Ratings - please read before commenting - Archived

[BigEars]

I appreciate what boxrec are trying to do here, there is certainly a minority of fighters/promoters out there who manipulated the system to achieve a better ranking than they deserved. But they were very much a minority.
In the main the system was very accurate, both in terms of rating top level fighters and those lower down the ladder, and I believe most people would believe it was a fair ranking system.

From a statistical perspective I'm sure the new system is sound, and can be explained through connected opponents etc.
It still seems to do a reasonable enough job of ranking the top level fighters and perhaps it still works well as a predictive model.....but once you move outside of the top level fighters it seems it can be way off.

Take for example the show in Ireland last weekend. Alan Phelan is now ranked #136 in the world after beating Aidan Metcalfe. A very generous ranking. Over 200 places behind him at #365 is Eric Donovan, the shows headliner who beat a much better fighter in Stephen McAfee.
Donovan has also beaten much better previous opponents in Dai Davies and Juan Luis Gonzalez. In reality it is highly probable either of those two men, or McAfee would beat Phelan, and unquestionable that Donovan would with ease.
Now I'm sure connected opponents can explain this statistically, and Phelan's narrow loss to the 2-40-2 (at the time) Jamie Quinn probably somehow helps him. But the old system would never have had such glaring an error in ranking ability level between these two fighters.

For promoters/matchmakers who use the system to help find opponents, and fans who like to track fighters progress from the way up (and not just when they get within the top 20% of fighters in the World) this new system is far less user friendly.
Without considerable tinkering I suspect people will start to use the system less and try to find an alternative. Which is sad as boxrec has been a great tool for such a long time.

[computerrank]

I appreciate what boxrec are trying to do here, there is certainly a minority of fighters/promoters out there who manipulated the system to achieve a better ranking than they deserved. But they were very much a minority.
In the main the system was very accurate, both in terms of rating top level fighters and those lower down the ladder, and I believe most people would believe it was a fair ranking system.

From a statistical perspective I'm sure the new system is sound, and can be explained through connected opponents etc.
It still seems to do a reasonable enough job of ranking the top level fighters and perhaps it still works well as a predictive model.....but once you move outside of the top level fighters it seems it can be way off.

Take for example the show in Ireland last weekend. Alan Phelan is now ranked #136 in the world after beating Aidan Metcalfe. A very generous ranking. Over 200 places behind him at #365 is Eric Donovan, the shows headliner who beat a much better fighter in Stephen McAfee.
Donovan has also beaten much better previous opponents in Dai Davies and Juan Luis Gonzalez. In reality it is highly probable either of those two men, or McAfee would beat Phelan, and unquestionable that Donovan would with ease.
Now I'm sure connected opponents can explain this statistically, and Phelan's narrow loss to the 2-40-2 (at the time) Jamie Quinn probably somehow helps him. But the old system would never have had such glaring an error in ranking ability level between these two fighters.

For promoters/matchmakers who use the system to help find opponents, and fans who like to track fighters progress from the way up (and not just when they get within the top 20% of fighters in the World) this new system is far less user friendly.
Without considerable tinkering I suspect people will start to use the system less and try to find an alternative. Which is sad as boxrec has been a great tool for such a long time.

We already found some outliers in the current system, getting too many points. We reacted as already described above and tightened the connectness requirements.

Regarding your example Phelan will be down to #495, while Donavan will keep his rank.

Considering the lower general points level we also will multiply all the points by 5. Thus we enhance the visability of very low ratings and have top ratings around 1000 points as in former releases.

[Manrae]

We already found some outliers in the current system, getting too many points. We reacted as already described above and tightened the connectness requirements.

Regarding your example Phelan will be down to #495, while Donavan will keep his rank.

Considering the lower general points level we also will multiply all the points by 5. Thus we enhance the visability of very low ratings and have top ratings around 1000 points as in former releases.

Sounds good

When should we anticipate the new change?

[computerrank]

The new release is ready, but the team is busy with some other work. So I guess, next week.

[Manrae]

The new release is ready, but the team is busy with some other work. So I guess, next week.

Cool

[marcianofan]

We already found some outliers in the current system, getting too many points. We reacted as already described above and tightened the connectness requirements.

Regarding your example Phelan will be down to #495, while Donavan will keep his rank.

Considering the lower general points level we also will multiply all the points by 5. Thus we enhance the visability of very low ratings and have top ratings around 1000 points as in former releases.

I'm glad you're making this change. I think higher resolution is always good, as is consistency with previous releases. Is there any good reason not to also reveal the decimal places? I know nobody much cares about the difference between a .003 fighter and a .005 fighter, but I guess I would argue that, contrary to the attitude you seem to be operating under, no fighter or fight is totally meaningless. When one guy beats another, it creates a valid assumption that the winner is better than the loser. Its meaning is on a much smaller scale than a #1 vs. #2 matchup obviously,, but I think that's what numbers are good for showing. It's not so much an issue of making sure "x" number of fighters have at least 1 point. It's more about showing the difference between an 0-1 fighter who got knocked out in 1 round against another debuting fighter, and the guy who knocked him out. To show them both at a flat "0" is to imply they and their fights practically don't exist. As-is, the difference between the top 50% and the bottom 1% appears to be negligible and arbitrary since they at least graphically have the same points. I'd love to see the difference between 0.0000000 and 0.999999, and all levels in between.

