Ratings - please read before commenting - Archived

[computerrank]

BoxRec Whole History Ratings are live now ...

[jerd16]

So let me get this straight - the inactive/historical fighters records will now only display the WHR alongside each fight? Why?

I much preferred the previous archived ratings, because then you could gauge a boxers standing at that time. WHR-only ratings obscures that.

[computerrank]

So let me get this straight - the inactive/historical fighters records will now only display the WHR alongside each fight? Why?

I much preferred the previous archived ratings, because then you could gauge a boxers standing at that time. WHR-only ratings obscures that.

I is just a different perspective. Before we had a rating before the bout and one after the bout. Now we have a rating at bout time. This rating changes with career from bout to bout. - The difference is: the ratings at any bout time depend on the results before AND after the bout. The perspective is from all history.

[jerd16]

Right. I understand the reasoning. I just think it doesn't provide an accurate assessment of the accomplishment at the time of the bout or that era. Look at John L Sullivan for instance. This system murders his accomplishments.

So the previous system is not returning?

And as I was just getting acclimatised with the previous iteration its changed AGAIN. How long before this one is overhauled?

[computerrank]

Right. I understand the reasoning. I just think it doesn't provide an accurate assessment of the accomplishment at the time of the bout or that era. Look at John L Sullivan for instance. This system murders his accomplishments.

So the previous system is not returning?

And as I was just getting acclimatised with the previous iteration its changed AGAIN. How long before this one is overhauled?

We are not fixed to a specific perspective. We try to provide ratings, which represent the real results as good as it is possible. We measure the differences between results and our ratings and try to minimize them.

A further opportunity now with these WHR ratings: We will be able to update the ratings with each bout coming new into the system or changed by the editors. This option is currently tested and will be added next.

[marcianofan]

So let me get this straight - the inactive/historical fighters records will now only display the WHR alongside each fight? Why?

I much preferred the previous archived ratings, because then you could gauge a boxers standing at that time. WHR-only ratings obscures that.

I is just a different perspective. Before we had a rating before the bout and one after the bout. Now we have a rating at bout time. This rating changes with career from bout to bout. - The difference is: the ratings at any bout time depend on the results before AND after the bout. The perspective is from all history.

If I understand you correctly, the new system applies hindsight to the results of fights. So Fighter A loses to Fighter B, and initially he takes whatever hit he would tend to take based on the result. But then Fighter B loses a fight to some nobody in the next fight, which lowers Fighter A's points going into his prior fight to Fighter B?

Is that right? I think that's a pretty terrible idea, if so. It completely ignores the idea of fighters having good or bad nights, or for the judges' scores not reflecting the fight, and the idea that styles make fights, among other things.

Looking back and editing history based on the future makes for a poor and unstable system, in my view. I know a big part of your goal is to make rankings predictive of fight outcomes for the purposes of increased credibility, but I think this goes way too far in that direction. In my view, the ratings ought to reflect a fighter's standing at the time the fight happened, period.

The constant fluctuation of the numbers and rankings are also a really big deal. I use your rankings as a basis for setting up my own. More specifically, I use your specific numbers to evaluate the fighters I don't yet rank in the top 20 that I publish so that I can put non-top-20 wins in context. The rankings update that was targeted at rewarding connectibility and eliminating the Raymi effect I think was a good one, but the massive shifts it caused made me stop publishing my rankings because I had to go back and reconcile all the previous numbers, though this was more manageable once top 50 fighters stopped having only a few points. With the lockdown in effect, I started doing the needed revision in my own files about 3 days ago. I paused on Eric Molina's profile last night, and he was ranked #37. When I woke up this morning, he was ranked #87, and instead of the #50 fighter having 32.89 points, the #50 fighter now has 6.705. Back in 2015 when I was still regularly updating my rankings, the borderline of the top 50 was about 116. When I started the rankings in 2010, the borderline was about 150. It's one thing to tweak the rules by which the rankings are determined, but to massively crash the scale itself is a disaster for any semblance of consistency.

