Ratings - please read before commenting - Archived

[Leonid]

... So this means that maximum rating of a boxer in a time period does not correlate well with the number of boxers/bouts in time period, right? That's my point - it doesn't and IMO number of boxers is a better factor of a strenth of division.

But its not my main point, main point is that the strength of the division should not be factored in all-time ratings as heavily as it is now (+ 14% of carreer high rating). ...

(1) The number of boxers in a time period is NOT a good measure for the top ratings in a time period.

(2) It looks like the top rating in a time period correlates with the number of boxers in the time periods before ... with some lag of 10 - 15 years. The model might be . if a lot of boxers start their career and peak their career 10 to 15 year later, the top rating will peak too ...

I will go into this - perhaps understanding this process will help ...

(1) - Your diagram shows that since 50s your measure of division strength (top ratings) was more or less steadily declining, while mine (number of boxers) shows it as steady till mid 80s and then increase. I say mine correlates better with the perceived strength, but surely it is highly subjective and 30s-40s appearing the strongest in both approached does not look appropriate. That's why one of my suggestions was to exclude this factor of a division strength at all from the all-time ratings.
(2) - Doubt it, but that is a possibility.

(3) - Why do you use as much as 14%? Right now you imply that HW 50s division is a few times better than current (ie. Rocky got 799 point = 60% of his total all-time points vs Wlad's 129 point = 10%of his total for this "14% of carreer high rating" which is based on the strength of the division). I'd say much lower than 14% would be more appropriate to disallow one-time success (maybe just luck) in any division mean more than long-time success in others. But lowering the percentage would still mean 50s HW div is better than current, when in reality it is hardly so.

Paradox: If Marciano faced a total bum in his last fight after Moore and got ko'ed by a lucky punch, his opponent would guarantee himself a place in like top-35 HWs of all time right above Vitali Kitschko and Ken Norton. He'd get to what - at least 3000 current rank points and all-time 200/1*33%+200/1*33% + 14%*3000 = 552. And how many of the HW boxers since 70s have got this many all-time points? Only several ATGs.

[mike1989]

and the rating Holyfield can lose ???

Holyfield looses points - from 1312 down to 1304

please reset the ratings Holyfield for years since you dumped KAV for Klitschko ????

[computerrank]

...
(3) - Why do you use as much as 14%? Right now you imply that HW 50s division is a few times better than current (ie. Rocky got 799 point = 60% of his total all-time points vs Wlad's 129 point = 10%of his total for this "14% of carreer high rating" which is based on the strength of the division). I'd say much lower than 14% would be more appropriate to disallow one-time success (maybe just luck) in any division mean more than long-time success in others. But lowering the percentage would still mean 50s HW div is better than current, when in reality it is hardly so.

Paradox: If Marciano faced a total bum in his last fight after Moore and got ko'ed by a lucky punch, his opponent would guarantee himself a place in like top-35 HWs of all time right above Vitali Kitschko and Ken Norton. He'd get to what - at least 3000 current rank points and all-time 200/1*33%+200/1*33% + 14%*3000 = 552. And how many of the HW boxers since 70s have got this many all-time points? Only several ATGs.

I see your points. I am not sure about the right way. The current all time ratings are experimental at all. The 14% weight was the result of a regression calculation with various all time lists. This value showed the least rank differences.

I really appreciate your comments and suggestions

[computerrank]

and the rating Holyfield can lose ???

Holyfield looses points - from 1312 down to 1304

please reset the ratings Holyfield for years since you dumped KAV for Klitschko ????

Code: Select all

Light Heavyweight    1984     5     0     0     5    499 Evander Holyfield
Cruiserweight        1985    42     8     0    34    499 Evander Holyfield
Cruiserweight        1986   234    74    66    94    499 Evander Holyfield
Cruiserweight        1987   372   140    82   150    499 Evander Holyfield
Cruiserweight        1988   461   206   104   150    499 Evander Holyfield
Heavyweight          1989   499   239   109   150    499 Evander Holyfield
Heavyweight          1990   632   305   175   151    499 Evander Holyfield
Heavyweight          1991   708   371   185   153    499 Evander Holyfield
Heavyweight          1992   749   404   192   153    499 Evander Holyfield
Heavyweight          1993   881   470   258   153    499 Evander Holyfield
Heavyweight          1994   912   492   267   153    499 Evander Holyfield
Heavyweight          1995   920   500   267   153    499 Evander Holyfield
Heavyweight          1996  1019   534   333   153    499 Evander Holyfield
Heavyweight          1997  1066   566   347   153    499 Evander Holyfield
Heavyweight          1998  1099   600   347   153    499 Evander Holyfield
Heavyweight          1999  1198   632   413   153    499 Evander Holyfield
Heavyweight          2000  1223   654   416   153    499 Evander Holyfield
Heavyweight          2001  1243   661   429   153    499 Evander Holyfield
Heavyweight          2002  1268   669   446   153    499 Evander Holyfield
Heavyweight          2006  1268   669   446   153    499 Evander Holyfield
Heavyweight          2007  1268   669   446   153    499 Evander Holyfield
Heavyweight          2008  1305   673   479   153    499 Evander Holyfield
Heavyweight          2010  1305   673   479   153    499 Evander Holyfield
Heavyweight          2011  1305   673   479   153    499 Evander Holyfield

