(Algorithmic) HEAVYWEIGHT RANKINGS

Welcome to our Community
Wanting to join the rest of our members? Feel free to Sign Up today.
Sign up

Haulport

Quis Custodiet Ipsos Custodes
First 100
Jan 17, 2015
3,111
4,213
And I just want to brag a little :) that my purely mathematical algo is giving me results that may not be exactly perfect in everyone's eyes, but pretty fucking close to what most humans believe is correct and there is ZERO massaging of the numbers to get there. Well, Cain's inactivity is a massaged but, honestly, after he pulled out of this last fight I'm not sure if anyone would have a problem with me taking him out of the rankings until he starts fighting regularly again.

What say you?
 

Wild

Zi Nazi
Admin
Dec 31, 2014
95,148
138,132
And I just want to brag a little :) that my purely mathematical algo is giving me results that may not be exactly perfect in everyone's eyes, but pretty fucking close to what most humans believe is correct and there is ZERO massaging of the numbers to get there. Well, Cain's inactivity is a massaged but, honestly, after he pulled out of this last fight I'm not sure if anyone would have a problem with me taking him out of the rankings until he starts fighting regularly again.

What say you?
I wouldnt mind seeing him removed. He just looked like total garbage his last time out and now he's injured again.
 

WoodenPupa

Member
Feb 14, 2015
2,919
3,564
I think a lot of what motivates your sense of rankings, and I say this with no disrespect, is that you work under the philosophy that a certain rank limits someone's opportunities, particularly title shots. To me those are separate decisions. #2 does not have to face the champ automatically. Rankings, imo, should serve as a guideline for tourneys, championship fights and general match-ups and reflect the achievements of fighters at any given point. This philosophy frees me to simply examine what has been happening for each fighter within a time range (I keep a cutoff point that the algo doesn't go back any further).
I'll throw my 2 cents in here. I have to disagree whole-heartedly on your rankings rationale. IMO by far the greatest benefit in achieving a good algorithm is that it will neutrally determine matchups. If the algorithm is good enough to accurately reflect the achievements of fighters, it's good enough to determine who they fight. Otherwise, what are the achievements for? Why should any fighter go into a fight not knowing how much closer the bout brings him to a title fight?

I do understand you're not the biggest fan of title-based MMA and prefer tournaments. I'm okay with either, so long as they're approached correctly. And the way to approach them correctly is by having an impersonal method of determining who fights who. This method is what a rankings algorithm should serve. Otherwise, we're just playing with numbers for the sake of it. If you believe in your algorithm, you should believe in its power to match up fighters based on justice.

What are 90% of thread topics about whenever a new, significant matchup is announced? About whether the fight is justifiable based on accomplishments. If the algorithm is good, it should tell us who deserves what.

As for Ben, his win over Alistair and the chain of wins of Alistair and his opponents had created Ben's opportunity to rise in the ranks. He goes ahead of Cain because Ben has a greater point total for the time period and in reality, I broke my own rule because Cain has been SO inactive that technically he doesn't even qualify to be ranked. I put him in and shared this with everyone because I am at a bit of a loss regarding how to handle this kind of inactivity. You have the former champ barely fighting to the point of not deserving to be ranked but if you lower the amount of qualified fights a guy needs to be ranked than you may very well get some very stupid fighters/can-crushers introduced into your rankings (which is a COMMON and a HORRIBLE mess that the vast majority of ranking sites out there have).
Heard! Totally agree with you---this is probably the biggest headache a rankings programmer can have. I don't think I have any big insights here. But let me try.

First, I think most will agree that the kind of uncertainty we're dealing with in Cain is different from the kind of uncertainty that we have with brand new fighters, guys fresh from the amateur/small promotion circuit. We know more about Cain that we do about someone newly signed, with 6 bouts under his belt, who has looked amazing but against limited competition.

But what exactly is the uncertainty with Cain? In his case, it isn't really ring rust IMO. With Cain, it's more about his injuries. The fact is, Cain is constantly hurt, and this should make us very hesitant to reinstate our full confidence in him as a fighter. It's much easier to have this confidence in a fighter whose layoff was due to contractual disputes, as opposed to injuries.

So this sets up the two main categories for inactivity: (1) injury based, and (2) contractual.

