Rating system - yet again

[needle]

As oiuyt said, this has been discussed previously (extensively.) I think the USFA Board of Directors has even discussed it formally and rejected the idea. I like the idea, in theory. It's a pretty good system that forces competitors to "live up" to their rating. Basically, if you're the seeded #3 in a competition, you gain points for placing 1st or 2nd, break even if you finish 3rd, and loose points if you finish 4th or below. Of course, there are a lot of disadvantages, but I think the biggest problem is the issue that keeps coming up: organizers of local competitions and even the USFA have a hard time keeping the current system straight despite the fact that it could not be much simpler. I am sure there are many, many fencers out there who have classifications that they earned due to an error in the reporting of the results or a misapplied rule and others who do not have a classification that they should have due to such oversights. I have acted as BC chairman at a dozen or so events and have received more than one call from fencers who were still waiting for the USFA to change their classification weeks after I reported the results.

Thanks. If it was discussed before, I'm not going to restart it. I get the explanations (yours and oiuyt's) - I'm not happy with them, but I understand the reasons why this was rejected

[Feraud]

I understand the math geniuses out there all wanting to come up with some new perfect system and maybe it’s a fun exercise for your master’s thesis.

There’s just no point or useful purpose to develop a new rating system, which has the main purpose of seeding competitions.
And it’s also worth pointing out that any flaws that exist in the current model would also be present in any new system as well. This is because you’re trying to assign mathematical perfection to something that is inherently random; i.e. the results of a fencing tournament. You could have the same competition with the same participants, held two times and get two different results. So there’s completely no use in trying to assign some hyper-accurate skill classification to each fencer in the country. For top-level athletes there’s already a point standing, used to determine important rankings such as selecting a National team. But there’s no need to have hyper-accurate perfect seeding in every competition for every fencer. And it would be completely irrelevant for this classification to change up or down after every single competition.
Never mind the enormous practical limitation on implementing something that even more complicated than the current model. How would the National Office handle the change in ranking every weekend from thousands of fencers? How are the fencing moms who run tournaments going to understand a system involving logarithms? Are you people just out of your minds?

[Inquartata]

In all candour, I don't these new "improved" ratings schemes that crop up every now and then ( like weeds ) are meant as seeding methods. I think some people want a hyper-accurate system so that they can know exactly how many people they are "better" than...

[veeco]

Hi!



That is tenth - not natural - logarithm, right?

Right. Otherwise I would have noted it ln(), not sure if that's the correct English notation, but that's the notation I grew up with...

One could have used FIE and French rating points directly, and gotten less leveling effects. As it is now, competitions with quite different real strengths will get the same SF.

Yes, but then the math-challenged people wouldn't have understood it ;-).

The 0.8 floor rule will increase the value of weak competitions, and thus spread out points to many more fencers. If nobody in the top-32 is rated either in the FIE or French top list, then this rule will bump the SF from 0.32 to 0.8. You could include 5 fencers from the French top-8 list, and the competition would still not get a higher SF - bump from 0.77 to 0.8 in that case. Therefore, it is possible to get a lot of points by searching out weak competitions. Drive around the countryside, win locals, and get your rating up!

Yes, I think it was done with that purpose in mind exactly. The point was that the designer of the system wanted all competitions to have a minimal SF to make sure that all tournaments had at least a minimal appeal in terms of ranking. That was because they viewed the non floored system as a self-fulfilling prophecy: people who fenced in a .32 SF comp wouldn't come back the next year, meaning that those you where there the next year wouldn't come back, and very quickly noone would show up to the least strong tournaments.

Oh, and I forgot to mention, that this system is only used for national competitions. There are no A,B,C... rankings in France. You fence in a national tournament, you get points (as little as 0.8 sometimes, but sometimes much much more if you happen to win the national championships were everyone shows up, or the Monal GP, where most of the top FIE fencers come). So your idea about driving up and down the countryside doesn't really work. But it might still be a nice ride though ;-).

As it is now, there are vrey few tournaments where noone from the top 16 of the FFE points shows up, so the competition is strong at all national tournaments, and I guess that warrants in some way the .8 SF.

There was actually a tournament that was being held for veterans with an other interesting formula in France lately, that used some transitivity to rank your results in the tournament. Sort of unrelated, but if you're into maths I could try and dig up the formula that was being used for you and translate it.

[michel]

Just for kicks (and the non-math challenged) I thought I'd put down here the formula used to compute national points in France. It came up from the brain of a "Polytechnicien" which means someone who graduated from Ecole Polytechnique, which can be summarized as "the French MIT".

And he was a fencer too. I think he was even French Fencing Federation Chairman.

Let p be your final placement in the tournament.
Let n be the number of fencers in the tournament.
Let SF be the strength factor of the tournament (more on that later).

(1.01 - log(p) / log(n)) * SF * 100.

To compute the strength factor, one must do the following:

This computes with the top 32 ranked fencers in the tournament.
FIE ranked fencers from 1 to 32 in FIE points are worth 10 points.
FIE ranked fencers from 33 to 64 are worth 5 points.
French ranked fencers from 1 to 8 are worth 10 points.
French ranked fencers from 8 to 16 are worth 5 points.
If there are more fencers in the top 32 ranked fencers who haven't been counted at this point, they are worth 1 point.

Add up all the numbers above, and divide it by 100, and if the number you obtain is lower than 0.8, then change it to 0.8. And you've got your SF. Why divide by 100 to multiply it again after? I have no idea why, don't ask me, that's the way it is.

I think it is because of the Clas ranking software code.

[KD5MDK]

The problem with that system is that it presumes fencers from the FIE points list or from the top 32 in the country are present. Well, I suppose the .8 factor means they don't have to be, but all tournaments which have none of them are then the same. I can't imagine that working in the US.

[veeco]

Well, as a matter of fact, some of them generally aren't... But the others make up for it in the bigger tournaments, and the flooring of the strength factor gives a nice incentive to middle of the pack fencers to attend as many tournaments as possible.

Generally there are about 4 big tournaments in the ME national circuit with a SF over .8. Add to that the Monal challenge, which is included in the French national points ranking, and the national championships, which usually have SF of over 3.00 you get some good opportunities to rack up a lot of points if you are one of the top 20 fencers in the nation. For others you need to work your way up, and that is probably IMO the biggest flaw of the system. Only a truely exceptional fencer will break into the top standings in just a season.

[swordsen]

Teh best way to implament one of these systems woudl be for Peet to just stick it onto ask fred and start doing it with every tournament that posts results there. I think it would be a very interesting experiment.