[mike1989]

Code: Select all

Light Heavyweight    1984     5     0     0     5    499 Evander Holyfield
Cruiserweight        1985    42     8     0    34    499 Evander Holyfield
Cruiserweight        1986   234    74    66    94    499 Evander Holyfield
Cruiserweight        1987   372   140    82   150    499 Evander Holyfield
Cruiserweight        1988   461   206   104   150    499 Evander Holyfield
Heavyweight          1989   499   239   109   150    499 Evander Holyfield
Heavyweight          1990   632   305   175   151    499 Evander Holyfield
Heavyweight          1991   708   371   185   153    499 Evander Holyfield
Heavyweight          1992   749   404   192   153    499 Evander Holyfield
Heavyweight          1993   881   470   258   153    499 Evander Holyfield
Heavyweight          1994   912   492   267   153    499 Evander Holyfield
Heavyweight          1995   920   500   267   153    499 Evander Holyfield
Heavyweight          1996  1019   534   333   153    499 Evander Holyfield
Heavyweight          1997  1066   566   347   153    499 Evander Holyfield
Heavyweight          1998  1099   600   347   153    499 Evander Holyfield
Heavyweight          1999  1198   632   413   153    499 Evander Holyfield
Heavyweight          2000  1223   654   416   153    499 Evander Holyfield
Heavyweight          2001  1243   661   429   153    499 Evander Holyfield
Heavyweight          2002  1268   669   446   153    499 Evander Holyfield
Heavyweight          2006  1268   669   446   153    499 Evander Holyfield
Heavyweight          2007  1268   669   446   153    499 Evander Holyfield
Heavyweight          2008  1305   673   479   153    499 Evander Holyfield
Heavyweight          2010  1305   673   479   153    499 Evander Holyfield
Heavyweight          2011  1305   673   479   153    499 Evander Holyfield

on a brush Holyfield rating:
1985-1988 year annual rank points without taking into account the cluster;
1996 annual rank opponent has not exactly +66 points (first was Ridik Bowie and not Mike Tyson);
in 2008, Holyfield lost to Valuev because annual rank opponent has +0 and 33. ???
in 2008, Holyfield allowed annual rating is not for top 10 ???

[computerrank]

on a brush Holyfield rating:
1985-1988 year annual rank points without taking into account the cluster;
1996 annual rank opponent has not exactly +66 points (first was Ridik Bowie and not Mike Tyson);
in 2008, Holyfield lost to Valuev because annual rank opponent has +0 and 33. ???
in 2008, Holyfield allowed annual rating is not for top 10 ???

- 1996: for opponent rank #1 and #2 give same points
- 2008: it wasn't a clear loss against Valuew, it was as MD
- 2008: I changed it to top 20 for annual rank and annual best opponent rank. Holyfield got 4 points - so he was around #16

- 1985-1988 year annual rank points without taking into account the cluster; <= I don't understand here, what you mean really - but cluster ratings are applied

[Leonid]

BTW I really like your current ratings in each division, which are underappreciated on the forum. But clearly an all-time rating system pose a bigger challenge. Regression analysis, huh? I see you've given these rating systems a thought. Still there's a way to go, since my hypothetical example have many real instances, like The Cinderella Man Jim Braddock being a top-20 all-time HW mostly due to his single win over Max Baer. There's probably not one person on this forum who would agree with that. Is it too much trouble to produce a top-20 HW list that would be a result of exclusion of this 14% factor from the ratings at all?

[computerrank]

BTW I really like your current ratings in each division, which are underappreciated on the forum. But clearly an all-time rating system pose a bigger challenge. Regression analysis, huh? I see you've given these rating systems a thought. Still there's a way to go, since my hypothetical example have many real instances, like The Cinderella Man Jim Braddock being a top-20 all-time HW mostly due to his single win over Max Baer. There's probably not one person on this forum who would agree with that. Is it too much trouble to produce a top-20 HW list that would be a result of exclusion of this 14% factor from the ratings at all?

