Ratings - please read before commenting - Archived

computerrank
Editor
Editor
Posts: 2492
Joined: 04 Jan 2003, 18:59

Ratings Release 41 (Planned) - Last Test Run

Post by computerrank »

Jens S wrote: 30 Apr 2019, 08:52 Well, it is definitely better, though I think, that it still has some weird ratings. Do fighters get the same bonus fighting a top10% fighter? Or does it vary on the opponents ranking?
Current changes of release 41 planned:

# - additional points for defeating an opponent in the top decile p4p ranks
# -- who already defeated such an opponent himself
# -- 1 point at least or
# -- opponent's current pre-bout rating at least, if higher
# -- 5 points at most
# -- weighed by clearness of decision and rounds boxed
# -- weighed by number of opponent's losses since last defeated opponent in the top decile p4p ranks
# --- weight=1 for no loss, weight=0.5 for 1 loss, weight=0.25 for 2 losses etc.

Now you can inspect these planned ratings here:

http://151.boxrec.com/~martin/inc/ratings.php
Jens S
Heavyweight
Heavyweight
Posts: 61
Joined: 13 Oct 2003, 14:45

Re: Ratings - please read before commenting

Post by Jens S »

I just checked the ratings of the fighters from Denmark, and the seem quite good now. Did you change some more? A real improvement. I hope to get time to make a check in the coming days, but it seems, that you are on track :-)
Jens S
Heavyweight
Heavyweight
Posts: 61
Joined: 13 Oct 2003, 14:45

Re: Ratings - please read before commenting

Post by Jens S »

Just to be precise. It is this link I used: http://151.boxrec.com/~martin/ratings_test2.php

Test2 seems really good.
daparg
Super Middleweight
Posts: 9
Joined: 28 Apr 2013, 11:21

Re: Ratings - please read before commenting

Post by daparg »

from 0 to 26 for beating an opponent of only 2 bouts? Someone can explain me?


http://boxrec.com/en/event/775509/2281735
computerrank
Editor
Editor
Posts: 2492
Joined: 04 Jan 2003, 18:59

Re: Ratings - please read before commenting

Post by computerrank »

Jens S wrote: 01 May 2019, 19:05 Just to be precise. It is this link I used: http://151.boxrec.com/~martin/ratings.test2.php

Test2 seems really good.
Jens,

http://151.boxrec.com/~martin/ratings.php

contains the latest test version.

test2 is used for internal tests. So I am not sure, what you looked into.
computerrank
Editor
Editor
Posts: 2492
Joined: 04 Jan 2003, 18:59

Re: Ratings - please read before commenting

Post by computerrank »

daparg wrote: 01 May 2019, 20:32 from 0 to 26 for beating an opponent of only 2 bouts? Someone can explain me?


http://boxrec.com/en/event/775509/2281735
Corbalan will be down to #655 from #188 with the planned new release
Manrae
Heavyweight
Heavyweight
Posts: 352
Joined: 28 Nov 2002, 18:57

Re: Ratings - please read before commenting

Post by Manrae »

computerrank wrote: 02 May 2019, 03:18 Jens,

http://151.boxrec.com/~martin/ratings.php

contains the latest test version.

test2 is used for internal tests. So I am not sure, what you looked into.
This is new, I haven't seen this one yet

I'll take a dive into it once I get home :salut:
AA484
Super Lightweight
Posts: 14
Joined: 13 Jun 2016, 10:21

Re: Ratings - please read before commenting

Post by AA484 »

There appears to be some discrepancy in the ratings between some boxers pages and their corresponding "score" in the ratings section. I assume the ratings page figure is the more accurate one?
computerrank
Editor
Editor
Posts: 2492
Joined: 04 Jan 2003, 18:59

Re: Ratings - please read before commenting

Post by computerrank »

AA484 wrote: 02 May 2019, 12:01 There appears to be some discrepancy in the ratings between some boxers pages and their corresponding "score" in the ratings section. I assume the ratings page figure is the more accurate one?
Would you please give an example.
AA484
Super Lightweight
Posts: 14
Joined: 13 Jun 2016, 10:21

Re: Ratings - please read before commenting

Post by AA484 »

computerrank wrote: 02 May 2019, 13:09 Would you please give an example.
Sorry, it was actually the bout/card page which has different numbers.

