mmaranks makes no sense

"I've entered tens of thousands of fights since, (before there were even Sherdog or FCFighter databases), so I wouldn't chalk it up to laziness."

My apologies. I realize you've put a lot of work into this and that obviously you have not been lazy about it. It's just frustrating because it feels like you're on the right track but have stopped just short of something really great.

"I understamd every point you have tried to make about where and why some tweaking might produce what you think it should. I've heard them all before. Will you beliveve me if I tell you that some of your ideas HAVE been tried, and simply didn't work?"

Honestly, I wouldn't. Nothing I've suggested has not been implemented with great success by the BoxRec people. I almost feel bad constantly comparing you to them but they really are the gold standard in computerized boxing ratings and I feel we would do well to emulate them.

"If a very successful fighter goes on some ridiculous spree of fighting unworthy opponents, then I WANT his rank to suffer. I like that part and you don't. Therein lies a difference between us."

You misunderstand. Of course that should hurt someone's rankings. The scenario I came up with is just one way in which the arbitrary 6-fight limit can cause problems. My point is simply that having a fight drop from being counted 100% to being counted 0% all of a sudden is a major weakness of your system. There needs to be a more "sliding scale" type of mechanism in place. The way things are set up now, it's possible for a fighter to fight twice on the same day, and have one of the fights count and the other not count. That just doesn't make any sense.

"Both stats and votes have shortcomings."

Stats have positives and negatives. Voting systems, on the other hand, I see no positives for. IMO, they are not a sensible way of doing things at all. Man-made ratings should be created in an intuitive fashion by a single person or a committee. There isn't a single highly-regarded set of boxing rankings that are voted on. Either one person like Dan Rafael of ESPN.com makes his own rankings, or a publication's editors get together and come up with rankings they can all agree on.

Not really. If you just look at the stat rankings many of them are totally absurd. Jardine is the #4 LHW whereas Silva isn't even in the top 10. Rory Markham is #1 at 185, not counting Almeida who is retired. Goulet is #3 at 170 despite just getting KTFO by Ludwig who barely ever fights. The system is not the worst thing I've ever seen, but it has SERIOUS flaws.

I think you're tricking yourself into seeing "serious flaws" because you're not allowing yourself to accept what you're looking at in the stat ranks.

You and I would both see Wanderlei as the favoured fighter if Jardine vs. Silva happened, but that's not what we're searching for here.

The stats ranks is measuring recent success rates, so why would it seem "totally absurd" to suggest that being 5-0-1 recently would be better than being 4-2 recently?

Nowhere is anyone suggesting that Jardine is a better fighter than Silva. All it is saying is that he is a more successful fighter recently.

And no, it's not just a simple matter of all 1.000 records are better than all .500 records.

It is weighted, but Wanderlei has two split decision wins, Mark Hunt doen't have an overly impressive profile despite having a couple good things on it. Yoshida's profile isn't mind-blowing either. Quinton has some weaker-thans on his recent dance card, Nakamuara is half-and-half lately..... all this adds up.

Get past what you've seen and know about the skills and/or level of each fighter, and look at the numbers.

Were talking about;

Fighter A:

5 clean wins and a draw

versus

Fighter B:

2 clean wins, 2 split wins, 1 clean loss and 1 split loss

Hopefully you can see where pure stats would favour Fighter A.

"I think you're tricking yourself into seeing "serious flaws" because you're not allowing yourself to accept what you're looking at in the stat ranks."

I think you're tricking yourself into not seeing serious flaws because you're forcing yourself to accept what you're looking at in the stat ranks.

"The stats ranks is measuring recent success rates, so why would it seem "totally absurd" to suggest that being 5-0-1 recently would be better than being 4-2 recently?"

Well, first of all, this brings back the issue of the 6 fight limit. Second of all, Vanderlei has 2 wins against top 5 competition. Jardine has never fought anyone anywhere near that level. Also, personally I would count Jardine's loss to Rashad.

I understand the point of "aging" a fight. I really do. Bellieve me when I say that I've toyed around with it a lot.

But it became one of those things where it made perfect sense in one set of circumstances, but totally counter-productive in another.

