Ratings - please read before commenting - Archived

[jujigatame]

Promoter advantage: ruining computerized boxing rankings since 2008!!

[computerrank]

Release N is launched - changes:

1. Bug fix for close decisions. In some cases the winner got less points for a clearer decision than for a closer decision.

2. Upset loser lost up to 70% of his points - such as a boxer with promoter advantage losing clearly by points.

It is more systematic to use the same ratio now. If the winner gets 70% (factor 1.7), the loser goes down to 100%/1.7 = 59% and loses 41%.

3. The prebout rating of a boxer coming back successfully from inactivity is now set up to the opponents prebout rating, depending on the returning boxer's prior rating (see Vitali Klitschko).

[jujigatame]

You now have Munoz #1 at 115 despite coming off a loss, and Hasegawa slipping farther down the 118 rankings despite being the clear #1 in the division. This change also gave boosts to Roy Jones (who is now above Johnson and Tarver) and Julio Diaz who has been boosted above Juan Diaz.

Other delights include Raul Garcia being pushed down to #7 at 105, and Molitor getting boosted way back up into the top 5, surpassing even Juanma Lopez. Also Wayne Elcock has gotten a major boost into the MW top 5 for god knows what reason, while Sylvester has been pushed back to the outskirts of the top 10 below Elcock, Castillejo, and Gevor.

What is the purpose of all this? It seems like you've been working feverishly for months now but every "improvement" has just caused the rankings to get uglier and uglier. All you've accomplished is reduced transparency and credibility for a once very solid system.

[Grant]

The Boxrec ratings are very handy and better than most, but an improvement would be IMO that any fighter who fights someone more than say 20 places beneath him or her should not accrue points.
Also, I have posted this elsewhere, It would be good to see the fighters positions (at the time of the fight) against their names in the record section.
It would mean that looking back on Mundines record in 20 years would paint a better image of his career rather than bullshit titles and 25 and 0 records

[computerrank]

jujigatame,

all these changes are just fixes - transparent and comprehensible to every capable person.

I see, the issue you cannot accept is the ratings taking into account the home advantage.

The effect of home advantage is clearly identified in sports science research and even obvious to the public. It drives the effort of every manager and boxer to get the bouts to the home turf.

The methods of implementation in a rating (parameters, grade of compensation) maybe arguable. But there is no way to say, there is no such impact.

We decided to compensate for this effect, as further looking away would be a gross injustice.

We decided to use the criteria home promoter, as this effect could be clearly and significantly identified in our bouts data base. And we compensated for it by 50% of the measured impact, to be sure not to overcompensate.

[computerrank]

The Boxrec ratings are very handy and better than most, but an improvement would be IMO that any fighter who fights someone more than say 20 places beneath him or her should not accrue points.
Also, I have posted this elsewhere, It would be good to see the fighters positions (at the time of the fight) against their names in the record section.
It would mean that looking back on Mundines record in 20 years would paint a better image of his career rather than bullshit titles and 25 and 0 records

getting points

This concept is implemented already. A boxer will not get points for even clearly defeating an opponent lower than 25% of his own rating. For closer decisions this limit is even higher.

records with pre-bout ratings

This conecpt is already implemented for internal analysis. Currently there are deliberations to integrate it into the public record pages.

[Grant]

The Boxrec ratings are very handy and better than most, but an improvement would be IMO that any fighter who fights someone more than say 20 places beneath him or her should not accrue points.
Also, I have posted this elsewhere, It would be good to see the fighters positions (at the time of the fight) against their names in the record section.
It would mean that looking back on Mundines record in 20 years would paint a better image of his career rather than bullshit titles and 25 and 0 records

getting points

This concept is implemented already. A boxer will not get points for even clearly defeating an opponent lower than 25% of his own rating. For closer decisions this limit is even higher.

records with pre-bout ratings

This conecpt is already implemented for internal analysis. Currently there are deliberations to integrate it into the public record pages.

Dopey me....thanks

[jujigatame]

jujigatame,

all these changes are just fixes - transparent and comprehensible to every capable person.

I see, the issue you cannot accept is the ratings taking into account the home advantage.

The effect of home advantage is clearly identified in sports science research and even obvious to the public. It drives the effort of every manager and boxer to get the bouts to the home turf.

The methods of implementation in a rating (parameters, grade of compensation) maybe arguable. But there is no way to say, there is no such impact.

We decided to compensate for this effect, as further looking away would be a gross injustice.

We decided to use the criteria home promoter, as this effect could be clearly and significantly identified in our bouts data base. And we compensated for it by 50% of the measured impact, to be sure not to overcompensate.

If these are simple fixes, then it simply reveals your home promoter implementation to be even more problematic than previously thought.

The biggest problem is that you are attempting to quantify something that you do not fully understand. You have taken some very general results from a sports science paper and attempted to apply them in a grossly over-reaching, over-simplified way. There are most likely dozens if not hundreds of variables that could go into determining the presence and magnitude of the advantage a home fighter has. I'm sure statisticians could spend years attempting to figure out all of the various correlations. You have reduced this to essentially 2 variables because you do not have the knowledge required to do a more thorough examination. However, despite the fact that the information you're using is quite limited, you have still allowed this to have an extremely sizable effect on the rankings. Considering the extremely simplified approach you are using, the effect of the home advantage should be reduced by much more than 50%. Something more like a 5-10% effect would be appropriate.

