BoxRec

If you were tasked with designing a computerised ranking list for active boxers, how would you go about it?

What criteria would you use for rating opponents, how far back or at how many fights would you look, how would you account for losses, etc.?

Pretty much how BoxRec already does it. Reading the justification for their ratings, it's pretty comprehensive.

BoxRec ratings are understandable except the launch state has never been explained to my satisfaction.

Dennis wrote:BoxRec ratings are understandable except the launch state has never been explained to my satisfaction.

When you look at the old time guys their points are rather insane. There definitely seems to be an issue the further back you go. Or the current fighters just suck way worse than all the old timers as some of the older school fans would suggest I'm sure lol.

It would seem like at some point you just gotta make up a number of points for the first guys so everything is gonna be birthed from that.

The starting point is the key. Who you put into the initial top 100, top 50, top 20, etc dictates a lot of what happens from there.

BoxRec has the right idea. I think it needs some tweaking but in general it's a solid system.

Dennis wrote:The starting point is the key. Who you put into the initial top 100, top 50, top 20, etc dictates a lot of what happens from there.

Yea thats what I mean. The rankings are based on the points. Thus at some point when you go back far enough you got a handful of guys if not just the first two guys to have a fight listed in the boxrec database you gotta manually create a point value for a boxer. I'd assume you'd have to do it one of those ways. But at some point yea you gotta invent a number value for guys & that number would reflect on everyone from there.

ReggieDiggs wrote:

Dennis wrote:The starting point is the key. Who you put into the initial top 100, top 50, top 20, etc dictates a lot of what happens from there.

Yea thats what I mean. The rankings are based on the points. Thus at some point when you go back far enough you got a handful of guys if not just the first two guys to have a fight listed in the boxrec database you gotta manually create a point value for a boxer. I'd assume you'd have to do it one of those ways. But at some point yea you gotta invent a number value for guys & that number would reflect on everyone from there.

Pretty sure BoxRec just starts all fighters at 0, and has rules for what happens when two 0-point fighters face each other. The whole thing grows organically from there.

jujigatame wrote:

ReggieDiggs wrote:

Dennis wrote:The starting point is the key. Who you put into the initial top 100, top 50, top 20, etc dictates a lot of what happens from there.

Yea thats what I mean. The rankings are based on the points. Thus at some point when you go back far enough you got a handful of guys if not just the first two guys to have a fight listed in the boxrec database you gotta manually create a point value for a boxer. I'd assume you'd have to do it one of those ways. But at some point yea you gotta invent a number value for guys & that number would reflect on everyone from there.

Pretty sure BoxRec just starts all fighters at 0, and has rules for what happens when two 0-point fighters face each other. The whole thing grows organically from there.

That's right.. unless the boxer debuts with an unusually impressive performance (Lomachenko) in this case they start from higher than 0 (even if the Fighter Page says 0, its not really 0). You can thank me for that one.

I would make it similar to boxrec's implementation. There might be slight modifications to the way weight division shifting is handled. The poor opposition penalty probably needs some minor reworking, such as less penalization if a good opponent in the division doesn't exist, a more gradual decline overall, and possibly a rollback of some of the lost points if the fighter beats his next good opponent. I might also consider linking a poor opposition penalty / inactivity penalty with a fighter's age. I would probably make a slight rating penalty if a fighter is KO'd, since KO's realistically do tend to set a fighter back more than a decision does. There might be a need to slightly offset the effect of fighters carrying their points upward as they move-up in weight, which results in the points pooling in higher divisions. I would also look for a way to reduce the potential point gains in those prizefighter tournaments, because it leads to things like Audley Harrison in the top 25 on the back of 3 straight "legitimate" wins in a single night. I have no clue why past fighters have such absurd ratings compared to present fighters, but I assume it pertains to larger talent pools, and considering that boxing fundamentals and sports science are both improving over time, I think I would implement something to normalize the average / peak ratings across eras, with possibly a minor upward trend.

lLLLLLEEEEEEEEEEEENNNNNNNNNNNNOOOOOOOXXXXXXXX


your time is here.....

