Ratings - please read before commenting - Archived

TerribleTim2 · Post by **TerribleTim2** » 01 Mar 2006, 05:50

Today the McCall/Gomez fight was changed to a NC. McCall climbed 14 spots when this lose was removed from his record. Shouldn't Gomez been drop a few spots?

JCS · Post by **JCS** » 01 Mar 2006, 09:21

TerribleTim2 wrote:Today the McCall/Gomez fight was changed to a NC. McCall climbed 14 spots when this lose was removed from his record. Shouldn't Gomez been drop a few spots?

Probably not.. Gomez gained very little or nothing for the win.

probert24 · Post by **probert24** » 19 Mar 2006, 05:20

Question on the parameter values. I was curious as to why DQ is only worth 50. Isn't it actually the worse possible lose? I mean its a total and complete quit job, right? For example Golota v Grant. Golota didn't want to be there but due to his previous DQs he decided to just quit and take a TKO. By the established parameters Andrew would have been better off cranking Grant in the nuts four or five times after the 10th round knockdown.

Shouldn't it be worth a bit more?

----------------
* DQ = 50
-----------------

thanks

Post by **John** » 19 Mar 2006, 06:28

The parameters aren't arbitrary, we plot a range of values and choose the one which gives the greatest accuracy.

JCS · Post by **JCS** » 19 Mar 2006, 11:59

JohnShep wrote:The parameters aren't arbitrary, we plot a range of values and choose the one which gives the greatest accuracy.

in my testing I found DQ to have a slightly larger impact but its so hard to get the ball rolling with you and c-rank!

jujigatame · Post by **jujigatame** » 19 Mar 2006, 13:17

probert24 wrote:Question on the parameter values. I was curious as to why DQ is only worth 50. Isn't it actually the worse possible lose? I mean its a total and complete quit job, right? For example Golota v Grant. Golota didn't want to be there but due to his previous DQs he decided to just quit and take a TKO. By the established parameters Andrew would have been better off cranking Grant in the nuts four or five times after the 10th round knockdown.

Shouldn't it be worth a bit more?

----------------
* DQ = 50
-----------------

thanks

Yea but what about a fight like Bowe/Golota where the Golota was dominating but then just got DQed because he's a total headcase?

probert24 · Post by **probert24** » 19 Mar 2006, 15:52

jujigatame wrote:
probert24 wrote:
----------------
* DQ = 50
-----------------
Yea but what about a fight like Bowe/Golota where the Golota was dominating but then just got DQed because he's a total headcase?

I think that further proves that the DQ should be worth more. Golota getting DQ'd is a worse reflection on him than any of the other listed criteria. The DQ is a reflection of your deepest fiber as a person and a fighter.

JCS · Post by **JCS** » 19 Mar 2006, 20:03

probert24 wrote:
jujigatame wrote:
probert24 wrote:
----------------
* DQ = 50
-----------------
Yea but what about a fight like Bowe/Golota where the Golota was dominating but then just got DQed because he's a total headcase?
I think that further proves that the DQ should be worth more. Golota getting DQ'd is a worse reflection on him than any of the other listed criteria. The DQ is a reflection of your deepest fiber as a person and a fighter.

The DQ setting is ABOUT right.. It may be a fraction low.. but thats it. I've tested it.

jbryniar1 · Post by **jbryniar1** » 30 Mar 2006, 17:05

What will Joe Mesi be ranked after he beats Ron Bellemy on April 1?

JCS · Post by **JCS** » 30 Mar 2006, 17:09

jbryniar1 wrote:What will Joe Mesi be ranked after he beats Ron Bellemy on April 1?

Probably bottom of Top 10

jbryniar1 · Post by **jbryniar1** » 30 Mar 2006, 17:30

jcs83md wrote:
jbryniar1 wrote:What will Joe Mesi be ranked after he beats Ron Bellemy on April 1?
Probably bottom of Top 10

Sounds fair.

computerrank · Post by **computerrank** » 31 Mar 2006, 06:07

... more around #30 - he lost 180 points for 2 complete years of inactivity.

JCS · Post by **JCS** » 31 Mar 2006, 10:23

Oops forgot about the 1yr inactivity automatic drop.

pundit · Post by **pundit** » 07 Apr 2006, 13:42

Can someone please explain how Jack Dempsey is not even top 100 in the all-time heavyweight ranking???

