighter? He beat Michel Soro and drew with Erislandy Lara.

4. At welterweight, there's a fighter that jumped from #342 to #30. Raul Rodriguez. I searched for him on boxrec and here's what I found...

https://boxrec.com/en/proboxer/901073

...I'm just going to stop right here, no need to go any further

WHR considers the uncertainty of a rating. The rating uncertainty increases with time between bouts, when the rating is not confirmed by a result. The rating shown is the lowest value of the trust interval.

Tom Schwarz's rating ls estimated to be between 3.058 ... 25.8, a span factor of 8.5
Otto Waliin's rating is estimated to be 0.981 ... 21.8, a span factor of 22.4

Wallin loss against Fury was wide. His win against Grant is past 18 months.
Schwarz has 4 valid wins within the last 18 months.

Braehmer decends from a high level - see his win against Brant.

Castrano drew with Lara - but this i no win. and Soro is past 24 months now. And Vitu is in decline.

Raul Rodriguez' rating is an artefact, when updating the database. This will be correct with next database update.

Dears,

Why Deontay Wilder is not in this list ?!

https://boxrec.com/en/ratings?ZtQ%5Bcou ... %5D=&r_go=

Thanks
AmirD