My thought has always been that long periods of inactivity don't demote a fighter's ranking, but rather, take him off the list entirely. Especially if the rankings are used to determine matchups, a long-dormant fighter can't retain a spot anywhere because it would affect the ability of other, lower-ranked fighters to get closer to their title shots.

If we take a fighter off the list, the question obviously becomes where to reinstate him on the list when he comes back. I think we can put him right where he was if his inactivity was contractually based. We have no reason to think him a lesser fighter unless such a long time has gone by that the game as a whole looks to have surpassed him. His accomplishments don't diminish, either.

The only exception to this would be if the fighter in question had a title shot coming when he dropped out. Since in the meantime someone else was probably given this shot, this reinstated fighter can't, if that bout is still upcoming, just take the contender's place. He should have to wait the results of the title bout and then fight the winner of it. If we aren't talking about title shot, then the reinstated fighter resumes his former ranking, displacing whoever happens to be in that spot.

If the layoff was injury-based, another set of considerations come into play. These considerations won't matter for short layoffs. Or maybe for anything up to a year. It's only when you have a fighter with extremely long layoffs that we should not allow him to retake his former ranking. And I think injury-based layoffs should affect our confidence in a fighter. To accept this reasoning, you have to be down with the idea that rankings aren't just about accomplishments, but confidence in the ability to perform. If it were only about accomplishments, then no one should ever be penalized in the rankings, no matter how much time he takes off. That to me just doesn't seem right.

So rankings have to reflect not just accomplishments, but expectations of performance. If you're ranked #3 in the division, you should be able to fight like it. All the more so if you're the champ. I'm among those who would be very skeptical of GSP reaching his former levels of performance if he were to come back in, say, June of this year. I simply don't have that confidence. He would need to establish himself again. To what extent, I'm not sure. I surely don't demote him out of the top 10. I would say top 5---but it's very debatable.

As for Cain? I have even less confidence in him than GSP at this point. He's just constantly hurt, and I have a lot of doubt that he can return to the level he was at when he was champ. I wouldn't have put him in a title rematch in the first place, after losing to Werdum. I'm very against instant rematches like that anyway.

At any rate, I don't think we can ever say the guy who just lost a title fight should be ranked out of the top 10. Neither should his injuries place him out of that range. What complicates my line of thought is that Cain's recent injury won't keep him out that long. 4 weeks, isn't it? But consider all the injuries he's had, and the cumulative amount of time he's had off because of those injuries. We can't possibly place the same faith in him that we had in him even AFTER the loss to Werdum. He's now not just the guy who lost to Werdum, but he's injured yet again!

Having said that, I can understand a rejection of this entire rationale of "confidence" in how we expect a fighter to perform. But as I said, a purely accomplishment-based rankings system seems unfair in some respects. So the problem is how to weigh time off in terms of a penalty. My (conjectured) solution involves the recognition of two categories of layoff: contractual and injury-based. Long amounts of time off, even contractually based, should come with a penalty because we're not sure how the fighter is keeping up with his skills, and so on. Long amounts of time off (whether linear or broken up over several instances) that involve injury should bring into play an extra penalty because the fighter has something to prove that he didn't before, namely that his body will perform as it did before the injury.
 

WoodenPupa

Member
Feb 14, 2015
2,919
3,564
And I just want to brag a little :) that my purely mathematical algo is giving me results that may not be exactly perfect in everyone's eyes, but pretty fucking close to what most humans believe is correct and there is ZERO massaging of the numbers to get there. Well, Cain's inactivity is a massaged but, honestly, after he pulled out of this last fight I'm not sure if anyone would have a problem with me taking him out of the rankings until he starts fighting regularly again.

What say you?
One question I wanted to ask is, in coming up with the algorithm, do you make a rankings list that you deem fair and then try to make the algorithm duplicate it? If so, at what point will you concede that the algorithm is right? (keeping in mind this involves multiple layers of circular mental fuckery---you're the author of the algorithm, after all, so you're always the person who "told" it to come up with the rankings it spews out).
 

