Ratings - please read before commenting - Archived

JCS · Post by **JCS** » 18 Apr 2006, 08:55

Deram wrote:Oh, but I don't doubt that it is a difficult task. I just think that noone should defend the current ratings as really saying much at all - because they don't. They cannot even compete with the less-than-perfect subjective rankings at all. So, instead of even trying to defend the current program, I think it would be more true and valuable to recognice that currently it is very flawed and can hardly be used for anything at all, BUT that it hopefully is a step towrds better and better computer rankings.

I hope the system get updated more and more and in the end will resemble and hopefully even surpass the subjective rankings we have now - but I think that goal is a good way out in the future.

That is my hope too. That's why I have participated my help and have actually got a few changes implemented, but Computerrank is too busy for me :)

Lennox · Post by **Lennox** » 18 Apr 2006, 14:15

IBO rankings are the proper computer rankings. Why not approach them to have theirs instead. Boxrec rankings ruin what is a great site.

computerrank · Post by **computerrank** » 18 Apr 2006, 14:38

Lennox wrote:IBO rankings are the proper computer rankings. Why not approach them to have theirs instead. Boxrec rankings ruin what is a great site.

They simply are objectivly worse, measured against their conformance with results.

I tested their formula in the past (guessed, as they don't publish the rules, but I got pretty the same ratings) and they don't show competitive prediction quality.

jujigatame · Post by **jujigatame** » 18 Apr 2006, 14:51

Now, I haven't looked closely at the IBO's ranking system or anything, but do you think it's possible that predictive rate isn't the best evaluation system on its own? I mean, if you managed to get a 99.9% predictive rate you'd effectively have done away with upsets, which seems wacky. You should be striving to measure who has accomplished the most, not who is most favored to accomplish the most in the future.

JCS · Post by **JCS** » 18 Apr 2006, 14:56

jujigatame wrote:Now, I haven't looked closely at the IBO's ranking system or anything, but do you think it's possible that predictive rate isn't the best evaluation system on its own? I mean, if you managed to get a 99.9% predictive rate you'd effectively have done away with upsets, which seems wacky. You should be striving to measure who has accomplished the most, not who is most favored to accomplish the most in the future.

How do you measure who has accomplished the most? At least we can look back on wins and losses to derive a prediction-based measurement to utilize.

You all are forgetting one thing, this rating system must "look official" for it to hit the main page. Thats one of the reasons why mine will die on the test server.

Lennox · Post by **Lennox** » 18 Apr 2006, 15:26

Quite clearly a set of ratings have to look right, in my mind prediction quality is not the way to do it. i have just looked at Boxrec ratings at Cruiserweight and in all honesty, if anyone could say they are a good set of ratings that depict the best 25 fighters they should seek medication. Whatever system has been developed that has produced that should be scrapped. How can Johnny Nelson be number 15 and Guillermo Jones be number 2...Virgil Hill is number 29. Nelson has won his last 3 fights against Cantatore, Petkovic and May. How is he number 15, when Kamel Amrane is number 7. Boxrec look plain awful. I know you will post some theory but you need a reality check, the soup does not taste nice.

JCS · Post by **JCS** » 18 Apr 2006, 15:36

Lennox wrote:Quite clearly a set of ratings have to look right, in my mind prediction quality is not the way to do it. i have just looked at Boxrec ratings at Cruiserweight and in all honesty, if anyone could say they are a good set of ratings that depict the best 25 fighters they should seek medication. Whatever system has been developed that has produced that should be scrapped. How can Johnny Nelson be number 15 and Guillermo Jones be number 2...Virgil Hill is number 29. Nelson has won his last 3 fights against Cantatore, Petkovic and May. How is he number 15, when Kamel Amrane is number 7. Boxrec look plain awful. I know you will post some theory but you need a reality check, the soup does not taste nice.

I have made a test system on pure predictive quality. Tell me which you prefer, because no system will ever get it right.