Can you at least agree that a system that places Alex Munoz at #1 coming off of a loss needs further examination and tuning?

[jujigatame]

We decided to compensate for this effect, as further looking away would be a gross injustice.

One more response to this particular statement. Do you think it is a gross injustice that in each football match, the visiting team is not spotted 1 goal? My point is that most sports can agree that some small measure of home advantage exists, but they also recognize that any attempts to compensate for this would be misguided and likely ineffective.

[computerrank]

We decided to compensate for this effect, as further looking away would be a gross injustice.

One more response to this particular statement. Do you think it is a gross injustice that in each football match, the visiting team is not spotted 1 goal? My point is that most sports can agree that some small measure of home advantage exists, but they also recognize that any attempts to compensate for this would be misguided and likely ineffective.

The games are regularly matched as double tests - equally distributed at both parties' home turf. This would be the best way for boxing too, but it seems not to be possible for some reasons.

Yes, there is a big home advantage in all these games, look at the statistics.

The same is found for boxing. The statistics say, the home promoted boxers won 75% of all decisions on points - regarding equally rated boxers only. This is not a neglectable margin. Otherwise we wouldn't discuss this.

The misguiding is given by boxers hiding on their home turf - and failing as soon as they leave it.

The really best prove themselves on foreign terrain too.

[JCS]

Martin,

I think juji is trying to represent the fact that some promoters may have a greater or lesser effect than others, so it would be unfair to use the same statistic for every situation.

[computerrank]

Martin,

I think juji is trying to represent the fact that some promoters may have a greater or lesser effect than others, so it would be unfair to use the same statistic for every situation.

There are some variations, sure, although the research results state, that the advantage seems to be unavoidable to some amount.

But even if there is some amount of variation, it would imply more injustice to ignore the impact, than to compensate some inherent part of the advantage, as we do by compensating for 50% of the mean observed value.

Assuming a distribution (regarding the findings above)

- nearly every home bout shows an advantage of 50% of the mean value (assumed as the unavoidable part)
- up to 50% of the bouts show an advantage up to the mean value
- and the rest shows an advantage of more than this mean value

If we compensate for this unavoidably occuring, reasonable 50% of the mean observed advantage, the correction can be exspected to never go into the wrong direction - at least not by a noteworthy amount.

[jujigatame]

There are many variables you are not taking into account:

Differences between promoters
Differences between countries/states
Relationship between home/away promoter
Contracts stipulating options on home/away fighter
Popularity of fighters
Television coverage

Additionally, you are attempting to quantify the advantage in a way that the original study does not attempt. If in 50/50 fights the home fighter wins by a ratio of 75/25, that does not necessarily indicate a quantifiable advantage of 50%. In fact, it doesn't mean anything even close to that. Remember a win or loss in a boxing match can be the matter of 1 point on the scorecards. If a judge sways the decision of a fight 75% of the time to the home fighter, he may only be averaging a sway of 1-2 points per fight. This is something you need to consider.

Also, I am curious about the following questions:

How does your definition of "home fighter" compare to the study's definition?

How does the study define a 50/50 fight? Has it occurred to you that a home promoter would seek out fights for his fighter against overrated opponents, or ones that provide a favorable style matchup?

What is your basis for assuming that 50% of the presumed advantage is inherent? This seems very arbitrary.

Do you think the current system has issues that need to be resolved? If so, what are they? Doesn't it make you uneasy to be basing such large ranking swings on so little information?

[BobbyDobbs]

Can someone explain to me why Carson Jones' rating went from 72 to 109 overnight?

[computerrank]

jujigatame - just commenting your post

There are many variables you are not taking into account:

Differences between promoters
Differences between countries/states
Relationship between home/away promoter
Contracts stipulating options on home/away fighter
Popularity of fighters
Television coverage

-> Yes, I am sure the impacts are multiple and correlated . But designing a rating, you have to keep things operable. The home promoter factor was identified as an important and strongly significant one. An other option was boxer nationality/show country, and this factor is significant too. But the current choice appears more related to business practice.

Additionally, you are attempting to quantify the advantage in a way that the original study does not attempt. If in 50/50 fights the home fighter wins by a ratio of 75/25, that does not necessarily indicate a quantifiable advantage of 50%. In fact, it doesn't mean anything even close to that. Remember a win or loss in a boxing match can be the matter of 1 point on the scorecards. If a judge sways the decision of a fight 75% of the time to the home fighter, he may only be averaging a sway of 1-2 points per fight. This is something you need to consider.

-> This was tested, and didn't change the results. Even KO/TKO results show a ratio of 2.3:1 for the home promoter boxer.

Also, I am curious about the following questions:

How does your definition of "home fighter" compare to the study's definition?