Honestly, I'd take some idiosyncratic or questionable rankings if they were consistent and predictable, as opposed to perfect rankings that shift massively beneath us every few months.

Also, I know there are fighters who take Boxrec rankings very seriously, and make their opponent's ranking a significant factor in their choice of opposition. How would you feel if you'd scheduled a fight against a top 50 fighter yesterday, only to find out he's now ranked #105 or something, and you're now just staying busy instead of climbing the ladder?

If I'm understanding correctly and shifting numerical rankings for past fights are now going to be a feature and not a bug, I'd like to respectfully request that we be given access to previous models via radio button so that we can keep consistency.

And please, please, please, when the numerical scale collapses like it seems to do with every update, could you do us a favor and adjust it so that we're at least seeing similar numbers for similar rankings from one day to the next? It's one thing to say you'd overrated Eric Molina and he's now #86, but to say a #50 ranking is now worth less than 1/5 what it was points-wise previously just seems blatantly arbitrary. If fighters shift around based on different modeling, I could handle that, but when the scale itself changes drastically, it makes me scrap everything I've been doing for the last 10 years. Or at least re-calculate everything, which I could do if it wasn't for the fact that everything could massively re-scale while I'm in the middle of doing so, as it did today.

[Rad#33]

I really hate this new rating system, maybe something could be done to give members the option to look at whatever ratings system that they prefer.

[computerrank]

@marcianofan,
- You are right. For the past later results influence former ratings at bout time now, too. An it is very logical to consider these later results for former ratings, exactly because a former result might have been due to a good day and not the whole truth, when rating a former career point.
- Additionally the WHR ratings have better predictive power than the past release. And I mean for future bouts without knowing the future.
- Yes, the scales changed, but in fact the rankings based on these ratings are the point that counts.
- And marcianofan, I well understand your issues, when you rely on a fixed Boxrec ratings system. But we always said, there is no Boxrec preference for a specific system, as long as it matches past results and predictive power best.

[marcianofan]

Respectfully, I think you're trying to be both reflective and predictive in equal measure, but you're never going to get both in equilibrium. So you're going to have a severely handicapped version of both all mixed up in the same pot. If you're going to change the rankings of past fights based on later performances, you might more accurately reflect how good that fighter was at the time- I'm not denying that. But it says nothing (or at least it says much less) about what that fighter had accomplished, which I think is the higher purpose when you're doing rankings. I don't begrudge the effort to be predictive or to show a fighter's ability level rather than his accomplishment level (ie Mike Tyson has 50-something points going into his first fight, while other debut fighters are near zero, because he went on to accomplish more). That's actually an interesting and innovative approach.

But I think what is obvious and, to me, more important is that we understand the context of the fight at the time it happened, no matter what happens thereafter. If Fighter A switches trainers and goes on a winning streak and revives his career, it doesn't mean that he wasn't a washed-up no-hoper at the time he lost to fighter B several months before. I don't think history is well-served when we turn upsets into non-upsets and vice-versa retroactively based on what happens after. I think the rankings have, to this point, done a solid job of predicting outcomes simply by reflecting a fighter's quality based on his present resume. The future is uncertain always, and I think the effort to mitigate that uncertainty by essentially rewriting the past is to destroy the very good in a Quixotic quest for the perfect.

Would you consider implementing a ratings toggle for predictive vs. reflective, so that the system in effect prior to today will still be accessible (and ideally fairly static), as well as the new one for the purposes you're leaning more toward now? Obviously it goes without saying at this point that if time or computing resources make parallel systems impractical, I'd strongly prefer the previous, reflective-emphasized system be the one that survives. But I'm hoping everyone can be made happy still, since I know just from this thread that I'm not the only one who feels this way.

By the way, just so I'm not being overly negative, I want to congratulate you on the new scoring feature, which I love.

[JCS]

I had my opinions of this system before it launched.. but man, I never appreciated how strange it would be going back in history and looking at pre-bout ratings... seeing a peak fighter's rating decrease multiple fights before it deserved to.