Last Updated: April 10

1 O'Neil Bell
2 Guillermo Jones
3 Jean Marc Mormeck
4 David Haye
5 Johnny Nelson
6 Krzysztof Wlodarczyk
7 Vadim Tokarev
8 Grigory Drozd
9 Virgil Hill
10 Steve Cunningham
11 Enzo Maccarinelli
12 Felix Cora Jr.
13 Carl Thompson
14 Marco Huck
15 Ola Afolabi
16 Johny Jensen
17 Matt Godfrey
18 Emmanuel Nwodo
19 Firat Arslan
20 Mark Hobson
21 Wayne Braithwaite
22 Kamel Amrane
23 Luis Andres Pineda
24 Vincenzo Cantatore
25 Jean Marc Monrose

Lennox · Post by **Lennox** » 18 Apr 2006, 15:46

Yeah that looks miles better at first glance. Why Jones 2? he lost to Cunningham. IBO looks best to me
1 Bell
2 Nelson
3 Mormeck
4 Jirov
5 Cunningham
6 Jones
7 Braithwaite
8 Castillo
9 Haye
10 Thompson
11 Hill
12 Davis
13 Brown
14 Wlodarcy
15 Cantatore

http://www.iboboxing.com

JCS · Post by **JCS** » 18 Apr 2006, 15:48

Lennox wrote:Yeah that looks miles better at first glance. Why Jones 2? he lost to Cunningham. IBO looks best to me
1 Bell
2 Nelson
3 Mormeck
4 Jirov
5 Cunningham
6 Jones
7 Braithwaite
8 Castillo
9 Haye
10 Thompson
11 Hill
12 Davis
13 Brown
14 Wlodarcy
15 Cantatore

http://www.iboboxing.com

My formula did it.. not me. Probably because the loss was only a split decision, while Jones went on to blow out Braithwaite in 4. Cunningham hasn't done much since..

Lennox · Post by **Lennox** » 18 Apr 2006, 16:04

Well those ratings are a big improvement on the current boxrec ones, they sort of 'look' right. I see someone 6-0 who has beaten no one is 38 at cruiser while Virgil Hill is 29. If the formula produces that for prediction rate then clearly its not a formula that produces good boxing ratings. I know the IBO is about what has happened not what is going to happen which is obviously the difference.

JCS · Post by **JCS** » 18 Apr 2006, 16:05

Lennox wrote:Well those ratings are a big improvement on the current boxrec ones, they sort of 'look' right. I see someone 6-0 who has beaten no one is 38 at cruiser while Virgil Hill is 29. If the formula produces that for prediction rate then clearly its not a formula that produces good boxing ratings. I know the IBO is about what has happened not what is going to happen which is obviously the difference.

Ironically.. The ratings I sent you were from a system purely based around what is going to happen.. much more so than the current BoxRec ratings.. so maybe theres more to that than you think ;)

JCS · Post by **JCS** » 18 Apr 2006, 19:41

Lennox wrote:Quite clearly a set of ratings have to look right, in my mind prediction quality is not the way to do it. i have just looked at Boxrec ratings at Cruiserweight and in all honesty, if anyone could say they are a good set of ratings that depict the best 25 fighters they should seek medication. Whatever system has been developed that has produced that should be scrapped. How can Johnny Nelson be number 15 and Guillermo Jones be number 2...Virgil Hill is number 29. Nelson has won his last 3 fights against Cantatore, Petkovic and May. How is he number 15, when Kamel Amrane is number 7. Boxrec look plain awful. I know you will post some theory but you need a reality check, the soup does not taste nice.

Martin,

He does have a point here, especially with Virgil Hill. How is he not ranked higher than #29 considering his EXTREMELY close loss to Mormeck in '04, not to mention his title win over Brudov? I'd figure his close loss over Mormeck who was pretty proven by '04 should've at least assured him a spot in the Top 25.. not to mention Virgil's win over Brudov.

computerrank · Post by **computerrank** » 19 Apr 2006, 06:19

Hill lost about 90 + 40 Points for 1.5 years of inactivity. Otherwise he would be near to top10.

Nelson lost 90 +25 points for his inactivity.

Lennox · Post by **Lennox** » 19 Apr 2006, 09:21

So its a shit formula then. Boxrec Ratings are a disgrace.

JCS · Post by **JCS** » 19 Apr 2006, 09:22

That extra inactivity hit is a bit harsh, I've been saying it for months.