-> home boxer >= 4 bouts with show promoter, opponent <2; study is by boxer nationality

How does the study define a 50/50 fight? Has it occurred to you that a home promoter would seek out fights for his fighter against overrated opponents, or ones that provide a favorable style matchup?

-> Boxrec pre-bout-rating, both boxers >=50 points; study by win ratio

What is your basis for assuming that 50% of the presumed advantage is inherent? This seems very arbitrary.

-> This is just an explanation model, might be more or less

Do you think the current system has issues that need to be resolved? If so, what are they? Doesn't it make you uneasy to be basing such large ranking swings on so little information?

-> Home promoter is no way little information. It is an important factor. The whole business is built around this factor. And the participants very well know, why.

So massive effects were exspected.

[BobbyDobbs]

Can someone explain to me why Carson Jones' rating went from 72 to 109 overnight?

Please respond

[JCS]

Can someone explain to me why Carson Jones' rating went from 72 to 109 overnight?

Please respond

There were recently bugfixes and minor modifications, as seen earlier in the thread. The lower a fighter's rating is, the more volatile it is.

Furthermore, there's an 18 month sliding window that looks at a fighter's performances within that window to see if they are still proving their current rating, within reason.

Promoter data could have been updated/inserted, which also now has an effect on the ratings.

[BobbyDobbs]

Thanks for explaining.

[computerrank]

@ Carson Jones

This was all due to the general rating correction - see Jason's.

His rating grew step by step till now. And his defining bout was on 2008-04-18 against Gonzalez.

So at this rating level there is no risk for missing opponent's quality until 2009-10-17.

rating values at the end of every line:
- 1st boxer - before bout
- 1st boxer - after bout
- 2nd boxer - before bout
- 2nd boxer - after bout

Rating values are changed by daily varied adaption factor for all boxer in a division - so the relative ranking is not touched by this.

Code: Select all

2004-10-05 Light Middleweight   Jones                Mosley               W  TKO     0    18     6     4
2004-12-10 Middleweight         Jones                Molton               W  TKO    15    15     0     0
2005-01-13 Light Middleweight   Jones                Vincent              W  TKO    15    76    52    30
2005-04-15 Middleweight         Jones                Sherman              NC ND      0     0     0     0
2005-06-16 Light Middleweight   Jones                Medina               D  PTS    76    67    22    25
2005-07-22 Light Middleweight   Jones                Cunningham           W  TKO    67    67     0     0
2005-08-02 Light Middleweight   Jones                Castaneda            W  SD     67    82    86    79
2005-10-04 Middleweight         Jones                Smith                W  UD     76    94    51    48
2005-10-27 Light Middleweight   Jones                Medina               L  UD     94    64    25    68
2005-11-18 Light Middleweight   Jones                Johnson              W  UD     64    87    46    41
2006-02-02 Middleweight         Jones                Townsend             W  MD     87   105    45    42
2006-02-17 Light Middleweight   Jones                Perez                L  TKO   105    78    98   183
2006-04-26 Light Middleweight   Jones                Capo                 W  TKO    78    78     7     7
2006-05-06 Light Middleweight   Jones                Hill                 W  KO     78    78     0     0
2006-06-15 Light Middleweight   Jones                Carpenter            W  TKO    78   121    40    35
2006-07-21 Light Middleweight   Jones                Flanagan             W  UD    121   121    10    10
2006-08-25 Light Middleweight   Jones                Gomez                L  TKO   121   113   349   374
2006-10-27 Light Middleweight   Jones                Hernandez            L  SD    113   143   330   262
2007-06-08 Light Middleweight   Jones                Garcia               L  UD    143   143   366   366
2008-01-04 Light Middleweight   Jones                Gonzalez             W  MD    143   169   182   165
2008-01-19 Light Middleweight   Jones                Gray                 L  UD    169   123    70   130
2008-04-18 Super Middleweight   Jones                Gonzalez             W  UD    105   184   156   115
2008-07-22 Middleweight         Jones                Huskey               W  TKO   184   184     0     0
2008-09-20 Middleweight         Jones                Castaneda            W  UD    184   184    24    24
2008-11-24 Light Middleweight   Jones                Hinkle               W  TKO   184   184     0     0
2009-01-16 Light Middleweight   Jones                Divison              W  TKO   184   184    28    28
2009-02-06 Light Middleweight   Jones                Soto Karass          SC NC      0     0     0     0

[BobbyDobbs]

And if her were to hypothetically defeat Soto-Karass what would that do to his rating?

[JCS]

Depends on the outcome. If it was a dominant win, I'd expect Jones to move up to about #25-30... but Martin can give you a specific answer.

[computerrank]

Depends on the outcome. If it was a dominant win, I'd expect Jones to move up to about #25-30... but Martin can give you a specific answer.

... about #35 for a clear win

[BobbyDobbs]

Thanks.

[El Raincoat]

What is the ranking's all about on the home page???

Thsi is crazy.

[jujigatame]

I've been bitching about this for months. It's the absurdly inflated "home promoter advantage" in effect. Munoz basically gets massive credit for only losing a close decision against Mijares.

Post on the ratings thread about it. If we get enough popular support maybe we can get it changed.