Ratings - please read before commenting - Archived

[mike1989]

why did Vasilia Lomachenko have a rating of 925 points and became 814 points ???

His last bout matching 50% of his rating was against Nicholas Walters on 2016-11-27,

18 months later on 2018-05-27 his best opponent Jason Sosa only had 359 points. This was less than 50% of Lomchenko's 925.

there is no Nicholas Walters. and Jason Sosa 2017-04-08 rating
Vasyl Lomachenko to fight 680 points and Jason Sosa 359 points and this is more than 50% ???

[computerrank]

why did Vasilia Lomachenko have a rating of 925 points and became 814 points ???

His last bout matching 50% of his rating was against Nicholas Walters on 2016-11-27,

18 months later on 2018-05-27 his best opponent Jason Sosa only had 359 points. This was less than 50% of Lomchenko's 925.

there is no Nicholas Walters. and Jason Sosa 2017-04-08 rating
Vasyl Lomachenko to fight 680 points and Jason Sosa 359 points and this is more than 50% ???

The best opponent rating in relation to the boxer's at THE END of the 18 months period counts. It is NOT the rating at bout time.

[mike1989]

His last bout matching 50% of his rating was against Nicholas Walters on 2016-11-27,

18 months later on 2018-05-27 his best opponent Jason Sosa only had 359 points. This was less than 50% of Lomchenko's 925.

there is no Nicholas Walters. and Jason Sosa 2017-04-08 rating
Vasyl Lomachenko to fight 680 points and Jason Sosa 359 points and this is more than 50% ???

The best opponent rating in relation to the boxer's at THE END of the 18 months period counts. It is NOT the rating at bout time.

I do not understand...???
50% and more from own rating and there is a new date of counting of 18 months ??? Explain please, Martin ??

[computerrank]

there is no Nicholas Walters. and Jason Sosa 2017-04-08 rating
Vasyl Lomachenko to fight 680 points and Jason Sosa 359 points and this is more than 50% ???

The best opponent rating in relation to the boxer's at THE END of the 18 months period counts. It is NOT the rating at bout time.

I do not understand...???
50% and more from own rating and there is a new date of counting of 18 months ??? Explain please, Martin ??

There is nothing new. When the Walters bout got out of 18 months scope, Sosa was left as the best opponent and Lomachenko's rating was reduced. Now this bout is valid until it gets out of scope or Lomachenko meets an opponent with half of his own rating.

[mike1989]

The best opponent rating in relation to the boxer's at THE END of the 18 months period counts. It is NOT the rating at bout time.

I do not understand...???
50% and more from own rating and there is a new date of counting of 18 months ??? Explain please, Martin ??

There is nothing new. When the Walters bout got out of 18 months scope, Sosa was left as the best opponent and Lomachenko's rating was reduced. Now this bout is valid until it gets out of scope or Lomachenko meets an opponent with half of his own rating.

at the time of the battle with Soso, the rating of lomachenko was 680 points and Sosa 359 points if we ranked Sosa 359 * 2 = 718 points, which is more than 50% of Lomachenko's rating ???
because the new reference point is a fight with Sosa 2017-04-08, and not Nikolas Volters ???
http://boxrec.com/en/event/745543/2135663

[JCS]

I do not understand...???
50% and more from own rating and there is a new date of counting of 18 months ??? Explain please, Martin ??

There is nothing new. When the Walters bout got out of 18 months scope, Sosa was left as the best opponent and Lomachenko's rating was reduced. Now this bout is valid until it gets out of scope or Lomachenko meets an opponent with half of his own rating.

at the time of the battle with Soso, the rating of lomachenko was 680 points and Sosa 359 points if we ranked Sosa 359 * 2 = 718 points, which is more than 50% of Lomachenko's rating ???
because the new reference point is a fight with Sosa 2017-04-08, and not Nikolas Volters ???
http://boxrec.com/en/event/745543/2135663

Follow the instructions..

r_red-10 = (r_old-10) * (1 - 0.5*(1 - 2*best_opp/r_old)) for men

r_red-10 = (925-10) * (1 - 0.5*(1 - 2*359/925))

r_red-10 = ~813

823-10 = 813

The fight with Sosa was at 130... not 135.. so perhaps that accounts for the slight difference.. maybe rounding impacts as well.

[computerrank]

There is nothing new. When the Walters bout got out of 18 months scope, Sosa was left as the best opponent and Lomachenko's rating was reduced. Now this bout is valid until it gets out of scope or Lomachenko meets an opponent with half of his own rating.

at the time of the battle with Soso, the rating of lomachenko was 680 points and Sosa 359 points if we ranked Sosa 359 * 2 = 718 points, which is more than 50% of Lomachenko's rating ???
because the new reference point is a fight with Sosa 2017-04-08, and not Nikolas Volters ???
http://boxrec.com/en/event/745543/2135663

Follow the instructions..

r_red-10 = (r_old-10) * (1 - 0.5*(1 - 2*best_opp/r_old)) for men

r_red-10 = (925-10) * (1 - 0.5*(1 - 2*359/925))

r_red-10 = ~813

823-10 = 813

The fight with Sosa was at 130... not 135.. so perhaps that accounts for the slight difference.. maybe rounding impacts as well.