I will do a test run without the 14% ...

[mike1989]

- 1996: for opponent rank #1 and #2 give same points
it's understood )
- 2008: it wasn't a clear loss against Valuew, it was as MD
as Holyfield lost to Valuev ? how +33 ?? )
- 2008: I changed it to top 20 for annual rank and annual best opponent rank. Holyfield got 4 points - so he was around #16
understand )... and then where Klitschko points for 2004 ???
- 1985-1988 year annual rank points without taking into account the cluster; <= I don't understand here, what you mean really - but cluster ratings are applied

over these years, Holyfield has received +66 for each year, as I understand that would get +66 need to be No. 1 in the cluster, not the weight ????

[mike1989]

- 1996: for opponent rank #1 and #2 give same points
- 2008: it wasn't a clear loss against Valuew, it was as MD
- 2008: I changed it to top 20 for annual rank and annual best opponent rank. Holyfield got 4 points - so he was around #16
- 1985-1988 year annual rank points without taking into account the cluster; <= I don't understand here, what you mean really - but cluster ratings are applied

it's understood )
as Holyfield lost to Valuev ? how +33 ?? )
understand )... and then where Klitschko points for 2004 ???
over these years, Holyfield has received +66 for each year, as I understand that would get +66 need to be No. 1 in the cluster, not the weight ????

[computerrank]

Pacmans all time rating is showing as his current points total in error

This is a bug in the presentation, yes.

All active boxers are listed with their current rating, but the ranking is based on the all time rating points actually. The right value should be displayed ...

[computerrank]

BTW I really like your current ratings in each division, which are underappreciated on the forum. But clearly an all-time rating system pose a bigger challenge. Regression analysis, huh? I see you've given these rating systems a thought. Still there's a way to go, since my hypothetical example have many real instances, like The Cinderella Man Jim Braddock being a top-20 all-time HW mostly due to his single win over Max Baer. There's probably not one person on this forum who would agree with that. Is it too much trouble to produce a top-20 HW list that would be a result of exclusion of this 14% factor from the ratings at all?

I will do a test run without the 14% ...

... top 100 all time heavweights ordered by all time rating without top rating share ...

all_time_heavies_no_top_rating.pdf

[mike1989]

on a brush Holyfield rating:
1985-1988 year annual rank points without taking into account the cluster;


- 1985-1988 year annual rank points without taking into account the cluster; <= I don't understand here, what you mean really - but cluster ratings are applied

All Time ("Inactive") Ratings
The All Time Rating for a boxer is the sum of 33% of the sum of annual rank points he gets for his annual ratings - and 33% of the sum of annual rank points of best defeated opponents - and 14% of his career top rating:
the annual rating is the rating at the end of every year the boxer was active
annual rank points = 200 / annual rank in division cluster

Holyfield gained +66 points was the first in the kruzerveyt but not in the cluster kruzerveyt+havivah ???

[Leonid]

... top 100 all time heavweights ordered by all time rating without top rating share ...

all_time_heavies_no_top_rating.pdf

Thanks a lot! I actually like it more than the current version. Marciano moved down, same as one hit wonders like Braddock and Johansson, while Lewis, Holmes, Tyson, Bowe, Holy, Johnson, Frazier, Dempsey, both Klitschkos moved up, and IMO deservingly so. At least several guys from this millenium not named Klitschko could make it to the top-100. That's all positive changes, 15 years in a hundred yr old sport deserve to be fairly represented.

Langford at #1 looks odd, but he is top 3 in any version. He had increadible longativity, but some of his points come from divisions very far from HW and even while fighting at HW he weighted less than today's middleweights do. He started at what - lightweigh? fought for welterweight title, continued at middleweight, so maybe there should be some coefficient applied to those points earned below his main cluster. All-time divisional ratings could have different number of points for a boxer than in his p4p all-time ratings, where such coefficient would not apply. Or old-time HWs should have cruiser-to-heavy coefficient applied when they and their opponents both fight under the CW limit of 200 pounds. Seems fair, since even CWs' fighting weight is above 200.

[computerrank]

on a brush Holyfield rating:
1985-1988 year annual rank points without taking into account the cluster;


- 1985-1988 year annual rank points without taking into account the cluster; <= I don't understand here, what you mean really - but cluster ratings are applied

All Time ("Inactive") Ratings
The All Time Rating for a boxer is the sum of 33% of the sum of annual rank points he gets for his annual ratings - and 33% of the sum of annual rank points of best defeated opponents - and 14% of his career top rating:
the annual rating is the rating at the end of every year the boxer was active
annual rank points = 200 / annual rank in division cluster

Holyfield gained +66 points was the first in the kruzerveyt but not in the cluster kruzerveyt+havivah ???

aha ... looks strange again ...