Look at Tyson's rating take a dump in fights before losing to Buster Douglas, while he's still blasting out his opposition. A guy who fought a lot and often, like Sugar Ray Robinson... has a rating that is on a continuous rollercoaster ride.

Of course it has a great prediction rate... but the movement will be nonsensical from a standard perspective as well some of the placements.

What a strange look.

[margaret thatcher]

I wonder who else has a p4p list with Fury at 2 and Wilder at 5

[HanzWorst]

This has to be a late april's fools joke. I don't get how a fighter loses points for winning fights. That makes no sense at all. Just look at Wladimir Klitschkos ratings. Or how can Tyson Fury gain 100 points for being inactive for 3 years? Please bring back the old ratings, it wasn't perfect but it made much more sense and was way more transparent.

[marcianofan]

I had my opinions of this system before it launched.. but man, I never appreciated how strange it would be going back in history and looking at pre-bout ratings... seeing a peak fighter's rating decrease multiple fights before it deserved to.

Look at Tyson's rating take a dump in fights before losing to Buster Douglas, while he's still blasting out his opposition. A guy who fought a lot and often, like Sugar Ray Robinson... has a rating that is on a continuous rollercoaster ride.

Of course it has a great prediction rate... but the movement will be nonsensical from a standard perspective as well some of the placements.

What a strange look.

Tugging at the Tyson thread a little more, I also think it's unsupportable that, despite his rating already having tanked for no reason prior to the Douglas fight, it was still higher that night than the night he beat Trevor Berbick. That I think highlights the issue. It's trying to analyze both what previous fights meant and what future fights tell us in retrospect, and I don't think those two pursuits are compatible in the same set of ratings.

[computerrank]

I had my opinions of this system before it launched.. but man, I never appreciated how strange it would be going back in history and looking at pre-bout ratings... seeing a peak fighter's rating decrease multiple fights before it deserved to.

Look at Tyson's rating take a dump in fights before losing to Buster Douglas, while he's still blasting out his opposition. A guy who fought a lot and often, like Sugar Ray Robinson... has a rating that is on a continuous rollercoaster ride.

Of course it has a great prediction rate... but the movement will be nonsensical from a standard perspective as well some of the placements.

What a strange look.

Tugging at the Tyson thread a little more, I also think it's unsupportable that, despite his rating already having tanked for no reason prior to the Douglas fight, it was still higher that night than the night he beat Trevor Berbick. That I think highlights the issue. It's trying to analyze both what previous fights meant and what future fights tell us in retrospect, and I don't think those two pursuits are compatible in the same set of ratings.

Berbick wasn' rated that high at that fight. It is quite consistent so.

[computerrank]

...
Would you consider implementing a ratings toggle for predictive vs. reflective, so that the system in effect prior to today will still be accessible (and ideally fairly static), as well as the new one for the purposes you're leaning more toward now? Obviously it goes without saying at this point that if time or computing resources make parallel systems impractical, I'd strongly prefer the previous, reflective-emphasized system be the one that survives. But I'm hoping everyone can be made happy still, since I know just from this thread that I'm not the only one who feels this way.
...

It is an option with the new ratings system to have multiple ratings in parallel. Boxrec could by selector show a version, witch adapts faster to the career results. This is due to the fact, that these ratings can be calculated with less computer power.

It will also allow to let the ratings respond instantaneously to each bout added or changed.

[marcianofan]

...
Would you consider implementing a ratings toggle for predictive vs. reflective, so that the system in effect prior to today will still be accessible (and ideally fairly static), as well as the new one for the purposes you're leaning more toward now? Obviously it goes without saying at this point that if time or computing resources make parallel systems impractical, I'd strongly prefer the previous, reflective-emphasized system be the one that survives. But I'm hoping everyone can be made happy still, since I know just from this thread that I'm not the only one who feels this way.
...

It is an option with the new ratings system to have multiple ratings in parallel. Boxrec could by selector show a version, witch adapts faster to the career results. This is due to the fact, that these ratings can be calculated with less computer power.