Breazeale is one example. On the ratings page he has a score of 264.2 (http://boxrec.com/en/ratings?offset=0&s ... eavyweight) but if you click on his bout page with Wilder (http://boxrec.com/en/event/785138/2336793) it is 209.4.
computerrank
Editor
Editor
Posts: 2492
Joined: 04 Jan 2003, 18:59

Re: Ratings - please read before commenting

Post by computerrank »

AA484 wrote: 02 May 2019, 15:33 Sorry, it was actually the bout/card page which has different numbers.

Breazeale is one example. On the ratings page he has a score of 264.2 (http://boxrec.com/en/ratings?offset=0&s ... eavyweight) but if you click on his bout page with Wilder (http://boxrec.com/en/event/785138/2336793) it is 209.4.
Thanks :TU:
The ratings page shows the current rating.
The bout page for scheduled bouts shows the projected rating at time of the scheduled bout. Breazeale's rating will drop on May 4 for missing opponent quality within 18 months.
Duran1970
Lightweight
Posts: 934
Joined: 03 Jan 2018, 14:20

Re: Boxrec Ratings - Read first before commenting on the ratings

Post by Duran1970 »

Jens S wrote: 01 Apr 2019, 19:19 As I read your answer, you wrote, that you were going to change the ratings so 50 % would have a point. Now you won't do it anyway? But I honestly really don't think, I got an answer. Why the sudden and extreme change in the level of points? Has Boxrec been wrong for a decade or so? It seems more like a personal opinion to me, than anything else. And it gives lack of transparency and therefore lack of credibility.

I would like to quote you:
"You cannot give points for meaningless bouts".
As you say, Boxrec gives fractional points. So you do give points

There are some meaningless fights, and I know, that you are aware of that. Bogus fights in Eastern Europe, that hardly can be called sparring. It''s more like shadowboxing. I know that Boxrec has demanded videos of fights from suspicious areas as verification. Thanks! A great step which I appreciate. But fights building up fighters or even fights at a lower level are not meaningless. It is a wrong way to look at boxing.

I think the owners of Boxrec should take a step back and look at what Boxrec is about, what the rankings should show, and who they are for. Is the idea of Boxrec not to give transparency in boxing? Instead of bogus rankings from governing bodies we get computerrankings based on results (which of course is not an exact science, but gives a very good idea of the strength of the boxers). Isn't the basis of Boxrec the users? Isn't it fulfilling their need of information, that makes the site great? Why keep information back from the users? If you really insist on only giving the fighters fractional points, then why not make them public?

I also think, that you should announce on the front page of Boxrec, every time you chance the ratings. In Denmark, where I live, there arose a feud between two promotors/managers, because you changed the ratings. One of them claimed that his fighter was the higher rated (which was correct at the time). A few days later, the other came out publicly calling the other one a liar, because his fighter was ranked higher (which was correct that day). Only because of a screendump was it possible to verify the first claim. That is not good for anybody. Not the promotors/managers, nor boxing or Boxrec.

But let me show you some things from the rankings:
http://boxrec.com/en/boxer/664616
Joe Jones (8-0, 7 KO) beats a guy with one point. He knocks out Rex Harris (4-5-2, 2 KO) in the third round. After that fight we have two fighters with no points. Joe Jones (Now 9-0, 8 KO) then gets knocked out by Shawndell Terell Winters (9-1, 8 KO) in the fifth round. Still no points.

Now lets look at Jonas Madsen versus Andreas Lynggaard:
http://boxrec.com/en/boxer/736932
Two fighters with both 0 points fights to a SD. Jonas Madsen moves to 2 points. So a close win over a opponent with fewer points give more.

That really seems weird. But maybe there is an explanation.

But this seems inexplicable:
Sarah Mahfoud (6-0, 3 KO) versus Stephanie Ducastel (8-4-2, 2 KO):
http://boxrec.com/en/boxer/786389
Ducastel has fought a draw with Jelena Mrdjenovic - the best in the division - just 16 months before. And Ducastel came in with two points. Sarah Mahfoud came in with 0 points, but won every single round against a fighter with 2 points on all three scorecards - 80-72. She only moves from 0 to 1 point.

