Ratings - please read before commenting - Archived

[Manrae]

This is in the pipeline. Great wins better balanced with bad losses.

http://151.80.18.199/~martin/inc/rating ... ationality

.

Henry Armstrong not in the top 100...

Chris John over Archie Moore...



The current formula needs work, but this is a step backwards IMO

Both have too many bad losses devaluating their big wins.

http://www.espn.com/sports/boxing/great ... reatest110
#3

https://www.thoughtco.com/ring-magazine ... gs-4153939
#2

IBHOF
2 fighter of the year awards
Simultaneously held 3 world titles in 3 weight divisions

...but not boxrec top 100 material? Seriously?

Chris John was great in his own right and very well may end up in hall of fame.

However, the difference between Moore//Armstrong is that they are in the argument of Greatest of all Time.

[computerrank]

Both have too many bad losses devaluating their big wins.

http://www.espn.com/sports/boxing/great ... reatest110
#3

https://www.thoughtco.com/ring-magazine ... gs-4153939
#2

IBHOF
2 fighter of the year awards
Simultaneously held 3 world titles in 3 weight divisions

...but not boxrec top 100 material? Seriously?

Chris John was great in his own right and very well may end up in hall of fame.

However, the difference between Moore//Armstrong is that they are in the argument of Greatest of all Time.

Boxrec cannot see styles, only the results. Compare their records with Pacquiao. Both had bad losses mixed into their tops.

[computerrank]

Manrae, I will look into the details and discuss it with you when I will be back home on March, 12.

[computerrank]

Another strong improvement:

# - all boxers connected by common bouts starting with seeds
#
# - losers' pre-bout rating increased by 3.5 * add_base for calculation
new ratings
#
# - all time ratings
# -- pairwise compensation of top wins within last year and low losses
# -- within 3 years - 1 year before and 1 year after
# -- compensation my geometric mean, before cut to low loss


winner prediction in percent for top decile bouts:
- based on ratings only 83.54 => 85.86
- based on ratings and additional 16 parameters 84.33 => 86.70 !!!

winner prediction in percent for all non debutant bouts_
- based on ratings only 74.93 => 79.77
- based on ratings and additional 16 parameters 77.87 => 82.18 !!!

[Manrae]

Manrae, I will look into the details and discuss it with you when I will be back home on March, 12.

Fair enough.

[Azmir]

These rankings are an absolute mess. The pound for pound list has been created by Adelaide Byrd, no question. She scored it 120-108 for Canelo.

Jarrell Miller at 6, Tom Schwarz at 12 makes absolutely no sense at HW, as well as Lerena at 6 and Szello at 18.

Just when you think things could get no worse, you see Golovkin and Loma in front of Usyk (IMO absolute #1 p4p) and Joshua (who almost unified the top division of the sport).

Then you see the female rankings and you just couldn't believe what you just read.

[Ruthless-RKO]

These rankings are an absolute mess. The pound for pound list has been created by Adelaide Byrd, no question. She scored it 120-108 for Canelo.

Jarrell Miller at 6, Tom Schwarz at 12 makes absolutely no sense at HW, as well as Lerena at 6 and Szello at 18.

Just when you think things could get no worse, you see Golovkin and Loma in front of Usyk (IMO absolute #1 p4p) and Joshua (who almost unified the top division of the sport).

Then you see the female rankings and you just couldn't believe what you just read.

It’s a point scoring system mate.

If you want rankings that are not based on points, visit ESPN, TheRing, TBRB, CyberZone

[computerrank]

Manrae, I will look into the details and discuss it with you when I will be back home on March, 12.

Fair enough.

Thanks Manrae,
I found a typo, which caused 0 annual points, when the boxer had any loss within a year before and after the reference year. The loss should only reduce the points, but it accidently cut down the annual points to 0.
The next release in March will include the fix.

[Manrae]

Manrae, I will look into the details and discuss it with you when I will be back home on March, 12.

Fair enough.

Thanks Manrae,
I found a typo, which caused 0 annual points, when the boxer had any loss within a year before and after the reference year. The loss should only reduce the points, but it accidently cut down the annual points to 0.
The next release in March will include the fix.