It will also allow to let the ratings respond instantaneously to each bout added or changed.

Any idea when that might be implemented?

[computerrank]

It is an option with the new ratings system to have multiple ratings in parallel. Boxrec could by selector show a version, witch adapts faster to the career results. This is due to the fact, that these ratings can be calculated with less computer power.

It will also allow to let the ratings respond instantaneously to each bout added or changed.

Any idea when that might be implemented?

It is on my list. But there are a lot of other possible optional enhancements for Boxec. So there is no schedule for that so far ...

[jerd16]

Any idea when that might be implemented?

It is on my list. But there are a lot of other possible optional enhancements for Boxec. So there is no schedule for that so far ...

Is there any way to get access to the previous ratings database?

[computerrank]

It is on my list. But there are a lot of other possible optional enhancements for Boxec. So there is no schedule for that so far ...

Is there any way to get access to the previous ratings database?

No

[Perutech]

Hello, I am the manager/matchmaker by Wanik Awdijan. He got very bad rankings about the new system. He was much better at the world and german rankings too. I would like to get an info How can happen that? Example his opponent from the past Mohlongo has got much better rankings. Wanik was second at the german rankings now he is 7th. He was 61. at the world, now 101. We are so sad about that! Can you help us about that? Thank you very much!!! https://boxrec.com/en/proboxer/631859

All the best!

[computerrank]

Hello, I am the manager/matchmaker by Wanik Awdijan. He got very bad rankings about the new system. He was much better at the world and german rankings too. I would like to get an info How can happen that? Example his opponent from the past Mohlongo has got much better rankings. Wanik was second at the german rankings now he is 7th. He was 61. at the world, now 101. We are so sad about that! Can you help us about that? Thank you very much!!! https://boxrec.com/en/proboxer/631859

All the best!

BoxRec changed to a new ratings version yesterday. These ratings more reward continuous recent good results, more than single good results. Wanik Awdijan's last best opponent was Mateo Damian Veron and he dropped in the ratings, because he only had a single good result against Lolenga Mock too.

[marcianofan]

I don't know how much additional criticism is welcome, but I guess I want to point this out to highlight the need for some significant adjustments, even taking for granted the underlying principles of the new system.

All-Time Heavyweight Rankings:
Melio Bettina is pretty significantly better than Jack Dempsey, who is only slightly better than John Ruiz. John Ruiz is significantly better than Ken Norton.

Bettina, along with other 2nd-tier Joe Louis-era contenders like Jimmy Bivens, Bob Pastor, Elmer Ray, Clarence Henry, and Roscoe Toles (all ranked above Ken Norton and all but Toles ranked above Dempsey) seem to be vastly and systematically overrated. It doesn't pass the eyeball test, I think most would agree.

[DazBoxingFan]

So tevin farmer is number 2 in the world, yet he lost to joe diaz who’s now number 7. How can you beat a fighter and be ranked below him? It has no logic

[computerrank]

I don't know how much additional criticism is welcome, but I guess I want to point this out to highlight the need for some significant adjustments, even taking for granted the underlying principles of the new system.

All-Time Heavyweight Rankings:
Melio Bettina is pretty significantly better than Jack Dempsey, who is only slightly better than John Ruiz. John Ruiz is significantly better than Ken Norton.

Bettina, along with other 2nd-tier Joe Louis-era contenders like Jimmy Bivens, Bob Pastor, Elmer Ray, Clarence Henry, and Roscoe Toles (all ranked above Ken Norton and all but Toles ranked above Dempsey) seem to be vastly and systematically overrated. It doesn't pass the eyeball test, I think most would agree.

All time ratings are a very special discipline. In fact you cannot compare the times. So nobody knows really.

[computerrank]

So tevin farmer is number 2 in the world, yet he lost to joe diaz who’s now number 7. How can you beat a fighter and be ranked below him? It has no logic

Not a fight makes a boxer, but good continuous results over a time period. Let Diaz confirm his good result and we will see.