In the same division:
http://boxrec.com/en/boxer/718233
Licia Boudersa (3-1-1, 0 KO) versus Elodie Bouchlaka (1-0-2, 0 KO)
None of Licia Boudersas opponents has ever had a point before or after.
None of Elodie Bouchlakas opponents has ever had a point before or after - except Boudersa.
Boudersa wins 79-73, 78-75 and 78-74. A fairly clear win, but not a shutout as Mahfoud did in her fight against Ducastel.
They were both at 0 points. Still Boudersa moved from 0 to 2 points.

How can this happen??? Mahfoud clearly fought better opponents before her fight than Boudersa, she had better results - all wins, she beat a clearly better fighter, and she did it more convincingly. Yet she didn't even get the same amount of points. No, she got less.

There is something wrong in the system!

About the stars - there are problems too.

http://boxrec.com/en/ratings?offset=7750
The fighter ranked 7790 has half a star. The fighter ranked 7977 has one star. So the higher ranked fighter has half a star less. He is ranked 187 spots above.

Here it is the same in a weight division
Look a number 695, Larry Smith. He has a full star. Jarvis Shephard at number 672 has half a star.
http://boxrec.com/en/ratings?offset=650 ... ddleweight
The same story here:
http://boxrec.com/en/ratings?offset=450 ... therweight

How can this happen?

Now if such things had happened in the rankings of WB-something - how would you look upon it? Transparency and comprehensibility is needed in boxing. That also goes for Boxrec.

Best regards

Jens S

PS. Duran1970: Hall of Famer, all time great, four time world champion George Dixon is ranked 194 at the all time featherweight:
http://boxrec.com/en/ratings?r_go=&yrR% ... offset=150
194 th eh?.....what a shame..and that's why this site's rankings cannot be taken seriously
computerrank
Editor
Editor
Posts: 2492
Joined: 04 Jan 2003, 18:59

Re: Ratings - please read before commenting

Post by computerrank »

Regarding George Dixon - I answered here in this thread.

viewtopic.php?f=22&t=28051&p=5129322&hi ... n#p5129322
Jens S
Heavyweight
Heavyweight
Posts: 61
Joined: 13 Oct 2003, 14:45

Re: Ratings - please read before commenting

Post by Jens S »

#Computerrank

I took them over to an Excel-sheet, so I actually have the rankings for fighters from Denmark. The new rankings here are really good! The best I have seen for a long time. They rank the boxers fair. No big surprises - just as their results should rank them. And the points reflects their accomplishments compared to each other well. I copied them and showed them to a friend, and he agreed. So those rankings I would be very excited about.

Name........................................... Old......... New
Lolenga Mock..................... 95,740 41,370
Micki Nielsen........................ 65,430 33,260
Enock Poulsen.................... 87,620 28,900
Jeppe Morell........................ 26,700 20,800
Landry Kore.......................... 42,550 18,820
Ditlev Rossing...................... 41,590 15,490
Dennis Ceylan.................. 40,330 14,710
Oliver Meng............................. 1,704.. 12,460
Abdul Khattab.................... 12,590 12,120
Mikkel Nielsen........................ 3,166... 8,325
Kem Ljungquist................... 0,394.. 4,629
Haris Dzindo............................ 0,055.. 3,396
Heva Sharif.............................. 0,057... 2,111
Andreas Lynggaard....... 1,402... 2,084
Frederik Hede Jensen.. 0,318... 2,084
Kim Poulsen............................. 2,950.. 1,871
Allan Mahfoud.................... 14,730.. 1,503
Simon Madsen.................. 25,970 0,798
Adam Bashanov............... 0,104... 0,777
Ali Hussein................................ 8,080.. 0,445
Oliver Mollenberg............ 0,052.. 0,361
Mahdi Jallaw......................... 0,083.. 0,091
Jacob Porsgaard............. 0,050.. 0,000
Danni Alwan........................... 0,049.. 0,000
computerrank
Editor
Editor
Posts: 2492
Joined: 04 Jan 2003, 18:59

Re: Ratings - please read before commenting

Post by computerrank »

Jens,

this is very close the latest test version
http://151.boxrec.com/~martin/ratings.php

Martin
Jens S
Heavyweight
Heavyweight
Posts: 61
Joined: 13 Oct 2003, 14:45

Re: Ratings - please read before commenting

Post by Jens S »

Well, there is a significant difference.