Lackeos wrote:I would make it similar to boxrec's implementation. There might be slight modifications to the way weight division shifting is handled. The poor opposition penalty probably needs some minor reworking, such as less penalization if a good opponent in the division doesn't exist, a more gradual decline overall, and possibly a rollback of some of the lost points if the fighter beats his next good opponent. I might also consider linking a poor opposition penalty / inactivity penalty with a fighter's age. I would probably make a slight rating penalty if a fighter is KO'd, since KO's realistically do tend to set a fighter back more than a decision does. There might be a need to slightly offset the effect of fighters carrying their points upward as they move-up in weight, which results in the points pooling in higher divisions. I would also look for a way to reduce the potential point gains in those prizefighter tournaments, because it leads to things like Audley Harrison in the top 25 on the back of 3 straight "legitimate" wins in a single night. I have no clue why past fighters have such absurd ratings compared to present fighters, but I assume it pertains to larger talent pools, and considering that boxing fundamentals and sports science are both improving over time, I think I would implement something to normalize the average / peak ratings across eras, with possibly a minor upward trend.

x2 I like your suggestions

Lackeos wrote:I would make it similar to boxrec's implementation. There might be slight modifications to the way weight division shifting is handled. The poor opposition penalty probably needs some minor reworking, such as less penalization if a good opponent in the division doesn't exist, a more gradual decline overall, and possibly a rollback of some of the lost points if the fighter beats his next good opponent. I might also consider linking a poor opposition penalty / inactivity penalty with a fighter's age. I would probably make a slight rating penalty if a fighter is KO'd, since KO's realistically do tend to set a fighter back more than a decision does. There might be a need to slightly offset the effect of fighters carrying their points upward as they move-up in weight, which results in the points pooling in higher divisions. I would also look for a way to reduce the potential point gains in those prizefighter tournaments, because it leads to things like Audley Harrison in the top 25 on the back of 3 straight "legitimate" wins in a single night. I have no clue why past fighters have such absurd ratings compared to present fighters, but I assume it pertains to larger talent pools, and considering that boxing fundamentals and sports science are both improving over time, I think I would implement something to normalize the average / peak ratings across eras, with possibly a minor upward trend.

Some of the ratings for fighters of past generations are head-scratching, and it's not just the cases of guys having 5000+ points. Herol Graham for instance would be today's P4P world #3 even when he was still competing in British and European title fights and had yet to face world class opponents. One could argue the talent pool was significantly stronger back then, but to that extent? I doubt it.

Deeper talent pool in the past is half the story - the other half is that top fighters used to fight each other far more often.

I would make it similar to boxrec's implementation. There might be slight modifications to the way weight division shifting is handled. The poor opposition penalty probably needs some minor reworking, such as less penalization if a good opponent in the division doesn't exist, a more gradual decline overall, and possibly a rollback of some of the lost points if the fighter beats his next good opponent. I might also consider linking a poor opposition penalty / inactivity penalty with a fighter's age. I would probably make a slight rating penalty if a fighter is KO'd, since KO's realistically do tend to set a fighter back more than a decision does. There might be a need to slightly offset the effect of fighters carrying their points upward as they move-up in weight, which results in the points pooling in higher divisions. I would also look for a way to reduce the potential point gains in those prizefighter tournaments, because it leads to things like Audley Harrison in the top 25 on the back of 3 straight "legitimate" wins in a single night. I have no clue why past fighters have such absurd ratings compared to present fighters, but I assume it pertains to larger talent pools, and considering that boxing fundamentals and sports science are both improving over time, I think I would implement something to normalize the average / peak ratings across eras, with possibly a minor upward trend

I agree with the rest, but would against this.

jezzamundo wrote:Deeper talent pool in the past is half the story - the other half is that top fighters used to fight each other far more often.

This will only get worse.Fighters today could not fight 8 contenders in a year.It just could not happen
with the politics of boxing and tv contracts.
There is no way that this will change in my lifetime.
Also fighters at least rated fighters make more in one match than past champions made in a year.
They will never fight that often when not needed.Their management would never allow it.

Boxrec is good.