That's awesome!
I'm looking forward to seeing it

[m1kee50]

we can search the BBBofC titles by Area... is there any chance of adding the Area filter into the fighter rankings too? So being able to see all the fighters eligible to fight for the Southern Area title at a given weight?

[computerrank]

we can search the BBBofC titles by Area... is there any chance of adding the Area filter into the fighter rankings too? So being able to see all the fighters eligible to fight for the Southern Area title at a given weight?

Hi Matt,
I guess, you should post this in this thread:

viewtopic.php?f=1&t=94835

[Manrae]

winner prediction in percent for top decile bouts:
- based on ratings only 83.54 => 85.86
- based on ratings and additional 16 parameters 84.33 => 86.70 !!!

winner prediction in percent for all non debutant bouts_
- based on ratings only 74.93 => 79.77
- based on ratings and additional 16 parameters 77.87 => 82.18 !!!

Hey I was wondering if you could explain this in "layman's terms" for me. Could you use examples as well?

[computerrank]

winner prediction in percent for top decile bouts:
- based on ratings only 83.54 => 85.86
- based on ratings and additional 16 parameters 84.33 => 86.70 !!!

winner prediction in percent for all non debutant bouts_
- based on ratings only 74.93 => 79.77
- based on ratings and additional 16 parameters 77.87 => 82.18 !!!

Hey I was wondering if you could explain this in "layman's terms" for me. Could you use examples as well?

winner prediction for all bouts, where the top boxer is in the top 10 percent of all boxers

- based on comparing their rating points: correct 83.54 percent (release 37) improved to correct 85.86 percent (release 38)

- based on comparing their rating points and additional record values: correct 84.33 percent (release 37) improved to correct 86.70 percent (release 38)

additional record values regarded:
-- results of last 6 bouts
-- stopped opponent in last 6 bouts
-- won in last 6 bouts
-- lost in last 6 bouts
-- stopped in last 6 bouts
-- boxed within last 2 years
-- best opponent stopped in last 6 bouts
-- best opponent defeated in last six bouts
-- worst opponent lost to in last 6 bouts
-- worst opponent stopped by in last 6 bouts
-- age

[BroughtonRulesRefuge]

- Mikey p4p ranking in every p4p ranking of active fighters before the Spence fight including boxrec.

Now, I go back to when the obese Bowe announced a comeback that immediately propelled him to #1 heavy rankings those early years. Took a few days, but he was soon back of a substantial pack of active heavies above him.

Here we have an elete p4p 4 division champ Mikey losing a lopsided decision to Spence in a classic "bridge too far" fight, and now Mikey below Mohamed Mimounce at 55th on your p4p rankings?

http://boxrec.com/en/ratings?&offset=50

Since your formula for all rankings is an ongoing project and I have little say so in that, I'm seldom fussed over what can be seen as intuitively bad results, but this current construct highlights a serious flaw in the formulae chosen.

Maybe some day the holy grail shall be achieved. Baseball thought they found it in Bill James, and maybe they did since baseball is moving closer to fantasy baseball which is where all the stats geeks have moved to. Big money to be had in fantasy baseball.

[computerrank]

- Mikey p4p ranking in every p4p ranking of active fighters before the Spence fight including boxrec.

Now, I go back to when the obese Bowe announced a comeback that immediately propelled him to #1 heavy rankings those early years. Took a few days, but he was soon back of a substantial pack of active heavies above him.

Here we have an elete p4p 4 division champ Mikey losing a lopsided decision to Spence in a classic "bridge too far" fight, and now Mikey below Mohamed Mimounce at 55th on your p4p rankings?

http://boxrec.com/en/ratings?&offset=50

Since your formula for all rankings is an ongoing project and I have little say so in that, I'm seldom fussed over what can be seen as intuitively bad results, but this current construct highlights a serious flaw in the formulae chosen.

Maybe some day the holy grail shall be achieved. Baseball thought they found it in Bill James, and maybe they did since baseball is moving closer to fantasy baseball which is where all the stats geeks have moved to. Big money to be had in fantasy baseball.

I cannot see the problem. Mickey Garcia now is #7 at Welterweight with 101 points. Mohamed Mimoune is #6 at Super Lightweight with 101 points. Titles do not count in any way.

