Ratings - please read before commenting - Archived

[JCS]

Ahh, I did not want to assume what "equilibrium-based" meant.. but it seems we have moved back one step closer to an ELO system. Is this really what we want to do?

[computerrank]

Ahh, I did not want to assume what "equilibrium-based" meant.. but it seems we have moved back one step closer to an ELO system. Is this really what we want to do?

... it is a bit more ...

Elo would update the ratings bout by bout as we do currrently.

These ratings concider all bouts of a rating time period of 10 years in one step, weighed decendend in the time domain. So results of boxers affect ratings of former opponents. And this results in drastic better predicition quality, regardless being still only based on achievements.

Yes, we want to be closer to ratings, confirmed by better match with the following results.

But we might make some cosmetical adaptions, which make them look more acceptable to the community.

[keithmoonhangover]

That's interesting, but there must be a flaw in the system. Jones is not the 10th best Light Heavy in the world.

He hasn't really lost to anyone bad enough to seriously damage his rating.

Losing in the 1st round to Danny Green surely means that Jones in not a top ten fighter.

[JCS]

Ahh, I did not want to assume what "equilibrium-based" meant.. but it seems we have moved back one step closer to an ELO system. Is this really what we want to do?

... it is a bit more ...

Elo would update the ratings bout by bout as we do currrently.

These ratings concider all bouts of a rating time period of 10 years in one step, weighed decendend in the time domain. So results of boxers affect ratings of former opponents. And this results in drastic better predicition quality, regardless being still only based on achievements.

Yes, we want to be closer to ratings, confirmed by better match with the following results.

But we might make some cosmetical adaptions, which make them look more acceptable to the community.

I still think the most recent bouts should hold the highest importance... After all, with top fighters only fighting a few times a year.. I think this is necessary.

[squiggy]

What is the reason for decreasing a boxer's rating for moving up in weight class? Is it just based on an assumption that he is a lesser fighter now that he weighs more?
Looking at a couple of its effects:
Pacquaio comes into 2008 with a rating of 906 after beating Barrera in 10/07. Then, at the end of the year, after beating Marquez at super feather, Diaz at light, and De la Hoya at welter, he is at 940. Had these wins all taken place in one weight class -- which would make them less of an accomplishment, not more -- he'd probably have gained several hundred points. Instead, he hasn't gained enough to move up a p4p list.
Similarly, Roy Jones Jr. comes out of his heavyweight championship victory over John Ruiz with almost 300 fewer points than he had after his last light heavy win six months earlier. Then his total almost doubles when he goes back down to light heavy.
It seems that, as a means of p4p ranking, this formula gives too much weight to weight, so to speak.

[JCS]

What is the reason for decreasing a boxer's rating for moving up in weight class?

Take Pacquiao's rating and compare it to Heavyweights. Where would he rank??

Ok.

[squiggy]

What is the reason for decreasing a boxer's rating for moving up in weight class?

Take Pacquiao's rating and compare it to Heavyweights. Where would he rank??

Ok.

I'm not sure what you mean or how your comment answers my question.
The hypothetical "how would Manny Pacquiao do against a heavyweight" has nothing to do with reality, whereas the questions I raised were about his actual rankings following his actual fights.

[m1kee50]

What is the reason for decreasing a boxer's rating for moving up in weight class?

Take Pacquiao's rating and compare it to Heavyweights. Where would he rank??

Ok.

I'm not sure what you mean or how your comment answers my question.
The hypothetical "how would Manny Pacquiao do against a heavyweight" has nothing to do with reality, whereas the questions I raised were about his actual rankings following his actual fights.

the rankings are about an actual depiction of relative acheivement - with the added factor that they are ideally able to be used for prediction... Pacman as a Heavy, with undecreased ratings, is the best HW in the world... except he isnt, is he? Same as Pacman at Cruiser, or LHW...

that he can compete at MW or JMW is a fact, but every boxer moving up in weight is an unknown quantity, if he was clear #1 at 135, he's a bit closer to the #2 at 140, he certainly cannot be assumed to be leap years ahead at 147 or 154 - to reflect this, his ratings have to drop.

[squiggy]

Okay. I see your point, as it pertains to rankings within a weight class. But unless I'm missing something, it does still seem to have an unfortunate effect on p4p rankings wherein a fighter gets little credit for winning fights at higher weights.

[JCS]

Okay. I see your point, as it pertains to rankings within a weight class. But unless I'm missing something, it does still seem to have an unfortunate effect on p4p rankings wherein a fighter gets little credit for winning fights at higher weights.

Yep - which makes P4P in a system like this, an unreasonable calculation.

[m1kee50]

Okay. I see your point, as it pertains to rankings within a weight class. But unless I'm missing something, it does still seem to have an unfortunate effect on p4p rankings wherein a fighter gets little credit for winning fights at higher weights.

the p4p ratings here are better looked at as a reflection of single-division dominance... they really should rename it... from that point of view it makes more sense doesnt it?

[squiggy]

It's a shame, though. A list of division-dominance numbers is much less interesting than a good p4p system would be.

[JCS]

It's a shame, though. A list of division-dominance numbers is much less interesting than a good p4p system would be.

How do you make a good P4P system?

