Ratings - please read before commenting - Archived

emile · Post by **emile** » 09 Oct 2006, 12:00

Seems to me some sort of gradual time diminishment off all results would be an eventual improvement. The problem with Gainer is that he hasn't won a useful fight in three years. Whether the time cut off is 15 or 18 months, someone could maintain their high ranking by having a pitty-pat fight every once in a long while. Gainer should ideally be dropping in the rankings as his last decent win gets farther and farther back, with nothing to replenish those points.

JCS · Post by **JCS** » 09 Oct 2006, 12:18

emile wrote:Seems to me some sort of gradual time diminishment off all results would be an eventual improvement. The problem with Gainer is that he hasn't won a useful fight in three years. Whether the time cut off is 15 or 18 months, someone could maintain their high ranking by having a pitty-pat fight every once in a long while. Gainer should ideally be dropping in the rankings as his last decent win gets farther and farther back, with nothing to replenish those points.

Well, I think the ratings now are made to mock those of sanctioning bodies, the Ring, etc. You can bet that they don't sit there with a calculator and gradually decay a fighter's rating based on his last bout.

A gradual step down that takes place every 6 months or so probably wouldnt be a bad idea... In my testing, I found 15 months better than 18 though... at a smaller penalty.

emile · Post by **emile** » 09 Oct 2006, 12:31

The ratings are not meant to mock those of the sanctioning bodies, obviously

As for The Ring, they may not use a calculator to diminish rankings, but they also don't have Derrick Gainer in the top 10. I'd assume they've taken his inactivity into account since, as Martin pointed out, his only 'recent' losses were to two of the best at their weight class.

computerrank · Post by **computerrank** » 09 Oct 2006, 13:11

The quality ensuring rule says, that every boxer has to meet a boxer with a rating within 1/4 of his own rating within 18 months time.

As Gainer did this last time with Chris John 2005-04-22 Chris John will lose 1/4 of his points on 2006-10-22.

Then he will be set back by additional 187 points.

JCS · Post by **JCS** » 09 Oct 2006, 18:58

emile wrote:The ratings are not meant to mock those of the sanctioning bodies, obviously

As for The Ring, they may not use a calculator to diminish rankings, but they also don't have Derrick Gainer in the top 10. I'd assume they've taken his inactivity into account since, as Martin pointed out, his only 'recent' losses were to two of the best at their weight class.

Its meant to do so in regards to some of the basic principles it follows.

JCS · Post by **JCS** » 10 Oct 2006, 15:23

Martin..

How did Chris John suddenly fall below JMM? It doesnt make sense he would lose points anytime soon due to lack of opponent quality because of JMM.

computerrank · Post by **computerrank** » 10 Oct 2006, 15:46

JCS83MD wrote:Martin..

How did Chris John suddenly fall below JMM? It doesnt make sense he would lose points anytime soon due to lack of opponent quality because of JMM.

There was a major bug in the ratings implementation. It caused jumps over 25% from day to day.

Now they are stable. The bug fix was launched on Sunday.

JCS · Post by **JCS** » 10 Oct 2006, 20:10

Martin.. Small recommendation

In cases such as Chris John beating Juan Manuel Marquez, I think that most would agree that John should still be ranked above Marquez, regardless of the fact that Juan recently beat Terdsak.

It got me thinking.

First, I am running on the assumption that John was ranked much lower than JMM, so when John won, he was placed just slightly above Marquez in points. I am now thinking in a case like this when a win is decisive, that the lower-ranked fighter should be rated a minimum amount of points ahead of the loser. Perhaps 10 or 15% higher?? Or maybe it should depend on the result.

For example (Not using exact calculations here)

Say John (500pts) UD Marquez (1200pts)

After the match.. their ratings would probably look something like

John (1051) - Marquez (1050)

Instead, I think John should be given some clearance over Marquez

10% would put John at 1156 and Marquez at 1050

Does that make sense?

In addition, I'm also interested to see the hometown advantage being increased slightly as there are some strictly foreign fighters ranked too high, and also I'd like to view results of inactivity decreases starting at 15 months (not 18) (40% not 50%)

computerrank · Post by **computerrank** » 12 Oct 2006, 10:09

@Jason,

1. Set upset winner above loser by a margin

In case of an upset, the winner is set above the normal win level already. So I don't think it makes sense to protect him longer than this bout.

2. Inactivity

I will let the time period with 18 months. The considerable point deductance of 50% should really come only with such clear inactivity. This time period is in line with the time period of opponent quality requirement, which deducts the half the value of 25%. Say: boxed against low level opposotion is not as bad as not boxing anyway.

Things must be kept simple too.