Jason, you met the point. The Linares bout with 347 points at Lightweight counts more than the Sosa bout with 359 at Super Featherweight. So the reference is 347 points at Lightweight.

[mike1989]

at the time of the battle with Soso, the rating of lomachenko was 680 points and Sosa 359 points if we ranked Sosa 359 * 2 = 718 points, which is more than 50% of Lomachenko's rating ???
because the new reference point is a fight with Sosa 2017-04-08, and not Nikolas Volters ???
http://boxrec.com/en/event/745543/2135663

Follow the instructions..

r_red-10 = (r_old-10) * (1 - 0.5*(1 - 2*best_opp/r_old)) for men

r_red-10 = (925-10) * (1 - 0.5*(1 - 2*359/925))

r_red-10 = ~813

823-10 = 813

The fight with Sosa was at 130... not 135.. so perhaps that accounts for the slight difference.. maybe rounding impacts as well.

Jason, you met the point. The Linares bout with 347 points at Lightweight counts more than the Sosa bout with 359 at Super Featherweight. So the reference is 347 points at Lightweight.

I can not understand why the point of otchetcheta Volters and not Sosa ???

[JCS]

Follow the instructions..

r_red-10 = (r_old-10) * (1 - 0.5*(1 - 2*best_opp/r_old)) for men

r_red-10 = (925-10) * (1 - 0.5*(1 - 2*359/925))

r_red-10 = ~813

823-10 = 813

The fight with Sosa was at 130... not 135.. so perhaps that accounts for the slight difference.. maybe rounding impacts as well.

Jason, you met the point. The Linares bout with 347 points at Lightweight counts more than the Sosa bout with 359 at Super Featherweight. So the reference is 347 points at Lightweight.

I can not understand why the point of otchetcheta Volters and not Sosa ???

I don't understand what you are saying.

[jamamb]

whats up with lenin castillos points jumping over 4x between fights?

lenin castillo http://boxrec.com/en/boxer/545460

After Lenin Castillo fought Andy Perez, Lenin's boxrec rating became 40. In his very next fight, against Aaron Mitchell, Lenin's starting rating was 185. There's no logical explanation for why Lenin's rating would have more than quadrupled for no reason. I could understand it if he dropped from heavyweight to minimumweight, the formula would've quadrupled his rating. But this doesn't make sense.

[computerrank]

whats up with lenin castillos points jumping over 4x between fights?

lenin castillo http://boxrec.com/en/boxer/545460

After Lenin Castillo fought Andy Perez, Lenin's boxrec rating became 40. In his very next fight, against Aaron Mitchell, Lenin's starting rating was 185. There's no logical explanation for why Lenin's rating would have more than quadrupled for no reason. I could understand it if he dropped from heavyweight to minimumweight, the formula would've quadrupled his rating. But this doesn't make sense.

The Mitchell fight was erroneously assigned to minimumweight. Now it is assigned to light heavyweight. The points will change with the next program run.

[jamamb]

[JCS]

Vanilla ELO sucks for combat sports. ELO only really excels in sports/games where there is frequent, consistent, fair matchmaking at all times. Glicko (and Glicko 2) are somewhat better, but again, these vanilla rating systems can't handle the variability that a sport like boxing brings to the table. Weight classes, layoffs, huge upsets, bad matching... these are the bullets that shoot holes in the ELO water barrel.

You need something that combines quality and quantity in order to make era-specific accomplishments comparable.

Baseball uses a statistic called JAWS to do this.

Additionally, and this is in my opinion, beating the #1 guy in a division of 50, should not be equivalent to beating the #1 guy in a division of 1000. How you tackle that is one of the biggest issues. Certainly, the latter should not be worth 20x (1000 / 50), but it shouldn't be equal either.. again, this is my opinion.

[Manrae]

The all-time ratings are looking strange these days.

Difficult to compare fighters across generations though. I wonder if there are standard algorithms that tackle this problem..elo perhaps?

The huge numbers of fights boxers used to have compared with now must also be a problem too.

Doesn’t look quite right though does it?

Zab Judah over Henry Armstrong
Vyacheslav Glazkov over Jersey Joe Walcott
There's a lot more...

Rule of thumb, if you can't defend these ratings in real life... something is wrong.

I've already brought this to his attention, but... nothing.

[Manrae]

Vanilla ELO sucks for combat sports. ELO only really excels in sports/games where there is frequent, consistent, fair matchmaking at all times. Glicko (and Glicko 2) are somewhat better, but again, these vanilla rating systems can't handle the variability that a sport like boxing brings to the table. Weight classes, layoffs, huge upsets, bad matching... these are the bullets that shoot holes in the ELO water barrel.

You need something that combines quality and quantity in order to make era-specific accomplishments comparable.

Baseball uses a statistic called JAWS to do this.

Additionally, and this is in my opinion, beating the #1 guy in a division of 50, should not be equivalent to beating the #1 guy in a division of 1000. How you tackle that is one of the biggest issues. Certainly, the latter should not be worth 20x (1000 / 50), but it shouldn't be equal either.. again, this is my opinion.