[Manrae]

winner prediction in percent for top decile bouts:
- based on ratings only 83.54 => 85.86
- based on ratings and additional 16 parameters 84.33 => 86.70 !!!

winner prediction in percent for all non debutant bouts_
- based on ratings only 74.93 => 79.77
- based on ratings and additional 16 parameters 77.87 => 82.18 !!!

Hey I was wondering if you could explain this in "layman's terms" for me. Could you use examples as well?

winner prediction for all bouts, where the top boxer is in the top 10 percent of all boxers

- based on comparing their rating points: correct 83.54 percent (release 37) improved to correct 85.86 percent (release 38)

- based on comparing their rating points and additional record values: correct 84.33 percent (release 37) improved to correct 86.70 percent (release 38)

additional record values regarded:
-- results of last 6 bouts
-- stopped opponent in last 6 bouts
-- won in last 6 bouts
-- lost in last 6 bouts
-- stopped in last 6 bouts
-- boxed within last 2 years
-- best opponent stopped in last 6 bouts
-- best opponent defeated in last six bouts
-- worst opponent lost to in last 6 bouts
-- worst opponent stopped by in last 6 bouts
-- age

So do these additional record values get factored in to create a new rating for each boxer?
For example) Boxer A has a rating of 100, but with the additional record values it goes to 130?

[computerrank]

winner prediction for all bouts, where the top boxer is in the top 10 percent of all boxers

- based on comparing their rating points: correct 83.54 percent (release 37) improved to correct 85.86 percent (release 38)

- based on comparing their rating points and additional record values: correct 84.33 percent (release 37) improved to correct 86.70 percent (release 38)

additional record values regarded:
-- results of last 6 bouts
-- stopped opponent in last 6 bouts
-- won in last 6 bouts
-- lost in last 6 bouts
-- stopped in last 6 bouts
-- boxed within last 2 years
-- best opponent stopped in last 6 bouts
-- best opponent defeated in last six bouts
-- worst opponent lost to in last 6 bouts
-- worst opponent stopped by in last 6 bouts
-- age

So do these additional record values get factored in to create a new rating for each boxer?
For example) Boxer A has a rating of 100, but with the additional record values it goes to 130?

These values are used to get an improved winner prediction and stoppage prediction for any scheduled bout. It is used on the event page:

http://boxrec.com/en/event/782880

[Manrae]

winner prediction for all bouts, where the top boxer is in the top 10 percent of all boxers

- based on comparing their rating points: correct 83.54 percent (release 37) improved to correct 85.86 percent (release 38)

- based on comparing their rating points and additional record values: correct 84.33 percent (release 37) improved to correct 86.70 percent (release 38)

additional record values regarded:
-- results of last 6 bouts
-- stopped opponent in last 6 bouts
-- won in last 6 bouts
-- lost in last 6 bouts
-- stopped in last 6 bouts
-- boxed within last 2 years
-- best opponent stopped in last 6 bouts
-- best opponent defeated in last six bouts
-- worst opponent lost to in last 6 bouts
-- worst opponent stopped by in last 6 bouts
-- age

So do these additional record values get factored in to create a new rating for each boxer?
For example) Boxer A has a rating of 100, but with the additional record values it goes to 130?

These values are used to get an improved winner prediction and stoppage prediction for any scheduled bout. It is used on the event page:

http://boxrec.com/en/event/782880

I'm sorry , I don't see either of these on this page.

Are they hidden? Is the 1st listed fighter the automatic favorite?

[computerrank]

winner prediction for all bouts, where the top boxer is in the top 10 percent of all boxers

- based on comparing their rating points: correct 83.54 percent (release 37) improved to correct 85.86 percent (release 38)

- based on comparing their rating points and additional record values: correct 84.33 percent (release 37) improved to correct 86.70 percent (release 38)

additional record values regarded:
-- results of last 6 bouts
-- stopped opponent in last 6 bouts
-- won in last 6 bouts
-- lost in last 6 bouts
-- stopped in last 6 bouts
-- boxed within last 2 years
-- best opponent stopped in last 6 bouts
-- best opponent defeated in last six bouts
-- worst opponent lost to in last 6 bouts
-- worst opponent stopped by in last 6 bouts
-- age

So do these additional record values get factored in to create a new rating for each boxer?
For example) Boxer A has a rating of 100, but with the additional record values it goes to 130?