Haulport

Quis Custodiet Ipsos Custodes
First 100
Jan 17, 2015
3,111
4,213
One question I wanted to ask is, in coming up with the algorithm, do you make a rankings list that you deem fair and then try to make the algorithm duplicate it? If so, at what point will you concede that the algorithm is right? (keeping in mind this involves multiple layers of circular mental fuckery---you're the author of the algorithm, after all, so you're always the person who "told" it to come up with the rankings it spews out).
Nope. It just takes people's records and how they win and lose into account. There is zero weighting according to previous (or "previous" if you get my meaning) rankings. The only thing like that is you can't be ahead of someone who beat you with some recency even if you have a higher point total and you have to be defeating people in "qualified" orgs who have a certain number of qualified fights in those orgs. For example: if you beat the crap out of a bunch of guys in King of the Cage and then come to the UFC and beat another first timer in the UFC who never fights in the UFC again, you won't be ranked and that UFC fight against the 1-timer won't ever count.
 

Haulport

Quis Custodiet Ipsos Custodes
First 100
Jan 17, 2015
3,111
4,213
WP: I will address your bigger post when I'm not at work : )
 

maurice

Posting Machine
Oct 21, 2015
1,359
2,300
It's a good list, but I have no idea why pre-retirement Schaub was rated so highly.
 

Ghost Bro

Wololo ~Leave no turn unstoned
Nov 13, 2015
8,511
10,799
Christ, I still haven't gotten to it...
The first point he makes about your algorithm doing matchups as well would be quite a nice addition to have as a theoretical blueprint, if the ranking system is running satisfactorily.
 

Haulport

Quis Custodiet Ipsos Custodes
First 100
Jan 17, 2015
3,111
4,213
The first point he makes about your algorithm doing matchups as well would be quite a nice addition to have as a theoretical blueprint, if the ranking system is running satisfactorily.
Eh, I don't think match-making should be mathematical. It should be about what is interesting to the fans within the confines (guidance) of a ranking system. Just pushing whoever hits the #1 slot against the champ could end up making some boring fights or make you miss some good ones. For example, if a fighter always loses to the champ (think Faber) but loses to no one else, he would clog the rankings. And really, how could you say that anyone is ranked ahead of him since he isn't losing to anyone besides the champ? You'd have to have him fight the guy you want to go against the champ but that could be a waste of a fight since this clogger isn't going to get a shot at a champ he already or recently lost to. BUT it would be interesting to see how that works out over a period of time to see what the fights would be:

Right now Werdum should be fighting Big Ben.

I'll throw my 2 cents in here. I have to disagree whole-heartedly on your rankings rationale. IMO by far the greatest benefit in achieving a good algorithm is that it will neutrally determine matchups. If the algorithm is good enough to accurately reflect the achievements of fighters, it's good enough to determine who they fight. Otherwise, what are the achievements for? Why should any fighter go into a fight not knowing how much closer the bout brings him to a title fight?
 

WoodenPupa

Member
Feb 14, 2015
2,919
3,564
Christ, I still haven't gotten to it...
LOL! No prob---I'm by far the worst offender of fading from interesting threads. Life is just a tad more important. Still, there are plenty of threads I should have gotten back to (mostly UG stuff, and going back years...the fog thickens in my mind, yo...I can't even remember if I...no wait...I DO know kung fu. I know kung fu!! *looks at claw hands in wonder*)
 

WoodenPupa

Member
Feb 14, 2015
2,919
3,564
Eh, I don't think match-making should be mathematical. It should be about what is interesting to the fans within the confines (guidance) of a ranking system. Just pushing whoever hits the #1 slot against the champ could end up making some boring fights or make you miss some good ones. For example, if a fighter always loses to the champ (think Faber) but loses to no one else, he would clog the rankings. And really, how could you say that anyone is ranked ahead of him since he isn't losing to anyone besides the champ? You'd have to have him fight the guy you want to go against the champ but that could be a waste of a fight since this clogger isn't going to get a shot at a champ he already or recently lost to. BUT it would be interesting to see how that works out over a period of time to see what the fights would be:

Right now Werdum should be fighting Big Ben.
I'm definitely alive to (at least some of) the limitations of filtering everything through an equation. Part of my motivation surely lies in my disdain for the fickle nature of Zuffa's decision makers. Dana in particular seems to let personal feelings alter the course of events when it comes to matchups. An algorithm would prevent this sort of thing---which of course is why Zuffa would never let the numbers dictate who fights who.

I don't buy the particular objections regarding fears of boring fights or missed matchups. There are no guarantees a fight will be boring, and I don't even think we're much better than 50% at predicting this kind of thing. Styles are important, but we are often surprised by what happens. As for great matchups missed---they're missed in the human-choice system all the time as well. As it stands, I think most people want to see fighters who have momentum face each other, and for the most part these end up being those fighters who are ranked by the media in the top ten.