The level of points is very low again, but I guess, that you could multiply by 7-8 times, then it would be the same as now. It is one of the quite frustrating things, by the many changes in the system. As a user, I get used to a certain level of points. After some times, I (and others) can "read" the ratings. If I look at the record of a fighter, I naturally understand how good the opponents are, because I have used the database so much, that I have a good idea about, how good a fighter with 1, 5, 10 or 25 points is. The more I use the database, the easier it gets. But when the level changes, I get disoriented and must start all over. Then I need to learn the new level of points. It is like having to learn how to read once again.

One question: compared to the present ratings, you have changed the level of the lower ranked fighters. That is if you compare the fighter ranked #1 with #100. In heavyweight right now, it #1 has 46 times more points than #100. In the new rankings, it is 75 times. At cruiser it is 53 times vs 108 times. The picture is the same through all the divisions. I am not saying, that it is wrong/better/worse, I am just wondering, why this change?
computerrank
Editor
Editor
Posts: 2492
Joined: 04 Jan 2003, 18:59

Re: Ratings - please read before commenting

Post by computerrank »

Jens S wrote: 03 May 2019, 17:36 Well, there is a significant difference.

The level of points is very low again, but I guess, that you could multiply by 7-8 times, then it would be the same as now. It is one of the quite frustrating things, by the many changes in the system. As a user, I get used to a certain level of points. After some times, I (and others) can "read" the ratings. If I look at the record of a fighter, I naturally understand how good the opponents are, because I have used the database so much, that I have a good idea about, how good a fighter with 1, 5, 10 or 25 points is. The more I use the database, the easier it gets. But when the level changes, I get disoriented and must start all over. Then I need to learn the new level of points. It is like having to learn how to read once again.

One question: compared to the present ratings, you have changed the level of the lower ranked fighters. That is if you compare the fighter ranked #1 with #100. In heavyweight right now, it #1 has 46 times more points than #100. In the new rankings, it is 75 times. At cruiser it is 53 times vs 108 times. The picture is the same through all the divisions. I am not saying, that it is wrong/better/worse, I am just wondering, why this change?
In the current release a boxer can get 25 additional points, independent of the opponents rating, which may be marginal. This is not plausible for many users, including you.

- So the new ratings limit the maximum amount of additional points as described above in relation to the opponent's rating.
-- 1 point at least
-- points as opponent's rating, if higher
-- 5 points at most

The combination of 5 points limit and setting into relation to opponent's rating result in a points level of about 1/10 compared with the current version. The switch from the connected opponent concept to the maximum opponent rank concept for the additional points may also have some influence on the points ratio change between #1 and #100.

A simple factor of 10 would just run into the same inacceptable, unbalanced point wins in some cases.

So the injection of additional points is limited to around 1 point plus a bit, to be acceptable.

The important point is, that the rankings and relations are valid now.
Jens S
Heavyweight
Heavyweight
Posts: 61
Joined: 13 Oct 2003, 14:45

Re: Ratings - please read before commenting

Post by Jens S »

Thank you for the answer, Computerrank.

As I understand your answer (and I may be wrong), the problem by multiplying is, that these multiplicand ratings will disturb future ratings?

But couldn't you leave all the ratings the same in the calculations, and then have what you could call "Display ratings"? What I mean is, that the "real" ratings would not be shown to the users.

If we take a simple multiplication by 10 as an example:
Fighter A has 10 points and fighter B has 2 points in the new ratings. Then in the ratings the users can see, they would have 100 and 20 points. But in the calculations if they fight, you would still use 10 and 2.