These values are used to get an improved winner prediction and stoppage prediction for any scheduled bout. It is used on the event page:

http://boxrec.com/en/event/782880

I'm sorry , I don't see either of these on this page.

Are they hidden? Is the 1st listed fighter the automatic favorite?

Maybe only the editors, commissions, matchmakers, promoters, managers can see it.

[Manrae]

winner prediction for all bouts, where the top boxer is in the top 10 percent of all boxers

- based on comparing their rating points: correct 83.54 percent (release 37) improved to correct 85.86 percent (release 38)

- based on comparing their rating points and additional record values: correct 84.33 percent (release 37) improved to correct 86.70 percent (release 38)

additional record values regarded:
-- results of last 6 bouts
-- stopped opponent in last 6 bouts
-- won in last 6 bouts
-- lost in last 6 bouts
-- stopped in last 6 bouts
-- boxed within last 2 years
-- best opponent stopped in last 6 bouts
-- best opponent defeated in last six bouts
-- worst opponent lost to in last 6 bouts
-- worst opponent stopped by in last 6 bouts
-- age

So do these additional record values get factored in to create a new rating for each boxer?
For example) Boxer A has a rating of 100, but with the additional record values it goes to 130?

These values are used to get an improved winner prediction and stoppage prediction for any scheduled bout. It is used on the event page:

http://boxrec.com/en/event/782880

I'm sorry , I don't see either of these on this page.

Are they hidden? Is the 1st listed fighter the automatic favorite?

Maybe only the editors, commissions, matchmakers, promoters, managers can see it.

Well that sucks

Looking forward to the new all-time rankings being implemented

Out of curiosity , have you ever played around with using age as a factor into a fighter's current rating?

[computerrank]

Maybe only the editors, commissions, matchmakers, promoters, managers can see it.

Well that sucks

Looking forward to the new all-time rankings being implemented

Out of curiosity , have you ever played around with using age as a factor into a fighter's current rating?

No, only official results are considered in the ratings.

[computerrank]

All is the same, the formula didn't change in any way, but

the only difference is the self-explaining restriction:, winners against opponents without any win or with last loss against such opponents, get no additional points now.

The effect is stunning.

This will stop the promotional and matchmaking games with worthless point transfers.

Fair play for all competitors.

[Paci]

All is the same, the formula didn't change in any way, but

the only difference is the self-explaining restriction:, winners against opponents without any win or with last loss against such opponents, get no additional points now.

The effect is stunning.

This will stop the promotional and matchmaking games with worthless point transfers.

Fair play for all competitors.

Yeah, padding is padding. Nothing else.

[jw9170]

Have your boxrec ratings changed or is there something I can do in my settings to see both the old and the new ratings. I am now seeing alot of fighters with 0 ratings. Everyones ratings seems to have been scaled down somehow so that it is smaller. This is confusing. An example would be:
Zafar Pariper vs Aran Dipaen both have ratings 0. However the former is ranked 168/890 whilst the latter is 411/890

[computerrank]

Have your boxrec ratings changed or is there something I can do in my settings to see both the old and the new ratings. I am now seeing alot of fighters with 0 ratings. Everyones ratings seems to have been scaled down somehow so that it is smaller. This is confusing. An example would be:
Zafar Pariper vs Aran Dipaen both have ratings 0. However the former is ranked 168/890 whilst the latter is 411/890

The ratings changed yesterday.

The formula itself didn't change in any way. The only difference is the self-explaining restriction: winners against opponents without any win or with last loss against such opponents, get no additional points now. Before, we noticed too many boxers collecting points against these opponents and instantly transfer these points to higher rated boxers.

The effect is stunning. Exclude points from obvious no-achievers and some records collapse.

The level decreased substantially. Especially at the non-substantial lower levels. Parpiev has 0.87 points and Dipaen hat 0.0051 points.

Fair play for all competitors.