JCS · Post by **JCS** » 21 Apr 2006, 10:53

jcs83md wrote:
Lennox wrote:Quite clearly a set of ratings have to look right, in my mind prediction quality is not the way to do it. i have just looked at Boxrec ratings at Cruiserweight and in all honesty, if anyone could say they are a good set of ratings that depict the best 25 fighters they should seek medication. Whatever system has been developed that has produced that should be scrapped. How can Johnny Nelson be number 15 and Guillermo Jones be number 2...Virgil Hill is number 29. Nelson has won his last 3 fights against Cantatore, Petkovic and May. How is he number 15, when Kamel Amrane is number 7. Boxrec look plain awful. I know you will post some theory but you need a reality check, the soup does not taste nice.
I have made a test system on pure predictive quality. Tell me which you prefer, because no system will ever get it right.

Last Updated: April 10

1 O'Neil Bell
2 Guillermo Jones
3 Jean Marc Mormeck
4 David Haye
5 Johnny Nelson
6 Krzysztof Wlodarczyk
7 Vadim Tokarev
8 Grigory Drozd
9 Virgil Hill
10 Steve Cunningham
11 Enzo Maccarinelli
12 Felix Cora Jr.
13 Carl Thompson
14 Marco Huck
15 Ola Afolabi
16 Johny Jensen
17 Matt Godfrey
18 Emmanuel Nwodo
19 Firat Arslan
20 Mark Hobson
21 Wayne Braithwaite
22 Kamel Amrane
23 Luis Andres Pineda
24 Vincenzo Cantatore
25 Jean Marc Monrose

I updated my routine a bit and updated ratings

April 21, 2006

1 O'Neil Bell
2 Guillermo Jones
3 Jean Marc Mormeck
4 David Haye
5 Johnny Nelson
6 Vadim Tokarev
7 Krzysztof Wlodarczyk
8 Rudy Markussen
9 Virgil Hill
10 Yusaf Mack (Wrong Division)
11 Felix Cora Jr.
12 Grigory Drozd
13 Marco Huck
14 Steve Cunningham
15 Rico Hoye
16 Enzo Maccarinelli
17 Vincenzo Cantatore
18 Carl Thompson
19 Ola Afolabi
20 Matt Godfrey
21 Johny Jensen
22 Emmanuel Nwodo
23 Firat Arslan
24 Kamel Amrane
25 Wayne Braithwaite

Lennox · Post by **Lennox** » 22 Apr 2006, 05:12

jcs83md wrote:
jcs83md wrote:
Lennox wrote:Quite clearly a set of ratings have to look right, in my mind prediction quality is not the way to do it. i have just looked at Boxrec ratings at Cruiserweight and in all honesty, if anyone could say they are a good set of ratings that depict the best 25 fighters they should seek medication. Whatever system has been developed that has produced that should be scrapped. How can Johnny Nelson be number 15 and Guillermo Jones be number 2...Virgil Hill is number 29. Nelson has won his last 3 fights against Cantatore, Petkovic and May. How is he number 15, when Kamel Amrane is number 7. Boxrec look plain awful. I know you will post some theory but you need a reality check, the soup does not taste nice.
I have made a test system on pure predictive quality. Tell me which you prefer, because no system will ever get it right.

Last Updated: April 10

1 O'Neil Bell
2 Guillermo Jones
3 Jean Marc Mormeck
4 David Haye
5 Johnny Nelson
6 Krzysztof Wlodarczyk
7 Vadim Tokarev
8 Grigory Drozd
9 Virgil Hill
10 Steve Cunningham
11 Enzo Maccarinelli
12 Felix Cora Jr.
13 Carl Thompson
14 Marco Huck
15 Ola Afolabi
16 Johny Jensen
17 Matt Godfrey
18 Emmanuel Nwodo
19 Firat Arslan
20 Mark Hobson
21 Wayne Braithwaite
22 Kamel Amrane
23 Luis Andres Pineda
24 Vincenzo Cantatore
25 Jean Marc Monrose
I updated my routine a bit and updated ratings

April 21, 2006

1 O'Neil Bell
2 Guillermo Jones
3 Jean Marc Mormeck
4 David Haye
5 Johnny Nelson
6 Vadim Tokarev
7 Krzysztof Wlodarczyk
8 Rudy Markussen
9 Virgil Hill
10 Yusaf Mack (Wrong Division)
11 Felix Cora Jr.
12 Grigory Drozd
13 Marco Huck
14 Steve Cunningham
15 Rico Hoye
16 Enzo Maccarinelli
17 Vincenzo Cantatore
18 Carl Thompson
19 Ola Afolabi
20 Matt Godfrey
21 Johny Jensen
22 Emmanuel Nwodo
23 Firat Arslan
24 Kamel Amrane
25 Wayne Braithwaite

Not so good that one, too many fighters in the wrong division and a few of your unbeaten fighters are too high. Cunningham must be higher than Jones to make sense to me. Braithwaite is too low. Eliseo Castillo is a glaring non appearer too, jirov too but maybe he fought after you did your list.