[squiggy]

Well... I'll think on it.

[keithmoonhangover]

That's interesting, but there must be a flaw in the system. Jones is not the 10th best Light Heavy in the world.

He hasn't really lost to anyone bad enough to seriously damage his rating.

Losing in the 1st round to Danny Green surely means that Jones in not a top ten fighter.

[JCS]

That's interesting, but there must be a flaw in the system. Jones is not the 10th best Light Heavy in the world.

He hasn't really lost to anyone bad enough to seriously damage his rating.

Losing in the 1st round to Danny Green surely means that Jones in not a top ten fighter.

Tell the computer that. I think he's due to fall some in March.

[m1kee50]

It's a shame, though. A list of division-dominance numbers is much less interesting than a good p4p system would be.

How do you make a good P4P system?

Q for you and computerrank i guess:

if everyone active retired tomorrow, and we used the all-time ratings to produce a p4p list - but only of the ones who were active - what would it look like?

[computerrank]

It's a shame, though. A list of division-dominance numbers is much less interesting than a good p4p system would be.

How do you make a good P4P system?

Q for you and computerrank i guess:

if everyone active retired tomorrow, and we used the all-time ratings to produce a p4p list - but only of the ones who were active - what would it look like?

... you won't be happy with this:

Code: Select all

+----------------+---------------+----------+
| first_name     | last_name     | r_at_act |
+----------------+---------------+----------+
| Bernard        | Hopkins       |     1259 |
| Roy            | Jones Jr      |     1126 |
| Floyd          | Mayweather Jr |     1124 |
| Evander        | Holyfield     |     1123 |
| Manny          | Pacquiao      |     1035 |
| Juan Manuel    | Marquez       |      944 |
| Shane          | Mosley        |      857 |
| Wladimir       | Klitschko     |      839 |
| Marco Antonio  | Barrera       |      772 |
| Rafael         | Marquez       |      666 |
| Antonio        | Tarver        |      584 |
| Erik           | Morales       |      493 |
| Vic            | Darchinyan    |      486 |
| Veeraphol      | Sahaprom      |      476 |
| Chris          | John          |      451 |
| Cory           | Spinks        |      443 |
| Vitali         | Klitschko     |      432 |
| Antonio        | Margarito     |      418 |
| Jose Luis      | Castillo      |      417 |
| Kelly          | Pavlik        |      417 |
| Pongsaklek     | Wonjongkam    |      388 |
| Sergio Gabriel | Martinez      |      382 |
| Glen           | Johnson       |      378 |
| Johnny         | Tapia         |      371 |
| Eric           | Morel         |      371 |
+----------------+---------------+----------+

[JCS]

I think the community strength needs to be looked into further... Some of the South Africans and Japanese are really high compared to the current system.

Here's an idea. Can you take the 100 "Biggest" fights in the past 5 years..

Compare prediction rates between the two using only those fights?

[computerrank]

I think the community strength needs to be looked into further... Some of the South Africans and Japanese are really high compared to the current system.

Here's an idea. Can you take the 100 "Biggest" fights in the past 5 years..

Compare prediction rates between the two using only those fights?

- only bouts with both boxers above a specific rating limit considered?

- then compare the results with the predictions

- segregated by South Africans, Japanese, Mexicans, Philippinoes, Thais ...

- identify significant deviations from the average predictions rates

something like this?

[JCS]

I think the community strength needs to be looked into further... Some of the South Africans and Japanese are really high compared to the current system.

Here's an idea. Can you take the 100 "Biggest" fights in the past 5 years..

Compare prediction rates between the two using only those fights?

- only bouts with both boxers above a specific rating limit considered?

- then compare the results with the predictions

- segregated by South Africans, Japanese, Mexicans, Philippinoes, Thais ...

- identify significant deviations from the average predictions rates

something like this?

Yes. Top 100 fights with both fighters meeting minimum rating.

Also If possible, I would also like to segregate career Japanese, Thais, Filipinos, and South Africans, and see the prediction rate of their first fights outside of their home country.

[JCS]

Also Martin, what is overall prediction rate only for 2000-2011?

[computerrank]

Also Martin, what is overall prediction rate only for 2000-2011?

Benchmarks (2000 - 2010):

- current Boxrec ratings 77.3 % correct predictions = right to false ratio = 3.4 : 1

- new approach ratings 83.4 % correct predictions = right to false ratio = 5 : 1

[JCS]

Martin --

How did Chauncy gain 7 points from beating a 0 rating boxer?

http://boxrec.com/list_bouts.php?human_ ... &cat=boxer

[computerrank]

Martin --

How did Chauncy gain 7 points from beating a 0 rating boxer?

http://boxrec.com/list_bouts.php?human_ ... &cat=boxer

A boxer can't gain regular points against an opponent with 0 points.

But there is the area of additional points.

A boxer can get additional points against an opponent with relevant launch category.

Try the >BoxRec Points calculator<, I implemented to answer the questions of boxers and managers planning their new bouts.

http://boxrec.dyndns.org/~martin/calc.php

- either enter the boxer id
- or enter a rating and a launch catergory
- mixed is possible

you will get:
- the point changes depending on the result
- the launch category changes