It sort of ties in with the cronological vs. dynamic debate too.

Aging, together with the static vs. dynamic point awards created some cases where it worked well, and others where it arguably worked poorly.

BJ Penn and Fujita are favourite examples of mine.

Both burst onto the scene and slowly started making waves. When we first saw them one would assume that they were newbies scoring major upsets. Fujita over Mark Kerr???? Surely a devastating upset loss for Kerr, right?

But as we became more familiar with Fujita through fights to come, we could say that Fujita was a legitimately solid competitor. I think even in our minds, we started to forgive Kerr a little more, each time Fujita showed well.

Same with BJ Penn. Surely we began to forgive Paul Rodriguez more and more for losing to BJ, once it became increasing evident how good BJ was.

I completley understand the flipside point, where it was debated that someone like Wanderlei could experience a life-changing scenario where his career could go down the drain. In this case each successive loss that Wanderlei might suffer would statistically hurt those who beat him when he was at a pinnacle. Fair point, and as you can see, using the same system provides a converse effect to the BJ and Fujita examples.

So is it a damned if you do, damned if you don't situation?

My take on this particular scenario has always settled on this other factor.... Sure Wanderlei losing 4 in a row would hurt Arona's rank, but none of that stops Arona from continuing to seek out wins over those who are now the elite.

If Wanderlei loses 4 in a row, surely we'd all essentally consider him out of the picture, so it's then up to Arona to beat whoever IS ranked, if he wishes to maintain a lofty status.

That all speaks more toward the dynamic system though. I think the dynamic system works best.

As for aging a fight result gradually... again it may seem perfectly fine in one case, but you'd see other cases where you wouldn't like the results it gave you.

It almost works opposite to those same scenarios I described for the dynamic system.

When Wanderlei was smaking Sak around three times, one might argue that Sak was at his best the first time, and became progressively worse and more broken-down each time. If older fights were worth a lesser percentage, then what was supposed to be Wanderlei's best win over Sak, would become his most worthless win, simply because it had aged longer.

I guess what I'm saying is, both the aging and the dynamic systems will help or hurt you depending on which way your opponent happens to be headed.

Actually, aging would help lessen the damage of a loss, which I guess is another reason I'm not sold.

Through aging, losing to Fedor would be less damaging to Nogueira than it would be for Cro Cop, simply because Nog did it first.

In my book, both guys lost to Fedor, so should both suffer the same deduction. I don't really care if one happened 8 months before the other. If anything, you can argue that Fedor was a more solidified fighter when he fought Cro Cop, thus should hurt Cro Cop less. But aging would do the opposite.

Then of coarse there's the deal where Fedor fights Nog on one NYE show, and Zuluzinho on the next. If I aged fights, it would reduce the value of the Nog win more than it would reduce the value of the Zulu fight.

I shudder to think that I'd be giving a win over Nog and a win over Zuluzinho similar point awards, simply because one happened earlier.

I want Fdor to get full marks for beating Nog until he moves on and fights enough other people to put Nog in the past.

I think you're tricking yourself into not seeing serious flaws because you're forcing yourself to accept what you're looking at in the stat ranks.

Sorry, I didn't mean what you're just supposed to accept what you see, I meant you weren't acknowledging what the stat ranks is intended to produce.

It can't list the most skilled guys. It's a math formula. It's only meant to evaluate success rates.

Well, first of all, this brings back the issue of the 6 fight limit. Second of all, Vanderlei has 2 wins against top 5 competition. Jardine has never fought anyone anywhere near that level. Also, personally I would count Jardine's loss to Rashad.

But again, you're reffing to something that the stats list is not claiming to portray. I'll assume you're refferring to Rampage and Arona as the top 5 competition Wandy has beat recently. Well STATISTICALLY, Rampage is hardly top 5, and Arona is currently 7th. So no, Wandy hasn't beaten 2 top 5 statistical guys, and that's what we're talking about here.

Actually, Nakamura is the best win in Wandy's recent past because Silva is only getting credit for 2/3 of the win over Arona, (split decision). In two fights with Arona, Wandy is behind... 4 judges to 2.