JCS · Post by **JCS** » 24 Apr 2006, 21:57

I think the hardest thing to calculate in computerized ratings.... Is, why does a higher rated fighter lose to a lower rated fighter?

Is it because the higher rated fighter declined.. or simply because the lower rated fighter was much better than anyone thought??

Perfect example being Liakhovich beating Brewster. Right now, Liakhovich is #10 and Brewster was #11.. but before the fight, Liakhovich was something like #45 while Brewster was #5 or so. Should Lamon have stayed at #5 or #6 thus propelling Liakhovich one spot above him?? That perhaps would've been a more accurate result than having them at 10,11.

The best the rating system can do is find a midpoint which is dependent on the decision of the fight.

Lennox · Post by **Lennox** » 25 Apr 2006, 03:42

I think its more of a case that fighters start off at basically their worst, then get better by each fight finally reaching a peak, they then get worse (often without knowing it themselves). I dont see anything wrong with a number 33 fighter beating a number 22 fighter. That is the aim, to seach for a higher ranked fighter with whom you think you can edge, and therefore elevate your status in the ratings. That is why I like about IBO ratings the fact that you have to beat someone to be high in the ratings. With Boxrec its more collect wins and up you go and the computer predicts that more often a 24-0 man will will more often than a 15-1 man. Its nonsense we all know there are a lot of 7-8-1 fighters that can beat some of these 13-0 fighters and we all know that keeping your record clean is often down to the promotional agreements rather than learning how to fight

computerrank · Post by **computerrank** » 25 Apr 2006, 06:24

Lennox wrote:I think its more of a case that fighters start off at basically their worst, then get better by each fight finally reaching a peak, they then get worse (often without knowing it themselves). I dont see anything wrong with a number 33 fighter beating a number 22 fighter. That is the aim, to seach for a higher ranked fighter with whom you think you can edge, and therefore elevate your status in the ratings. That is why I like about IBO ratings the fact that you have to beat someone to be high in the ratings. With Boxrec its more collect wins and up you go and the computer predicts that more often a 24-0 man will will more often than a 15-1 man. Its nonsense we all know there are a lot of 7-8-1 fighters that can beat some of these 13-0 fighters and we all know that keeping your record clean is often down to the promotional agreements rather than learning how to fight

Lennox,

BoxRec Ratings exactly doing this - the ratings are based on the ratings of the best opponents defeated (or with close results to them) in the last 4 years.

There is no way in BoxRec Ratings to get a good rating by simply defeating bums.

Best regards
Martin

Lennox · Post by **Lennox** » 25 Apr 2006, 07:50

computerrank wrote:
Lennox wrote:I think its more of a case that fighters start off at basically their worst, then get better by each fight finally reaching a peak, they then get worse (often without knowing it themselves). I dont see anything wrong with a number 33 fighter beating a number 22 fighter. That is the aim, to seach for a higher ranked fighter with whom you think you can edge, and therefore elevate your status in the ratings. That is why I like about IBO ratings the fact that you have to beat someone to be high in the ratings. With Boxrec its more collect wins and up you go and the computer predicts that more often a 24-0 man will will more often than a 15-1 man. Its nonsense we all know there are a lot of 7-8-1 fighters that can beat some of these 13-0 fighters and we all know that keeping your record clean is often down to the promotional agreements rather than learning how to fight
Lennox,

BoxRec Ratings exactly doing this - the ratings are based on the ratings of the best opponents defeated (or with close results to them) in the last 4 years.

There is no way in BoxRec Ratings to get a good rating by simply defeating bums.

Best regards
Martin

The problem with Boxrec though Martin, is that the ratings are aleady distorted, so will get further out of line in time. Ray Mercer ranked 300 when you have some fighters with 3 or 4 wins against weak opponents ranked higher is not right. Let me ask you one question. "Do you think Boxrec ratings are better than IBO, as a set of ratings that accurately puts the boxers in a line" A straight Question.