It's not a rating though of how good someone is but what they've already done. Most people would agree that Joshua is better than Teper, Tarver, or Arreola, but he is rated lower because he doesn't have the points yet. If someone beat someone, they should be ranked higher with very few exceptions but that is not always the case.

My lack of math skills will keep me from critiquing boxrec too closely. So I will stick to two points: the winner should usually be ranked higher, and it's a ranking of past performance, not ability.

Yes, I think that BoxRec's rankings are generally solid. There have certainly been some questionable placements (I think Mchunu was the #2 CW at some point), but that applies to virtually every set of rankings I've seen.

jezzamundo wrote:Deeper talent pool in the past is half the story - the other half is that top fighters used to fight each other far more often.

I would make it similar to boxrec's implementation. There might be slight modifications to the way weight division shifting is handled. The poor opposition penalty probably needs some minor reworking, such as less penalization if a good opponent in the division doesn't exist, a more gradual decline overall, and possibly a rollback of some of the lost points if the fighter beats his next good opponent. I might also consider linking a poor opposition penalty / inactivity penalty with a fighter's age. I would probably make a slight rating penalty if a fighter is KO'd, since KO's realistically do tend to set a fighter back more than a decision does. There might be a need to slightly offset the effect of fighters carrying their points upward as they move-up in weight, which results in the points pooling in higher divisions. I would also look for a way to reduce the potential point gains in those prizefighter tournaments, because it leads to things like Audley Harrison in the top 25 on the back of 3 straight "legitimate" wins in a single night. I have no clue why past fighters have such absurd ratings compared to present fighters, but I assume it pertains to larger talent pools, and considering that boxing fundamentals and sports science are both improving over time, I think I would implement something to normalize the average / peak ratings across eras, with possibly a minor upward trend

I agree with the rest, but would against this.

http://www.independentworldboxingrankings.com is different to Boxrec mainly for these very reasons. We worked with the Association of Boxing Commissions back in 1998 and with the Ali act/ideas that John McClain put in position.

The Key points why I think IWBR is better is our handling of meaningless fights. Someone like Ali Raymi has never fought anyone who has fought anyone that has or has had a rating. Some fighter may never have beat a top 100 fighter but they have fought/beat someone who has and by that it is easier to measure them. We have a BAR that you need to make before you enter the rankings (which go down to 200+ in most divisions) You could get there with 15 wins all by knock out against 15 'bodies' like Ali Raymi. You can take the route of beating one good opponent of course. The general route of a fighter is that he will have 10 or 12 learning fights where he is supposed to win over journeymen, perhaps learning against a southpaw or a spoiler before fighting another fighter perhaps will a similar record or regional title. If a fighter gets over that hurdle then usually they are stepped up in class until they meet top 100 opposition, then move through rankings ultimately towards a world title.

The IWBR has methods of compliance that you have to pass in order to MAINTAIN YOUR RANKING. At the top we ask that a fighter must have fought another in the top 50 to maintain a ranking in the first 35. If he does not he loses 10% of his points each month until he lies in a position where he is now compliant, ie beyond number 35. That is OUR policy on poor opposition. There are a few other areas of compliance one being that a fighter can't (normally) be in the top 50 without ever having beaten a top 100 fighter. A fighter is moved outside the top 50 if he has not beaten a top 100 fighter in 3 years.

Weight classes have the following codes. A fighter moving up in weight class takes 75% of his points with him in all divisions except Heavyweight where he takes 58% if he moves up from Cruiserweight. There is no penalty for moving up in the lowest divisions, ie up to Super-flyweight. A fighter moving down in weight takes all of his points.

Age is factored and the bar is different at each division a 30 year light-flyweight is ' same degree of older' as a 35 year old heavyweight. They do not get points reduced as long as they keep winning fights and maintain their compliance. When the lose their points for losing are doubled (fall faster). Fighters lose double points when they lose a third consecutive fight, same as 4, 5th etc. A fighter with 6 consecutive losses is unlikely to be seen in the top 100.

Fighters beaten on points do not lose so many points. Fighter losing by knock out can lose more, fighters winning by knock out don't get much more but they do get something but it is insignificant with top 50 fights, the winning is 95% of the point allocation.