[computerrank]

Maybe it is an idea to show the rating points down to 4 decimals.

It wouldn't make the ratings more readable, but is is perhaps the best way to show the last details.

[oogiebe]

Maybe it is an idea to show the rating points down to 4 decimals.

It wouldn't make the ratings more readable, but is is perhaps the best way to show the last details.

I don't see why fighters at those levels need to be differentiated from one another. It would occur to me that the points indicate the minute differences in accomplishments. Maybe two decimals would do.

[marcianofan]

I don't see why fighters at those levels need to be differentiated from one another. It would occur to me that the points indicate the minute differences in accomplishments. Maybe two decimals would do.

Maybe show a decimal or two on the rankings page, but then also show the points with more decimals on the fighter pages when you click on them? That way it doesn't ugly-up the rankings list, but you can find the information easily enough if you are looking for it. Or...and I don't know if this would be easy to implement or not...perhaps list the full 4 decimals as a mouse-over feature if you hover on the basic number? You wouldn't have to include decimals on the list of rankings at all that way.

As for why they need to be differentiated...since unconnected fighters are at very low levels, I would say the differences are far from minute in their own contexts. So a guy with .002 has theoretically twice the accomplishment as a guy with .001. It's just at a much lower level than a guy with 50 and a guy with 25.

[imaioral]

Much better scoring points now. We can have a wide points parameter for measuring. Don't change it ok.

[computerrank]

Ratings release 40 is live now.

Top ratings about factor 10 higher.

Lower ratings down to 3 decimals for better visability.

[marcianofan]

Ratings release 40 is live now.

Top ratings about factor 10 higher.

Lower ratings down to 3 decimals for better visability.

WONDERFUL update- thanks so much for all your efforts!

[Freiheit]

Can you explain why Dzemal Bosnjak gained so many points for defeating Osei

http://boxrec.com/en/boxer/834109

[computerrank]

Can you explain why Dzemal Bosnjak gained so many points for defeating Osei

http://boxrec.com/en/boxer/834109

Bosnjak got 2.5 points in the last release. Now he gets 25 points. It is simply a plain factor of 10 in the new release.

These are additional points, which are given for wins against connected boxers, such as Osei here.

Please read the updated ratings descripton. You will find the details, how a boxer can get connected and how a boxer gets the additional points.

[Bbalboa]

Hello my Name is Benjamin Blindert global id 780700. Yesterday i was Ranked 118 in the world (light heavyweight division). Now After the Ranking update i am ranked 474 (10w 0l 0D) Place number 473 has a Record of 1w 8L 0D, can you explain this to me?

[computerrank]

Hello my Name is Benjamin Blindert global id 780700. Yesterday i was Ranked 118 in the world (light heavyweight division). Now After the Ranking update i am ranked 474 (10w 0l 0D) Place number 473 has a Record of 1w 8L 0D, can you explain this to me?

Hello Benjamin,
Henry Mercer got .247 points against Leon Johnson and then declined to .056 points. But he is still a tick above you.
You never got any points from a connected opponent. So you are a tick behind him with .056 points.
In the ratings release before you had a connected opponent with Richard Hilgert. But the requirements were tightened and Hilgert is no longer a connected boxer.
Greetings
Martin

[Bbalboa]

Sorry but that makes absolutely no sense to me. You are talking about Place 472 by the way, not 473. Maybe the new Rating System is better than the old one, but in my case the drop from Place 118 to 474 is a joke and makes no sense, i hope the rating System will change again.

[computerrank]

Benjamin,
Regarding Ehsan Shakeri his win against Alessandro Mammoliti made the difference. More valid than any of your opponents.

[oogiebe]

I've seen some dramatic changes up and down. Do you still see anomalies, or are you satisfied with the current calculations?

[jamamb]

there will never be a rating system that everyone is happy with

[computerrank]

I've seen some dramatic changes up and down. Do you still see anomalies, or are you satisfied with the current calculations?

Drastical changes more down than up due to more restrictions to correct for detected outliers.

Nothing is left, but I am open to go into any case indicated.

[oogiebe]

Drastical changes more down than up due to more restrictions to correct for detected outliers.

Nothing is left, but I am open to go into any case indicated.

I'm good. Your explanation to the prior poster explained how it works.

[JCS]

I think the extra precision makes the ratings ugly as hell.

Why not just multiply by 100? Who gives a shit if somebody has 0.003 points.

[Manrae]

Ratings release 40 is live now.

Top ratings about factor 10 higher.

Lower ratings down to 3 decimals for better visability.

Hey buddy. A little while back I had asked for the rating of the 100th ranked fighter throughout the years, do you mind posting this again now that the ratings have changed ?
Also, will the annual ratings be updated to reflect the new release ?

[computerrank]

Hey buddy. A little while back I had asked for the rating of the 100th ranked fighter throughout the years, do you mind posting this again now that the ratings have changed ?
Also, will the annual ratings be updated to reflect the new release ?

Until next week.