3. Home advantage

It is already at 40%, which is quite high, I think.

example:

It assumes a home boxer with a neutral rating of 500 at the same level as a foreign boxer with rating 1000.

home boxer = 500*1.4 = 700
foreign boxer = 1000/1.4 = 700

Best regards
Martin

JCS · Post by **JCS** » 12 Oct 2006, 16:25

computerrank wrote:@Jason,

1. Set upset winner above loser by a margin

In case of an upset, the winner is set above the normal win level already. So I don't think it makes sense to protect him longer than this bout.

2. Inactivity

I will let the time period with 18 months. The considerable point deductance of 50% should really come only with such clear inactivity. This time period is in line with the time period of opponent quality requirement, which deducts the half the value of 25%. Say: boxed against low level opposotion is not as bad as not boxing anyway.

Things must be kept simple too.

3. Home advantage

It is already at 40%, which is quite high, I think.

example:

It assumes a home boxer with a neutral rating of 500 at the same level as a foreign boxer with rating 1000.

home boxer = 500*1.4 = 700
foreign boxer = 1000/1.4 = 700

Best regards
Martin

In reference to #1.. I am not saying to give him protection after the bout. I am merely saying to make sure an upset win ranks the winning fighter a certain amount above the losing fighter post-bout.

For instance.

If Tye Fields were to TKO Klitschko, the end result would most likely have Fields only 1 point above Klitschko.. just enough to give him the higher rank. My argument is that since it was a TKO, Fields should be a fair amount higher than Klitschko post-bout.. because it is obvious (at least on that night) he was much better.

On #2, I too found a problem with Derrick Gainer being ranked so high. The fact that we wait 18 months to do any type of deductions has to be considered a flaw. Smaller, more frequent deductions would probably result in a much better ranking for a fighter such as Gainer.

And I did not realize hometown factor was so high. I simply found fighter's rankings like Pongsaklek, Lorenzo Parra, and Omar Narvarez too high.

thismodernlove · Post by **thismodernlove** » 16 Oct 2006, 03:29

i have a problem with how Wlad Klitschko is rated ahead of Lamon Brewster

no offence to BoxRec but to me that makes ZERO FCKING sense.

thismodernlove · Post by **thismodernlove** » 16 Oct 2006, 03:31

there should be a rule in BoxRec rankings which states that if Fighter A beats Fighter B decisively, like a KO or TKO, (as in the case of Brewster/Klitschko) then no matter what Fighter B accomplishes, for say, 6 bouts, Fighter B must be rated below Fighter A.

JCS · Post by **JCS** » 16 Oct 2006, 08:59

thismodernlove wrote:there should be a rule in BoxRec rankings which states that if Fighter A beats Fighter B decisively, like a KO or TKO, (as in the case of Brewster/Klitschko) then no matter what Fighter B accomplishes, for say, 6 bouts, Fighter B must be rated below Fighter A.

You realize Lamon lost to Sergei while Wlad beat the recognized #1 heavyweight. I agree with your comment in some instances, but it becomes quite complicated over time.

IF you were trying to make a point, it should've probably been w/ Chris John and Juan Manuel Marquez or Barrera and Pacquiao.

thismodernlove · Post by **thismodernlove** » 17 Oct 2006, 03:07

JCS83MD wrote:
thismodernlove wrote:there should be a rule in BoxRec rankings which states that if Fighter A beats Fighter B decisively, like a KO or TKO, (as in the case of Brewster/Klitschko) then no matter what Fighter B accomplishes, for say, 6 bouts, Fighter B must be rated below Fighter A.
You realize Lamon lost to Sergei while Wlad beat the recognized #1 heavyweight. I agree with your comment in some instances, but it becomes quite complicated over time.

IF you were trying to make a point, it should've probably been w/ Chris John and Juan Manuel Marquez or Barrera and Pacquiao.

i donno man its just that to me if one guy KO's another guy it doesnt make sense that the guy who got KO'd is rated ahead of the other guy like only a few fights later...?

but guess ur right cause klits beat byrd... but still i dont agree that hes rated ahead of brewster.

JCS · Post by **JCS** » 20 Oct 2006, 09:14

Cobwebcat wrote:Tommy Gibbons highest ranking of any boxer ever?

Monzon twice as good as SRR?

Daft beyond belief

All-time points displayed are ONLY for the division that the fighter is currently assigned to. So.. according to the system, Tommy Gibbons has more division points than anyone else. I'm trying to formulate a statement to accurately reflect that..

The current ratings work the same way. So the P4P system isn't really P4P, its just got fighters ranked based on their current divisional points. Although obviously, if a fighter changes divisions ,some of their points follow them.

JCS · Post by **JCS** » 20 Oct 2006, 10:31

Cobwebcat wrote:That's ridiculous. No wonder Hatton has disappeared from the P4P list.

I can see the logic but it makes for a really stupid list.