Needless to say that most people would have Dempsey in the top 5.

JCS · Post by **JCS** » 07 Apr 2006, 13:43

pundit wrote:Can someone please explain how Jack Dempsey is not even top 100 in the all-time heavyweight ranking???

Needless to say that most people would have Dempsey in the top 5.

Incomplete records have caused guys he's fought (or guys that Dempsey's opponents have fought, etc etc.) to seem lesser than they were. Hence, Dempsey doesn't get as much credit from the computer.

Deram · Post by **Deram** » 17 Apr 2006, 01:20

Well, why even bother with these ratings? Most people don't consider boxrec ratings any serious thing at all.

Chanet refused to meet Kessler eventhough he was Kessler's mandatory. Furthermore - while Kessler has beaten Green, Thysse, former champ siaca, Former champ Mundine (who is now still ranked top 5 by the Ring) and many more... Chanet has only one half-decent name on his resume and that is Tsypcho.

And Reid is higher rated than Calzaghe is just as far out.

Perhaps it shows the complete and unwarranted bias that because Chanet has a few minor wins at LHW - it is much easier to earn points for these rankings in LHW than SMW because SMW is a newer division. Whoever wrote this progrm for boxrec does not have a clue about boxing and it would improve the reputation of boxrec if they simply discontinued those "rankings" for good or at least untill they found someone able to write a far better program.

Boxrec forum is good and boxrec name database is the best on the internet, but these "rankings" are ridiculous and only harm the seriousness of the whole site.

computerrank · Post by **computerrank** » 17 Apr 2006, 08:52

Chanet defeated Tsypko, who defeated Magee, who held the IBO title for a long time ...

... and Reid rated higher than Calzaghe??? you mean Kessler?

Reid got his rating by his win over Magee and close loss to Ottke.

Kessler defeated the boxers you mentioned, not at their best rating.

Deram · Post by **Deram** » 17 Apr 2006, 11:11

computerrank wrote:Chanet defeated Tsypko, who defeated Magee, who held the IBO title for a long time ...

... and Reid rated higher than Calzaghe??? you mean Kessler?

Reid got his rating by his win over Magee and close loss to Ottke.

Kessler defeated the boxers you mentioned, not at their best rating.

and you think Magee, Chanet or Tsypcho are better than ex WBA champ Mundine, ex WBA champ Siaca and current WBA champ Kessler?

There are levels of skill between those two groups and it is totally ridiculous that Chanet could ever be considered even remotely close to Kessler - and ahead of Kessler? Noone - NOONE in the entire boxing world would claim Chanet could beat Kessler... or that he has meet better opposition. Not even Chanets own promotors. They have openly said that they don't want Chanet to meet Kessler.

and Kessler didn't defeat the boxers at their best rating??????

Siaca was REIGNING WBA world champion when Kessler beat him.. Exactly WHEN was he higher ranked than that?

Mundine was ranked no. 4 by the Ring when Kessler fought him, down from no. 3 before the close defeat to Siaca.

Magee has never been a real champion (ibf, wba and wbc - wbo sometimes included). He has never even been a serious contender and has lost to quite ordinary fighters.
Tsypcho is the same level - mediocre at best.
Chanet similarly... he just lost to another mediocre - middleweight!

Siaca has never lost outside of worldchampionships fights and when Kessler beat him he was reigning champ.
Mundine beat Echols - who is better than anyone the three mediocre fighter magee, tsypcho and chamet have ever met - and Mundine is ex WBA world champion.
Kessler has beat both these and many many more.

You cannot be serious thinking that Chanet could in any way shape or form be even considered to rank above Kessler?????

JCS · Post by **JCS** » 17 Apr 2006, 11:14

Deram,

In summary - the ratings system is programmed to achieve the highest prediction rate possible (Did the higher rated fighter win the fight) while still looking acceptable.

I agree, there are some noticeable flaws in the system. I've been trying to kickstart c-rank and JohnShep to let me assist but all I've done is achieve a much higher prediction rate, silly me

Truth is, there is no perfect answer.. although it could be tweaked up a bit.

JCS · Post by **JCS** » 17 Apr 2006, 11:19

ComputerRank

There appears to be a bug in the system.

Check out Sergey Tatevosyan.. he just jumped to #12 in Middleweight after losing to a guy who is 3-0 who is not even in the Top 500.