Call Jardine the tortoise and Wandy the hare I guess. Jardine is ahead by virtue of a nice steady pace with no speedbumps. Wandy has some highlights, but also a couple of potholes.

In my perfect world, this spells time for Jardine to be tested against someone of his statistical equal... like Lambert or Liddell or Babalu or Arona. Then we'd get the answers we want.

And in the end, doesn't this spell out quite nicely why the voting aspect helps "correct" the numbers?

With voter input, Wandy ends up at #3, while Jardine gets adjusted to #12.

I think the mixing is a bad idea. A system should be entirely human or entirely computerized.

Well, mmaranks.com offers all of them.... just stats, just votes, and the combination of the two.

Whatever you like.

Maybe im not the brightest guy. But would you like to explain to me how according to the statistical ranking. Yves Edwards is ahead of Kawajiri?

Considering Kawajiri (regardless of how he won the stats still say a W) just beat the guy who beat Edwards and Shaolin win should still count since its his 6th fight

Well, first of all, the Kawajiri/Hansen fight was ruled a no contest.

Beyond that, both are 5-1 recently, with Yves beating more successful guys than Kawajiri overall.

When your statistical analysis come up with 337 people that are ranked higher than Sean Sherk at 170lb, you should start questioning your formula.

First, the system didn't come up with 337 people at 170 pounds who are above Sherk, it came up with 337 people at all weights.

Second, Sherk has these recent wins working for him... Gerald Strebendt, Darin Brudigan, Brodie Farber, Lee King, and Joel Blanton.

No disrespect to those guys, but I don't think it's a list of who's who in MMA.

You're assesing Sherk's skills instead of looking at the numbers associated with his recent accomplishments.

LOL @ Vera being ranked higher than Overeem.

If only someone laughed out loud to the awkwardly unfunny humour you displayed in the ring at Freedom Fight. :\

Uh, yeah anyway, I guess the rest of us already knew that Overeem is low because he'd only been in the division for 1 day when that list was updated, so his votes were obviously a little light.

But you see, the development of a statistical system has to be based on qualitative appraisal.

Let me know how to have a statistical system that measures the quality of a fighter without bias.

IMO a particular win over a fighter should be "crystalized" in association with the loser's ranking AT THE TIME OF THE FIGHT... ie. if you beat the #1 ranked fighter in your weight class, and then he loses a bunch of fights subsequently and falls to #10 in the rankings, your win should still be weighted as a win over the #1 fighter..... because fighers change over time. that is the biggest flaw in the mmaranks system: it doesn't account for the fact that fighters change over time.

"Plus, if he was top 10 at 170, you'd figure he'd be top 10 at 155, and he's nowhere to be found there, either. "

Wait a minute - how can a guy be ranked at a weight he's never fought at before? That makes no sense.

"Well STATISTICALLY, Rampage is hardly top 5"

However, he was much higher ranked when Silva beat him than he is now. So basically you're penalizing Silva for beating Rampage down so badly that it hampered the rest of his career. This is why I don't think later results should influence how much earlier results are counted. This is really a serious problem, because even if a guy was the best fighter in the world when you beat him, if he has a long injury afterward or simply starts going downhill, you don't get appropriate credit for beating him when he was on top of the heap.

"Let me know how to have a statistical system that measures the quality of a fighter without bias."

BoxRec actually uses a system of statistical self-appraisal combined with common sense. They actually take a predictive rate, i.e. at every point in history, at what percentage does their rankings correctly predict the winner. Right now I think this percentage is around 67% for them. However, someone there did come up with a system that had a higher predictive rate (like 70%) but they haven't adopted it because it allows prospects and inexperienced fighters to climb the rankings much more quickly. This may lead to them being ranked above guys they can probably beat, but it doesn't really produce a more accurate look at who's accomplished what.

Also, back to the 6-fight cap point. Have you tried decaying the point value of a fight as it ages, regardless of how many fights have come before and after it? Something like 2.5% per month, so after a while the fight counts for nothing. I simply think that quality of opposition being equal, a guy should be rewarded for fighting more often than his peers.