JCS · Post by **JCS** » 25 Apr 2006, 09:08

Lennox wrote:
computerrank wrote:
Lennox wrote:I think its more of a case that fighters start off at basically their worst, then get better by each fight finally reaching a peak, they then get worse (often without knowing it themselves). I dont see anything wrong with a number 33 fighter beating a number 22 fighter. That is the aim, to seach for a higher ranked fighter with whom you think you can edge, and therefore elevate your status in the ratings. That is why I like about IBO ratings the fact that you have to beat someone to be high in the ratings. With Boxrec its more collect wins and up you go and the computer predicts that more often a 24-0 man will will more often than a 15-1 man. Its nonsense we all know there are a lot of 7-8-1 fighters that can beat some of these 13-0 fighters and we all know that keeping your record clean is often down to the promotional agreements rather than learning how to fight
Lennox,

BoxRec Ratings exactly doing this - the ratings are based on the ratings of the best opponents defeated (or with close results to them) in the last 4 years.

There is no way in BoxRec Ratings to get a good rating by simply defeating bums.

Best regards
Martin
The problem with Boxrec though Martin, is that the ratings are aleady distorted, so will get further out of line in time. Ray Mercer ranked 300 when you have some fighters with 3 or 4 wins against weak opponents ranked higher is not right. Let me ask you one question. "Do you think Boxrec ratings are better than IBO, as a set of ratings that accurately puts the boxers in a line" A straight Question.

You're contradicting yourself. You claim fighters need big wins to rise, but then you feel Mercer should be ranked higher when Mercer hasn't done shit in.. how long? Mercer SHOULD Be ranked higher than #300ish... not much higher, but I believe its because of the 4 year dropoff Martin implemented. I still think a fighter should get SOME credit for his rating previous to the 4-year mark...

On my latest system he is #97.

Lennox · Post by **Lennox** » 25 Apr 2006, 10:38

The problem with Boxrec though Martin, is that the ratings are aleady distorted, so will get further out of line in time. Ray Mercer ranked 300 when you have some fighters with 3 or 4 wins against weak opponents ranked higher is not right. Let me ask you one question. "Do you think Boxrec ratings are better than IBO, as a set of ratings that accurately puts the boxers in a line" A straight Question.[/quote]

You're contradicting yourself. You claim fighters need big wins to rise, but then you feel Mercer should be ranked higher when Mercer hasn't done shit in.. how long? Mercer SHOULD Be ranked higher than #300ish... not much higher, but I believe its because of the 4 year dropoff Martin implemented. I still think a fighter should get SOME credit for his rating previous to the 4-year mark...

On my latest system he is #97.[/quote]

No contradiction at all. I say fair ratings should be the man who beat the man who beat the man etc. Mercer at your 97 is much higher than boxrec 300. Mercer is 60 in the IBO that does not seem to far out to me when I look at fighters like Robert Hawkins, Lou Savarese, Ribert Davis, I think Ray Mercer beats these. You are right a fighter should get some credit for achievements of yesteryear. Mercer is clearly not the fighter he was but he could still win fights at a lower level, the problem is Mercer is now unlikely to fight an opponent where he would be favoured to win, he would be used as a stepping stone by perhaps an overseas promoter. Let me ask you that same question I asked Computerrank. "better ratings IBO or Boxrec"?

computerrank · Post by **computerrank** » 25 Apr 2006, 10:46

computerrank wrote:
Lennox wrote:IBO rankings are the proper computer rankings. Why not approach them to have theirs instead. Boxrec rankings ruin what is a great site.
They simply are objectivly worse, measured against their conformance with results.

I tested their formula in the past (guessed, as they don't publish the rules, but I got pretty the same ratings) and they don't show competitive prediction quality.

Lennox,

I already answered ...

Best regards
Martin

Post by **Marina** » 25 Apr 2006, 10:59

The BoxRec ratings are approximately twice as good as the IBO because they are approximately twice as accurate.
To my mind the IBO seem to have too much opinion weighting in them and I am very sceptical that they are purely objective. Which begs the question as to why should anyone place any faith in a so called objective rating that is not transparent and verifyable.