We realised pretty early on it was impossible to use our system for old time fighters or for pound for pound. Some divisions are very weak, some strong. Cruiserweight and Light-heavyweight are very weak, wherby Junior-middle, welter and several classes below have a lot more boxers that have achieved.

IWBR rates on what has actually happened, it does not have fighters ranked high up that have 20 meaningless wins, Boxrec does. IWBR ratings will have a fighter you have beaten (in the last year) below you. In Boxrec there are absurd rankings wherby a fighter is above just because of that meaningless win against an Oklahoman two months later.

There is no comparison in the quality of the ratings between the two, but like any system each will have flaws. Glazkov is too high by what we can SEE, sadly it is a very weak division.

I don't think any good rating system can be totally computerised, the appointment of each fighter to his correct division is difficult. If you allot a fighter fighting at 148 to Jnr-middle then most fighters (90%) will be in the wrong division.

We have a major update on its way, we are listing the opposition quality of the last twenty fights (currently it is 6) so there is a lot more information about who has the best resume...... You can question IWBR ratings slightly but they are not far out.

Best regards

Lennox wrote: Weight classes have the following codes. A fighter moving up in weight class takes 75% of his points with him in all divisions except Heavyweight where he takes 58% if he moves up from Cruiserweight. There is no penalty for moving up in the lowest divisions, ie up to Super-flyweight. A fighter moving down in weight takes all of his points.

This stood out to me.... Some fighters move back and forth between divisions a lot. How do you handle this?

JCS wrote:

Lennox wrote: Weight classes have the following codes. A fighter moving up in weight class takes 75% of his points with him in all divisions except Heavyweight where he takes 58% if he moves up from Cruiserweight. There is no penalty for moving up in the lowest divisions, ie up to Super-flyweight. A fighter moving down in weight takes all of his points.

This stood out to me.... Some fighters move back and forth between divisions a lot. How do you handle this?

I credit them back if they are within 12 months.

Lennox wrote:

JCS wrote:

Lennox wrote: Weight classes have the following codes. A fighter moving up in weight class takes 75% of his points with him in all divisions except Heavyweight where he takes 58% if he moves up from Cruiserweight. There is no penalty for moving up in the lowest divisions, ie up to Super-flyweight. A fighter moving down in weight takes all of his points.

This stood out to me.... Some fighters move back and forth between divisions a lot. How do you handle this?

I credit them back if they are within 12 months.

Odd.

What is odd about it.

Fighter has 200 points a Welter,
fights next at 154, drops to 150 points, wins fight to add 30 points. (180)
next fight goes back to 147, and has 230 points.

Do you think he should have 180 points or 230 points?

Lennox wrote:What is odd about it.

Fighter has 200 points a Welter,
fights next at 154, drops to 150 points, wins fight to add 30 points. (180)
next fight goes back to 147, and has 230 points.

Do you think he should have 180 points or 230 points?

What's odd? The whole 12 month thing.. and you get the points going in one direction, or not the other.

In your example, I'd expect he'd have 240. Since he gained 20% at 154, I figure the gain would be proportional.

Lennox wrote:What is odd about it.

Fighter has 200 points a Welter,
fights next at 154, drops to 150 points, wins fight to add 30 points. (180)
next fight goes back to 147, and has 230 points.

Do you think he should have 180 points or 230 points?

Something more than 230 since he earned 30 additional points at the higher weight class which should translate to 40 at 147. That means he should have 240. If moving up makes you lose points then moving down should make you gain points. There are exceptions clearly as Chad Dawson was a dehydrated no-energy fighter at 168. Same thing when Chris Byrd went down to 175 at the end of his career. Some guys cut weight but don't overdue it and can be strong at the lower weight. Same thing with moving up. Some boxers are killing themselves to make a lower weight and actually are much better at the higher weigh class. Pacquiao is an example when he was fighting below at or below 122.

With our rules, no points are gained going back down a weight. The Chris Byrd situation is a good example of why it does not work the same in reverse, and where it does work is at the very lightest weights, where there is no penalty situation anyway for moving up.

He get's his 25% penalised points returned if he quickly returns back to his old division.

He only won 30 points in the last fight.