Ratings - please read before commenting - Archived

[pugilisticspecialist]

Do you think we will ever see pre-fight and post-fight RANKINGS of fighters on the "bout" page? For historical purposes, this would be Godly. I've suggested this before but, nothing.

Fightmatrix.com is like BoxRec for MMA and they have quarterly rankings for every fighter in the database - it's super convenient.

They have other cool features too, like an age filter and "Show Undefeated Fighters Only" option.

[computerrank]

Interesting...

Do you think we will ever see pre-fight and post-fight RANKINGS of fighters on the "bout" page? For historical purposes, this would be Godly. I've suggested this before but, nothing.

Honestly, "John" just gets curt with most people making suggestions, in the suggestion thread

To make it clear, the suggested corrected ratings release is accepted. But the core team hesitates to set it live, because we already had some massive changes in the ratings this year. So they might set it back until June.

I would like the rankings option too, but it costs too much computer ressources to keep sorted the rankings of up to 40,000 active boxers in a daily udate within the ratings run.

[computerrank]

Rating release 41 was activated now.

All reported issues should be corrected now.

Changes:

# - at most 40 additional points for defeating an opponent in the 15 percent p4p ranks
# -- who already defeated such an opponent himself
# -- 8 point at least
# -- opponents current rating at least
# -- 40 points at most
# -- weight by clearness of decision and rounds boxed
# -- weight by opponent's losses after his win against an opponent in the 15 percent p4p ranks
# --- weight 1 after no loss, 1/2 after 1 loss, 1/4 after 2 losses etc
# --- weight 1 until 18 months after win, decreasing to 1/2 after every 18 months

# - at most 1 points for defeating an opponent, who defeated a winner himself
# - at most 0.1 points for defeating an opponent with a win
# - at most 0.01 points for a win
# -- weight for all by clearness of decision and rounds boxed

[Bigroe]

When will results from York Hall show 4th May be shown and new ratings be show.
Thanks

[computerrank]

When will results from York Hall show 4th May be shown and new ratings be show.
Thanks

Please ask in the Record Queries & Updates forum here:
viewforum.php?f=1

This forum is for technical ratings questions only.

[Manrae]

To make it clear, the suggested corrected ratings release is accepted. But the core team hesitates to set it live, because we already had some massive changes in the ratings this year. So they might set it back until June.

I would like the rankings option too, but it costs too much computer ressources to keep sorted the rankings of up to 40,000 active boxers in a daily udate within the ratings run.

What about... just for retired boxers? As long as there's no ratings changes, the rankings wouldn't change. Would that take up too much resources as well? That way we could see where, for example, James "Buster" Douglas was ranked when he KO'd Mike Tyson in January of 1990 by clicking on either boxer's "bout" link, since they're both retired.

Also, can't wait to look around at the new update

[Manrae]

Rating release 41 was activated now.

All reported issues should be corrected now.

Changes:

# - at most 40 additional points for defeating an opponent in the 15 percent p4p ranks
# -- who already defeated such an opponent himself
# -- 8 point at least
# -- opponents current rating at least
# -- 40 points at most
# -- weight by clearness of decision and rounds boxed
# -- weight by opponent's losses after his win against an opponent in the 15 percent p4p ranks
# --- weight 1 after no loss, 1/2 after 1 loss, 1/4 after 2 losses etc
# --- weight 1 until 18 months after win, decreasing to 1/2 after every 18 months

# - at most 1 points for defeating an opponent, who defeated a winner himself
# - at most 0.1 points for defeating an opponent with a win
# - at most 0.01 points for a win
# -- weight for all by clearness of decision and rounds boxed

Gilberto Ramirez and Luke Campbell are both in the wrong weight class.
Ramirez is now campaigning at Light Heavyweight and has given up his WBO 168lb title
Campbell is close to finalizing a fight with Lomachenko for the vacant WBC Lightweight title
Unless a fighter has a significant fight at another weight, the constant weight class changes made by the admins is counter-productive and leads to inflated/deflated ratings. Historically it's a complete mess, but so are the records I guess.....