The point you make about the ranking-clogger is a good one, but in reality I don't think we have to worry about it too much for the simple reason that most fighters are not satisfied with being unable to beat the champ even while staying ahead of the rest of the pack. When that happens, the fighter in question usually jumps weight classes, takes a long break, moves to another org, or retires. Perfect example: Rich Franklin, who after the 2nd loss to Silva fought Lutter at 185, and then jumped weight to fight Dan Henderson at 205. As I recall, Franklin wasn't forced to jump a weight class. I do remember Dana/Zuffa being nonplussed at the idea of him getting a title shot again, but I don't think he was forced to do so.

At any rate, keep in mind that I fully realize how much easier it is to talk about an algorithm doing wondrous things than to actually formulate and program it. But that said, while I don't think technology is the savior of the human race, and I'm quite removed from the transhumanist fetishism which seems so popular these days, I do believe that mathematics and logic can be made to surrogate for excellent human judgment in the interest of fairness in certain matters. Over matters of life and death, no. But MMA rankings and matchups? I believe it's doable. But it's no more than a belief, based on some conjecture and maybe a bit of idealist optimism.

You get TONS of respect in my book, Haulport, for being a doer when it comes to all this stuff. Still a huge fan of your proposed rule set and judgment parameters, and I'm equally impressed by the fact that you've actually programmed a ranking algorithm. We might differ on the potential and ultimate purpose of the tool, but the conversation is easy. Actually doing it? Worlds difference. Respect, brother.
 

mmadotcomisufcdotcom

Well-Known Member
Oct 26, 2015
886
885
With Cain #3 my guess is algorithm does not take into account fights per year, long periods of inactivity, fights pulled out of for injury, or being prone to injury.

Excuse me I'm just a negative Nancy right now.
 

Haulport

Quis Custodiet Ipsos Custodes
First 100
Jan 17, 2015
3,111
4,213
With Cain #3 my guess is algorithm does not take into account fights per year, long periods of inactivity, fights pulled out of for injury, or being prone to injury.

Excuse me I'm just a negative Nancy right now.
Actually it does. My algo removed him from the rankings, but I put him in there to not catch shit : (

After his well-timed pull-out against Werdum, I think everyone would be ok with the algo leaving him out so I can leave that alone in the future unless something really goes sideways.
 

mmadotcomisufcdotcom

Well-Known Member
Oct 26, 2015
886
885
Actually it does. My algo removed him from the rankings, but I put him in there to not catch shit : (

After his well-timed pull-out against Werdum, I think everyone would be ok with the algo leaving him out so I can leave that alone in the future unless something really goes sideways.
Well it is the internet, someone is going to bitch on both sides. Damned if You do damned if you dont kind of thing.

Would you be willing to humor me? Would you be so kind to re run your algorithm but remove the short notice loss from both Nelson's and Hunt's records?
 

Haulport

Quis Custodiet Ipsos Custodes
First 100
Jan 17, 2015
3,111
4,213
Well it is the internet, someone is going to bitch on both sides. Damned if You do damned if you dont kind of thing.

Would you be willing to humor me? Would you be so kind to re run your algorithm but remove the short notice loss from both Nelson's and Hunt's records?
I would be happy to if I can find the time (a lot of manual crap on my part - unfortunately I'm not an engineer, just know some sql and R and all my engineer homies hate sports like it killed their moms - except the Russians like Soccer and the Indians like Cricket). And I will try and find the time. Could you give me a reference for which losses specifically you are referring to?
 

mmadotcomisufcdotcom

Well-Known Member
Oct 26, 2015
886
885
I would be happy to if I can find the time (a lot of manual crap on my part - unfortunately I'm not an engineer, just know some sql and R and all my engineer homies hate sports like it killed their moms - except the Russians like Soccer and the Indians like Cricket). And I will try and find the time. Could you give me a reference for which losses specifically you are referring to?
Sounds about right with cricket and soccer.

Hunt vs Werdum, Hunt on 3 weeks notice had to shed 22lb. Had been on a solid win streak, 5 maybe?

Nelson vs Miocic, Nelson on 3 weeks notice, maybe 5 weeks after a fight. 3 fight win streak going in.

Side note, Nelson v Miocic replace a Barao fight. Where is Barao?