Freiheit · Post by **Freiheit** » 17 Apr 2006, 11:23

jcs83md wrote:ComputerRank

There appears to be a bug in the system.

Check out Sergey Tatevosyan.. he just jumped to #12 in Middleweight after losing to a guy who is 3-0 who is not even in the Top 500.

Its not a bug.
The result was entered after the latest ratings calculation. I then changed him from super middle to middleweight.

JCS · Post by **JCS** » 17 Apr 2006, 11:26

Freiheit wrote:
jcs83md wrote:ComputerRank

There appears to be a bug in the system.

Check out Sergey Tatevosyan.. he just jumped to #12 in Middleweight after losing to a guy who is 3-0 who is not even in the Top 500.
Its not a bug.
The result was entered after the latest ratings calculation. I then changed him from super middle to middleweight.

Ah, ok.. I figured that was the alternate possibility.

Deram · Post by **Deram** » 17 Apr 2006, 19:54

jcs83md wrote:Deram,

In summary - the ratings system is programmed to achieve the highest prediction rate possible (Did the higher rated fighter win the fight) while still looking acceptable.

I agree, there are some noticeable flaws in the system. I've been trying to kickstart c-rank and JohnShep to let me assist but all I've done is achieve a much higher prediction rate, silly me Truth is, there is no perfect answer.. although it could be tweaked up a bit.

"..the highest prediction rate possible".

Well, the current system will not predict well. Noone in boxing doubted for a second that Kessler would beat Chanet. Infact there would be like 99% chance of Kessler win and about 1% for Chanet.

I do not know the algorythm so I do not know what is missing, all I can tell and most people can easily spot is that the system is far more flawed than all the "hand-made" rankings out there - and that is bad because they are often driven by politics etc. So they are not perfect due to politics etc. while boxrec rankings may be politics free, but they are just heavily flawed in general.

I have one suggestion. How about letting "manner of winning" count? So that a win is not just a win, but so that an avarage UD of 119-109 counts for far more than a 115-114 win? It would make sense because a supposedly good fighter only managing to do a 115-114 should be taken with a pinch of salt, while a fighter winning in shoot out fashion may be far better.

I know you do not have info about this for all fights, but once all fights are registered in the database the average UD win can be registered (perhaps for each division if that is of importance) and then assigned to the fights where we do not have info.

The system has very poor predictive abilities and has far more huge mistakes than other rankings. One further flaw is that when a fighter is "exposed" or in other words "over-the-hill - it often goes fast and I have noticed that bocrec rankings treats this much too slowly. Even if everyone with two eyes can see that a fighter is surely done, boxrec still holds him much too high.

JCS · Post by **JCS** » 17 Apr 2006, 22:18

Deram wrote:
"..the highest prediction rate possible"..

You forgot the last part.. There is apparently additional subjective criteria in that it must look acceptable.. You know.. for example, we can't assume a certain fighter is #4 per say.. but he must have done something to earn it.. or come close

Deram wrote: I have one suggestion. How about letting "manner of winning" count? So that a win is not just a win, but so that an avarage UD of 119-109 counts for far more than a 115-114 win? It would make sense because a supposedly good fighter only managing to do a 115-114 should be taken with a pinch of salt, while a fighter winning in shoot out fashion may be far better.

There is a bit of this in the system already.. and I agree, it could be taken one step further. I have proven this in my system. Further adjustments could be made.. as in, a early round KO counting more than a late round.. etc., etc. I could go on and on

Deram wrote: I know you do not have info about this for all fights, but once all fights are registered in the database the average UD win can be registered (perhaps for each division if that is of importance) and then assigned to the fights where we do not have info.

This is already setup.

Deram wrote: The system has very poor predictive abilities and has far more huge mistakes than other rankings. One further flaw is that when a fighter is "exposed" or in other words "over-the-hill - it often goes fast and I have noticed that bocrec rankings treats this much too slowly. Even if everyone with two eyes can see that a fighter is surely done, boxrec still holds him much too high

I agree, it could be improved upon.. but trust me, it is very difficult to make a computerized system to compete with subjective rankings. Also, in some occasions.. some fighters are hit too hard because they are considered "over the hill" .. and some fighters are not boosted enough for having early career success..

Deram · Post by **Deram** » 18 Apr 2006, 02:32

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.