I'm sure that one of the test links you provided had Caleb Plant, David Benavidez and Anthony Dirrell in the top 10
Now, they're not and John Ryder and Rocky Fielding remain...

Is

[jamamb]

these catz always unsatisfied

[Manrae]

these catz always unsatisfied

Yep

[computerrank]

Gilberto Ramirez and Luke Campbell are both in the wrong weight class.
Ramirez is now campaigning at Light Heavyweight and has given up his WBO 168lb title
Campbell is close to finalizing a fight with Lomachenko for the vacant WBC Lightweight title
Unless a fighter has a significant fight at another weight, the constant weight class changes made by the admins is counter-productive and leads to inflated/deflated ratings. Historically it's a complete mess, but so are the records I guess.....


I'm sure that one of the test links you provided had Caleb Plant, David Benavidez and Anthony Dirrell in the top 10
Now, they're not and John Ryder and Rocky Fielding remain...

Is

Boxers and their managers can claim to be rated in the division, they want to campaign in. You can post your remarks in the Records Queries & Updates forum.

Caleb Plant is #9. Benavidez #8 -> #11 and Dirrell #10 -> #12, some of their opponents are rated lower now. Benavidez lost 98 points, Dirrell lost 63 points. Ryder #5 -> #6 lost 33 points, Fielding #6 -> #7 lost 60 points.

[Manrae]

Boxers and their managers can claim to be rated in the division, they want to campaign in. You can post your remarks in the Records Queries & Updates forum.

Caleb Plant is #9. Benavidez #8 -> #11 and Dirrell #10 -> #12, some of their opponents are rated lower now. Benavidez lost 98 points, Dirrell lost 63 points. Ryder #5 -> #6 lost 33 points, Fielding #6 -> #7 lost 60 points.

It's still wrong.

Ah well, I'll just wait for a better release

[JCS]

They'll never be perfect.

Better to accept and move on.. or even better yet, create your own systems..

[JohnMcMinn]

The new release seems like an improvement to me. It definitely clears up some oddities and "points explosions" I was noticing.

[Manrae]

They'll never be perfect.

Better to accept and move on.. or even better yet, create your own systems..

I disagree. The absence of perfection is the opportunity to always improve. That goes for me, you, boxing analytics, anything... no need to accept stagnation in a perpetually moving existence

I'm no mathematician, so I can't create a system such as this, but I do have my own system for rating fighters, predictions and such but 90% of it is just looking at their fights and making guesses

[JCS]

I disagree. The absence of perfection is the opportunity to always improve. That goes for me, you, boxing analytics, anything... no need to accept stagnation in a perpetually moving existence

I'm no mathematician, so I can't create a system such as this, but I do have my own system for rating fighters, predictions and such but 90% of it is just looking at their fights and making guesses

But what is "perfect"?

We can base it on prediction rate alone.. that's the most objective way to do it.. but you won't like the results, I guarantee it.

[Manrae]

But what is "perfect"?

We can base it on prediction rate alone.. that's the most objective way to do it.. but you won't like the results, I guarantee it.

No problems, faults, impossible to be made any better. It's an idea, not an actual reality. Something to strive for even though we all ultimately fail, it makes us better. If I say something or someone is "perfect" I'm simply saying that it is as good as I can ever perceive it to be though it's not objectively true. I'm not asking for objective perfection, but a continual effort towards that goal, just as he's been doing. Nothing wrong with that, I'm not being negative by giving legitimate critiques. It makes the site better, we all should pitch in and give our thoughts and opinions

Wait, do you work for the site too? I'm asking because you used "we" and I always assumed it was just the one guy here who does the mathematics...

[JCS]

Wait, do you work for the site too? I'm asking because you used "we" and I always assumed it was just the one guy here who does the mathematics...

I used to chip in back in the day... maybe 12 years ago or so.. we went for a prediction-based system for a short time.. it didn't go so well. Ever since and even before this, I'd toss in some ideas to try and help move things along.

About 10 years ago.. I took what I learned and made an MMA system. Not sure there's a better one out there.