Actually it was ruled a DQ win for Kawajiri. Since your site is based on pure stats. A win is still a win. And that means Kawajiris last 6 wins include 2 former Shooto champions and one of wich recently beat Yves

The word I got was that is was later ruled a no-contest. Sherdog's DB has a DQ, but their play-by=play person who was covering the fight says it was ruled a no-contest.

The FCFighter DB has the no-contest listed.

Paul Rodriguez never fought BJ. Unless you were just making up an example ;)

Shit, I always mix up Joey Gilbert and Paul Rodriguez. I have no idea why. :[ I meant to say when BJ beat Joey Gilbert.

IMO a particular win over a fighter should be "crystalized" in association with the loser's ranking AT THE TIME OF THE FIGHT... ie. if you beat the #1 ranked fighter in your weight class, and then he loses a bunch of fights subsequently and falls to #10 in the rankings, your win should still be weighted as a win over the #1 fighter..... because fighers change over time.

I understand your point Jimmy, I honestly do. But as I said, it works both ways. The example you give is correct, but there is a flipside.

I don't know if I put a lot of stock into a person's ability to steal someone's "aura". I'm not sure I believe that Wandy beating Rampage the first time automatically signalled the demise of Quinton's career.

In any case, nothing stopped Wandy from taking those big points at the time, and continuing to add to them. Just because Quinton goes for a dump, doesn't mean Wandy has to sit idly by and watch his point total drop. He is free to beat up on the next guys who are up there.

See my examples of BJ Penn and Fujita, to see why it makes things more clear when it goes the other way.

Since a lot of times we have no way to really know how good someone is at a particular point in time, the dynamic aspect helps everyone's value to become better defined within the timeframe used.

When further developing how the formulas work, I was in talks with someone who knew about chess rankings. Ironically, he said that chess rankings often fail when they aren't dynamic. Like MMA, if in chess there isn't a lot of cross-over competition, then it's hard to gague where you're really at if you just take results at their "crystalized" or static value.

If a group of people play chess in Ohio for most of their lives, we can establish a ranking for them, but it bares no reference to how I stand amongst other groups. Now if one of them go and play in Texas, then we now have a reference point for how Ohio chess compares to Texas chess.

Then it's fair to reference everyone who has played Mr. Ohio and everyone who has played Mr. Texas.

Sure some people in both states may have become better or worse chess players since they played the people their refenced to, but if you keep the timeframe tight enough, then it is generally fair.

Maybe you disagree, and that's fine, but we figure that with the timeframe in use, (6 fights), we know fighters can improve or become worse, but maybe not a great deal. Obviously there are some exceptions, but it all seems to balance out OK.

If you want to see a stat sytem that does some of the things you guys are asking for, check out Sherdog's power stats. They award points based on how someone was doing at that particular time, and they allow for all fights to count, with no cap.

Their HW list includes the 3-7 Ray Seraille at #8. Why? Because Travis Fulton was doing well at the time Ray beat him.

Guys like Travis Fulton, Dan Severn and Jeremy Horn are exactly the reason why I changed my stats to dynamic ranks limited to 6 fights.

I want a snapshot of who has recently beaten guys with recent success of their own. That is what I want from the stat ranks, and that is what I get.

The play by play had a slight mixup because someone at the event incorrectly called it a no contest. It was ruled a DQ win for Kawajiri and it has not been changed. Other fights that ended the same way were DQs aswell. Ryan G or Rich can back me up on that. Shitty way to win but stats are stats :=)

OK, just going by what I could find. I will investigate further, and give TK the win if that's the way it was decided by Shooto.

BoxRec actually uses a system of statistical self-appraisal combined with common sense.

Self-appraisal and common sense don't sound like un-biased criteria to me. That actually sounds a lot like voter polling.

"Self-appraisal and common sense don't sound like un-biased criteria to me. That actually sounds a lot like voter polling."

Not at all. If you read my post, it is statistical self-appraisal based on a predictive rate. Of course, there must be a human common-sense element regulating things, but they largely depend on the predictive rate to judge the effectiveness of their system.

^^^^^

I have no idea what that means. :)