Very interesting... and very true observation in a subjective sport like boxing where corruption is rampant then and now.

[computerrank]

Vanilla ELO sucks for combat sports. ELO only really excels in sports/games where there is frequent, consistent, fair matchmaking at all times. Glicko (and Glicko 2) are somewhat better, but again, these vanilla rating systems can't handle the variability that a sport like boxing brings to the table. Weight classes, layoffs, huge upsets, bad matching... these are the bullets that shoot holes in the ELO water barrel.

You need something that combines quality and quantity in order to make era-specific accomplishments comparable.

Baseball uses a statistic called JAWS to do this.

Additionally, and this is in my opinion, beating the #1 guy in a division of 50, should not be equivalent to beating the #1 guy in a division of 1000. How you tackle that is one of the biggest issues. Certainly, the latter should not be worth 20x (1000 / 50), but it shouldn't be equal either.. again, this is my opinion.

In the Boxrec Ratings the number of boxers above a certain rating follows a power rule

number_boxers(above half_rating) / number_boxers(above rating) ~= 3

So boxers in less populated divisions show lower point values in the top ranks.

The Boxrec All Time Ratings award annual points for the best ranked opponent defeated. But these points are in relation to the ratio of the annual number of boxers above a point limit in the division.

[mike1989]

Tyson Fury has an error in the current ranking:

After the fight with Wladimir Klitschko 2015-11-28 the rating was 641 points

the fall of the rating occurred on 2017-05-28 and the Fury rating was 320.5 points

The next drop in the rating will occur on 2018-11-28

why is the current rating of Tyson Fury 206 points ???

[computerrank]

Tyson Fury has an error in the current ranking:

After the fight with Wladimir Klitschko 2015-11-28 the rating was 641 points

the fall of the rating occurred on 2017-05-28 and the Fury rating was 320.5 points

The next drop in the rating will occur on 2018-11-28

why is the current rating of Tyson Fury 206 points ???

Hi Michael,
yes, you are right with the old program release.
Rule 23 was extended to:
The rating of a boxer is reduced by 50% for every time period of inactivity of 18 months - and to even less in proportion to a longer time period of inactivity

[mike1989]

Tyson Fury has an error in the current ranking:

After the fight with Wladimir Klitschko 2015-11-28 the rating was 641 points

the fall of the rating occurred on 2017-05-28 and the Fury rating was 320.5 points

The next drop in the rating will occur on 2018-11-28

why is the current rating of Tyson Fury 206 points ???

Hi Michael,
yes, you are right with the old program release.
Rule 23 was extended to:
The rating of a boxer is reduced by 50% for every time period of inactivity of 18 months - and to even less in proportion to a longer time period of inactivity

Hi, Martin! You can please by example ??

[computerrank]

Tyson Fury has an error in the current ranking:

After the fight with Wladimir Klitschko 2015-11-28 the rating was 641 points

the fall of the rating occurred on 2017-05-28 and the Fury rating was 320.5 points

The next drop in the rating will occur on 2018-11-28

why is the current rating of Tyson Fury 206 points ???

Hi Michael,
yes, you are right with the old program release.
Rule 23 was extended to:
The rating of a boxer is reduced by 50% for every time period of inactivity of 18 months - and to even less in proportion to a longer time period of inactivity

Hi, Martin! You can please by example ??

r_red - 10 = (r_old - 10) * (1 - 0.5*(1 - 2*best_opp/r_old)) for men
best_opp=0 for time of inactivity
r_red - 10 = (r_old - 10) * (1 - 0.5) for men and 18 months of inactivity
extended to more than 18 months inactivity
r_red - 10 = (r_old - 10) * (1 - 0.5)**(time of inactivity / 18 months); 18 months = 365.24 days/year * 1.5 years

example Tyson Fury

inactivity from 2015-11-28 to 2018-06-09 = datedif("2015-11-28";"2018-06-09";"d") = 924 days
r_red - 10 = (641 - 10) * (1 - 0.5)**(924/547.86) = 631*0.5**1.687 = 631*0.3107 = 196
r_red = 196 + 10 = 206

[computerrank]

If Usyk moved up to heavyweight tomorrow where would he be ranked?

400 points and #2

[computerrank]

Wow those all time ratings are a mess.

SRR six places behind PAC Man?!

This didn’t used to be the case. Whatever happened was for the worse. Doesn’t pass the sense test.

Too many recent fighters at the top.

Thanks a lot, Cobwebcat
I found and fixed a bug in the implementation. Boxers with a loss in the year, could not compensate for it, even with a strong win. The fixed release is launched now..

[mike1989]

hello Martin and what has changed in the rating all time ???

[computerrank]

hello Martin and what has changed in the rating all time ???

The rules did not change. The rules were not implemented correctly.
- An annual win against an opponent is limited to the lower opponent rank of a worse loss.
- But by error ONE clear loss limited the annual achievement to 0.

[pugilisticspecialist]

Presuming Tyson Fury's next fight is against Wilder in December, what will his points rating be by then?
Considering it will be over 3 years since he fought an opponent in the top 100 at heavyweight, will he suffer another points reduction?