[mike1989]

aha ... looks strange again ...

)... and Klitschko will dositey annual ranking points for 2004. In 2008, Holyfield was the defeat of Valuev because for 2008 may not be a +33 ??? )

[computerrank]

All Time ("Inactive") Ratings
The All Time Rating for a boxer is the sum of 33% of the sum of annual rank points he gets for his annual ratings - and 33% of the sum of annual rank points of best defeated opponents - and 14% of his career top rating:
the annual rating is the rating at the end of every year the boxer was active
annual rank points = 200 / annual rank in division cluster

Holyfield gained +66 points was the first in the kruzerveyt but not in the cluster kruzerveyt+havivah ???

aha ... looks strange again ...

It was a bug ...

2008 he got 33 points for his close loss against Valuev.

Now it looks like this for Holyfield:

Code: Select all

Light Heavyweight    1984     5     0     0     5    499 Evander Holyfield
Heavyweight          1985    34     0     0    34    499 Evander Holyfield
Heavyweight          1986   173    13    66    94    499 Evander Holyfield
Heavyweight          1987   267    35    82   150    499 Evander Holyfield
Heavyweight          1988   323    68   104   150    499 Evander Holyfield
Heavyweight          1989   361   101   109   150    499 Evander Holyfield
Heavyweight          1990   494   167   175   151    499 Evander Holyfield
Heavyweight          1991   570   233   185   153    499 Evander Holyfield
Heavyweight          1992   611   266   192   153    499 Evander Holyfield
Heavyweight          1993   743   332   258   153    499 Evander Holyfield
Heavyweight          1994   774   354   267   153    499 Evander Holyfield
Heavyweight          1995   782   362   267   153    499 Evander Holyfield
Heavyweight          1996   881   395   333   153    499 Evander Holyfield
Heavyweight          1997   928   428   347   153    499 Evander Holyfield
Heavyweight          1998   961   461   347   153    499 Evander Holyfield
Heavyweight          1999  1060   494   413   153    499 Evander Holyfield
Heavyweight          2000  1085   516   416   153    499 Evander Holyfield
Heavyweight          2001  1104   522   429   153    499 Evander Holyfield
Heavyweight          2002  1128   530   446   153    499 Evander Holyfield
Heavyweight          2006  1128   530   446   153    499 Evander Holyfield
Heavyweight          2007  1128   530   446   153    499 Evander Holyfield
Heavyweight          2008  1165   533   479   153    499 Evander Holyfield
Heavyweight          2010  1165   533   479   153    499 Evander Holyfield
Heavyweight          2011  1165   533   479   153    499 Evander Holyfield

[mike1989]

Now it looks like this for Holyfield:

Code: Select all

Light Heavyweight    1984     5     0     0     5    499 Evander Holyfield
Heavyweight          1985    34     0     0    34    499 Evander Holyfield
Heavyweight          1986   173    13    66    94    499 Evander Holyfield
Heavyweight          1987   267    35    82   150    499 Evander Holyfield
Heavyweight          1988   323    68   104   150    499 Evander Holyfield
Heavyweight          1989   361   101   109   150    499 Evander Holyfield
Heavyweight          1990   494   167   175   151    499 Evander Holyfield
Heavyweight          1991   570   233   185   153    499 Evander Holyfield
Heavyweight          1992   611   266   192   153    499 Evander Holyfield
Heavyweight          1993   743   332   258   153    499 Evander Holyfield
Heavyweight          1994   774   354   267   153    499 Evander Holyfield
Heavyweight          1995   782   362   267   153    499 Evander Holyfield
Heavyweight          1996   881   395   333   153    499 Evander Holyfield
Heavyweight          1997   928   428   347   153    499 Evander Holyfield
Heavyweight          1998   961   461   347   153    499 Evander Holyfield
Heavyweight          1999  1060   494   413   153    499 Evander Holyfield
Heavyweight          2000  1085   516   416   153    499 Evander Holyfield
Heavyweight          2001  1104   522   429   153    499 Evander Holyfield
Heavyweight          2002  1128   530   446   153    499 Evander Holyfield
Heavyweight          2006  1128   530   446   153    499 Evander Holyfield
Heavyweight          2007  1128   530   446   153    499 Evander Holyfield
Heavyweight          2008  1165   533   479   153    499 Evander Holyfield
Heavyweight          2010  1165   533   479   153    499 Evander Holyfield
Heavyweight          2011  1165   533   479   153    499 Evander Holyfield

)
and what about the rating of Klitschko 2004, he is also in the top 20 ???