[Manrae]

I used to chip in back in the day... maybe 12 years ago or so.. we went for a prediction-based system for a short time.. it didn't go so well. Ever since and even before this, I'd toss in some ideas to try and help move things along.

About 10 years ago.. I took what I learned and made an MMA system. Not sure there's a better one out there.

That's awesome, if I ever have time I'd like to create something boxing related that the boxing community could be involved in even though I've never done it before, it just takes time and patience

[JCS]

That's awesome, if I ever have time I'd like to create something boxing related that the boxing community could be involved in even though I've never done it before, it just takes time and patience

A prediction-based system is the only objective way to do it.. set an objective and the output is a percentage of correctness.
It is the only way to prove that a system is "better" or that a change has "improved" the ratings. However, it is ugly compared to traditional style.

The state we are in now is just ridiculous --- "Why is this guy #456 and this guy #515??"... feels like a shell game.


In my view, the weak point is the all-time rankings and has been for a while. Not necessarily the calculations, but also the divisional grouping.

[Manrae]

A prediction-based system is the only objective way to do it.. set an objective and the output is a percentage of correctness.
It is the only way to prove that a system is "better" or that a change has "improved" the ratings. However, it is ugly compared to traditional style.

The state we are in now is just ridiculous --- "Why is this guy #456 and this guy #515??"... feels like a shell game.


In my view, the weak point is the all-time rankings and has been for a while. Not necessarily the calculations, but also the divisional grouping.

YES.

...and though I didn't realize how detrimental Boxrec rankings could be until I realized that it's used in countries with smaller boxing populations to signify potential rankings regionally (hope that made sense). If I were a boxer I'd feel some sort of way as well if I felt other boxers regionally were ranked higher than me than they should. Once you become known internationally, or once you become ranked by a sanctioning body, that no longer really matters... for now.

Oh, and I'm not interested in creating a points system, I want to do something else, but time is the hurdle

[JCS]

YES.

...and though I didn't realize how detrimental Boxrec rankings could be until I realized that it's used in countries with smaller boxing populations to signify potential rankings regionally (hope that made sense). If I were a boxer I'd feel some sort of way as well if I felt other boxers regionally were ranked higher than me than they should. Once you become known internationally, or once you become ranked by a sanctioning body, that no longer really matters... for now.

Oh, and I'm not interested in creating a points system, I want to do something else, but time is the hurdle

If BoxRec cut off the rankings at Top 500 (or 250, etc), it'd solve a lot of the bitching. If you don't rank, you simply aren't ranked, end of story.

That's what I do in my system... I don't have the time to argue why Fighter A is ranked #700 and Fighter B is ranked #800... I may even agree that it is "wrong", but it is so insignificant and will work its way out once those fighters jump into the Top X that is visible.

In my view, the all-time rankings need to allow boxers to be rated in multiple divisions as a first step for improvement.

[computerrank]

...
In my view, the all-time rankings need to allow boxers to be rated in multiple divisions as a first step for improvement.

And then ... ?

[JCS]

And then ... ?

Once you make this shift.. it'll cause another problem...

Old-time boxers will likely hold an advantage, right? There were less divisions and as a result, less division jumping. This means that modern boxers will have their points spread out more... Is this acceptable?

[computerrank]

Once you make this shift.. it'll cause another problem...

Old-time boxers will likely hold an advantage, right? There were less divisions and as a result, less division jumping. This means that modern boxers will have their points spread out more... Is this acceptable?

Boxrec all time ratings are collected over the divisions. For every year the best division is assigned. So you can see the distribution over the divisions and years.

Currently I try to assign the complete boxers to the division with most annual points. But there is no automatic.

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

[JCS]

Boxrec all time ratings are collected over the divisions. For every year the best division is assigned. So you can see the distribution over the divisions and years.

Currently I try to assign the complete boxers to the division with most annual points. But there is no automatic.

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

Right.. but...

As a result, Hopkins (for example) who generated considerable achievements in multiple divisions.. is limited to MW. Rightfully so, but he did a few things at LHW as well... I assume the LHW achievements inflate his MW rating as well.

Plenty of others fit this bill too...