I used to be a Lotus notes developer programming CMS-systems (years ago). We often used that kind of "manipulation" of data. The data in the databases would be different, from what people actually saw. It was just one single "layer" put on top of data. We created an extra "View", I believe it was called.
computerrank
Editor
Editor
Posts: 2492
Joined: 04 Jan 2003, 18:59

Re: Ratings - please read before commenting

Post by computerrank »

Jens,

I multiplied by factor 8, which keeps the p4p #100 level equal.

Look and see

http://151.boxrec.com/~martin/ratings.php
BroughtonRulesRefuge
Heavyweight
Heavyweight
Posts: 2762
Joined: 16 Dec 2008, 06:55

Re: Ratings - please read before commenting

Post by BroughtonRulesRefuge »

- Canelo and Crawford top two P4P, yet recent results yield gross discrepancies.

Crawford knocked out his #10 rated opponent and gained points, yet Canelo lost points against his #3 rated opponent for a UD.

Going into their fights, Khan was the sacrificial lamb and Jacobs the live dog.

I was disappointed in the answer to my last inquiry after Mikey dropped so low he was paired with what could be a journeyman as yet so unproven he hadn't yet endured the beatings due him.

Like I say, I have no input in your perpetual project and applaud boxrec adding back a more informative scale perspective, so I'm not so fussed as I an pointing out nonesensical results.
computerrank
Editor
Editor
Posts: 2492
Joined: 04 Jan 2003, 18:59

Re: Ratings - please read before commenting

Post by computerrank »

BroughtonRulesRefuge wrote: 05 May 2019, 15:23 - Canelo and Crawford top two P4P, yet recent results yield gross discrepancies.

Crawford knocked out his #10 rated opponent and gained points, yet Canelo lost points against his #3 rated opponent for a UD.

Going into their fights, Khan was the sacrificial lamb and Jacobs the live dog.

I was disappointed in the answer to my last inquiry after Mikey dropped so low he was paired with what could be a journeyman as yet so unproven he hadn't yet endured the beatings due him.

Like I say, I have no input in your perpetual project and applaud boxrec adding back a more informative scale perspective, so I'm not so fussed as I an pointing out nonesensical results.
Alvarez as the higher rated had a close win with a pre-bout rating ratio of 1:3 and lost some points.
Crawford as a higher rated had a TKO with a pre-bout rating ratio of 1:3.5 and got some points.
Garcia as a lower rated lost plain with a pre-bout rating of 1:1.6 and lost a lot of points points
Manrae
Heavyweight
Heavyweight
Posts: 352
Joined: 28 Nov 2002, 18:57

Re: Ratings - please read before commenting

Post by Manrae »

computerrank wrote: 04 May 2019, 05:13 Jens,

I multiplied by factor 8, which keeps the p4p #100 level equal.

Look and see

http://151.boxrec.com/~martin/ratings.php
The earlier test link looked better IMHO.

Either way, when can we expect a release?
computerrank
Editor
Editor
Posts: 2492
Joined: 04 Jan 2003, 18:59

Re: Ratings - please read before commenting

Post by computerrank »

Manrae wrote: 07 May 2019, 08:33 The earlier test link looked better IMHO.

Either way, when can we expect a release?
I suggested to go live with it, but the core team is hesltating.
Jens S
Heavyweight
Heavyweight
Posts: 61
Joined: 13 Oct 2003, 14:45

Re: Ratings - please read before commenting

Post by Jens S »

I think the ratings are really getting somewhere now. It is always difficult to evaluate precisely, how the ratings should be. We are subjective as human beings, but these ratings seems really credible. The "Exploding" ratings are gone. It is hard to find fighters, that are rated way to high or low.

The level of the points is fine as well. I think this is "as good as it gets". Of course there will be adjustments in the future, but this is a job well done!
Manrae
Heavyweight
Heavyweight
Posts: 352
Joined: 28 Nov 2002, 18:57

Re: Ratings - please read before commenting

Post by Manrae »

computerrank wrote: 07 May 2019, 10:19 I suggested to go live with it, but the core team is hesltating.
Interesting...

Do you think we will ever see pre-fight and post-fight RANKINGS of fighters on the "bout" page? For historical purposes, this would be Godly. I've suggested this before but, nothing.

Honestly, "John" just gets curt with most people making suggestions, in the suggestion thread :confused:
Locked