[mike1989]

can please even the rating of Max Schmeling to lose ???

[computerrank]

I dropped the component top rating completely ...

all_time_heavies_no_top_rating.pdf

[computerrank]

can please even the rating of Max Schmeling to lose ???

Max Schmeling now looks like this:

Code: Select all

Light Heavyweight    1927   854    17    13     4     0   9041 Max Schmeling
Light Heavyweight    1928   823    31    23     8     0   9041 Max Schmeling
Heavyweight          1929  1133    61    45    16     0   9041 Max Schmeling
Heavyweight          1930  2002   193   111    82     0   9041 Max Schmeling
Heavyweight          1931  2581   281   177   104     0   9041 Max Schmeling
Heavyweight          1932  2629   380   243   138     0   9041 Max Schmeling
Heavyweight          1933  1691   397   259   138     0   9041 Max Schmeling
Heavyweight          1934  1620   423   276   147     0   9041 Max Schmeling
Heavyweight          1935  2556   478   298   180     0   9041 Max Schmeling
Heavyweight          1936  4062   610   364   246     0   9041 Max Schmeling
Heavyweight          1937  2639   643   397   246     0   9041 Max Schmeling
Heavyweight          1938  1970   676   430   246     0   9041 Max Schmeling
Heavyweight          1939  2534   775   463   312     0   9041 Max Schmeling

[computerrank]

All time p4p look like this - no top rating regarded

all_time_p4p_no_top_rating.pdf

[Leonid]

All time p4p look like this - no top rating regarded

all_time_p4p_no_top_rating.pdf

Thanks! The changes are not as dramatic as one could predict, SRR and Moore are still top-2, but not by such a huge margin, and the more recent guys get a little more appreciation as they well deserve, while some not that great guys, especially from the 50s, go down. Who would have Jimmy Bivins as #21 p4p and Joey Maxim #33? They are barely making it to such positions in many subjectuve light-heavyweight all-time ratings.

I don't believe fighters were better back in the day, with their caveman training regimes and supplements and current system gives many of them quite a benefit compared to today's crop.

Therefore I like this rating better as well and think this should be the official system. Does anyone else support me in this?

[mike1989]

can please even the rating of Max Schmeling to lose ???

Max Schmeling now looks like this:

Code: Select all

Light Heavyweight    1927   854    17    13     4     0   9041 Max Schmeling
Light Heavyweight    1928   823    31    23     8     0   9041 Max Schmeling
Heavyweight          1929  1133    61    45    16     0   9041 Max Schmeling
Heavyweight          1930  2002   193   111    82     0   9041 Max Schmeling
Heavyweight          1931  2581   281   177   104     0   9041 Max Schmeling
Heavyweight          1932  2629   380   243   138     0   9041 Max Schmeling
Heavyweight          1933  1691   397   259   138     0   9041 Max Schmeling
Heavyweight          1934  1620   423   276   147     0   9041 Max Schmeling
Heavyweight          1935  2556   478   298   180     0   9041 Max Schmeling
Heavyweight          1936  4062   610   364   246     0   9041 Max Schmeling
Heavyweight          1937  2639   643   397   246     0   9041 Max Schmeling
Heavyweight          1938  1970   676   430   246     0   9041 Max Schmeling
Heavyweight          1939  2534   775   463   312     0   9041 Max Schmeling

what's wrong with it ??? )

[computerrank]

can please even the rating of Max Schmeling to lose ???

Max Schmeling now looks like this:

Code: Select all

Light Heavyweight    1927   854    17    13     4     0   9041 Max Schmeling
Light Heavyweight    1928   823    31    23     8     0   9041 Max Schmeling
Heavyweight          1929  1133    61    45    16     0   9041 Max Schmeling
Heavyweight          1930  2002   193   111    82     0   9041 Max Schmeling
Heavyweight          1931  2581   281   177   104     0   9041 Max Schmeling
Heavyweight          1932  2629   380   243   138     0   9041 Max Schmeling
Heavyweight          1933  1691   397   259   138     0   9041 Max Schmeling
Heavyweight          1934  1620   423   276   147     0   9041 Max Schmeling
Heavyweight          1935  2556   478   298   180     0   9041 Max Schmeling
Heavyweight          1936  4062   610   364   246     0   9041 Max Schmeling
Heavyweight          1937  2639   643   397   246     0   9041 Max Schmeling
Heavyweight          1938  1970   676   430   246     0   9041 Max Schmeling
Heavyweight          1939  2534   775   463   312     0   9041 Max Schmeling

what's wrong with it ??? )

Nothing is wrong for me, you wanted to see it, or?