That's very true. Logic suggests JohnShep/Martin need to do one of two things:

1. Remove the P4P List from the main page
2. Create a P4P database field,make a true P4P calculation to fill the field, and direct the main page P4P list to that field only.

computerrank · Post by **computerrank** » 20 Oct 2006, 14:44

1. If a boxer changes his division, his rating is changed in relation to the square of weight limits of the divisions.

2. P4p and all time ratings: the ratings are normalized to divisions' #9, #10 and #11 average ratings - this rating is set to 500. But the adaption factor may not be larger than 3.33.

3. Yes, p4ps base upon current top normalized ratings in different weight divisions. This has nothing to do with belts, unifications and other glamour.

3. This normalization compensates for different populations of different weight divisions. Higher populated divisions were in advantage before.

4. The normalization is adapted on an annual basis - so all time ratings are touched too.

5. This focuses on excellence (exellence over average) of boxers in every weight division and compensates for some historical deficencies in records (low number of registrated boxers in divisions impacts achievable rating level).

So be careful with words like ridiculous ...

Overall, these p4ps and all time ratings look better than the ratings before. Look at the old boys in heavyweight.

I also saw and see a case like Tommy Gibbons in light heavyweight. But he has tremendous ratings from 1917 to 1920 and 1924 in this weight division of over 4000 points. So, who really knows.

It makes no sense for me to base the discussion on single cases.

JCS · Post by **JCS** » 23 Oct 2006, 09:14

The P4P lists should not display ratings that have been normalized in any way that would negatively impact the ratings. Realistically, I think you could argue for normalization that would benefit a fighter who successfully moves decisions while continuing to win, which helps this fighter prove the term "pound for pound"

JCS · Post by **JCS** » 23 Oct 2006, 11:26

Cobwebcat wrote:On the Cruiserweight all time list Hearns is only 27 behind people like Glen Mcrory???!!!!

Either dont penalise fighters for going thru the weights (p4p surely all about this) or rank them in the division in which they had most of their fights, not the division where they fought last!!!!

Sorry to appear negative but the all time and p4p are easily the worst they have ever looked in the last 24 months because of this.

Thomas Hearns is still considered "active"

JCS · Post by **JCS** » 31 Oct 2006, 22:58

Martin,

Got a suggestion in regards to the inactivity drop.

Instead of dropping a fighter by an automatic 25% for not facing quality opposition.. how about using that average opposition in the last 18 months to determine their new rating in someway. Seems more "correct" than simply dropping them by 25%. This way its based on an exact representation.

computerrank · Post by **computerrank** » 01 Nov 2006, 06:02

JCS83MD wrote:Martin,

Got a suggestion in regards to the inactivity drop.

Instead of dropping a fighter by an automatic 25% for not facing quality opposition.. how about using that average opposition in the last 18 months to determine their new rating in someway. Seems more "correct" than simply dropping them by 25%. This way its based on an exact representation.

Jason,

the perspective of missing quality is not to adapt the ratings continiously related the opponents' quality. This is already done with the normal rating procedure.

It is a simple solution for the issue of opponents missing a limit of 25% of the own rating. The related rating reduction of 25% after 18 months is 1/3 of this gap.

In this limit rule it doesn't count whether the opponents are at level 25% or far below.

And there is no reduction (so no smoothing) for best opponents inside the margin of 25% and 100%. This just represents the legal value band of opponents.

There is a clear limit - and a clear reduction, if the limit is violated.

I think this is fair - and things shouldn't be too complicated.

Best regards
Martin

JCS · Post by **JCS** » 06 Nov 2006, 11:07

Everytime the ratings get updated they seem to act a little more like the old prediction system :)

computerrank · Post by **computerrank** » 06 Nov 2006, 11:34

... it was only a bug fix for 2 bugs ...

- home advantage was not calculated correctly, if boxer B won
- rating change was not calculated correctly, if decision was a draw or close decision

... but it really looks better, by the way ...

JCS · Post by **JCS** » 06 Nov 2006, 11:37

computerrank wrote:... it was only a bug fix for 2 bugs ...

- home advantage was not calculated correctly, if boxer B won
- rating change was not calculated correctly, if decision was a draw or close decision

... but it really looks better, by the way ...

you're making all the brits mad since Hatton isnt in Top 25 P4P!!

jujigatame · Post by **jujigatame** » 08 Nov 2006, 10:34

A couple of related notes:

1) JMM is above Chris John despite John remaining undefeated and JMM only beating Terdsak Jandang since losing to him.

2) Vlad Virchis is above Chagaev despite Chagaev remaining undefeated and Virchis only beating Vidoz since losing to him.

If you're not trying to make a predictive system anymore, I feel like you should be required to beat more than some lower-level guy you were expected to beat anyway to vault someone over someone you've lost to recently.

Ratings - please read before commenting - Archived

Re: All time rankings way way off

Re: rankings

Re: Way way way wrong