USFA membership/website/rating system update

[misha]

Originally Posted by oiuyt

I'm sure there are others that we won't even consider until after the system is in place and we find the hidden traps.
So we deal with them as we find them.

That is NOT the way to approach building of any system....

Example... Let's see .... ahh... USFA website...

Instead of using common methods, ask someone you know
without particular expertise in the subject to make something
special, using unconventional stuff...
Do not request any warranty, do not pay for support...

"So we deal with them as we find them."

oops website is up only 61% of the time...


.

[misha]

Originally Posted by KD5MDK

Then you get to an event where the top 100 people are in one chunk, the 101-180 are in the second chunk, and the 181-220 are in the third...

that means you need to make more chunks

.

[Inquartata]

Originally Posted by telkanuru

If only there were some sort of interconnected network of machines capable of transmitting and accessing vast quantities of data almost anywhere in the world that could be utilized. Some sort of international web, or network. That would be perfect!

You should drop Al Gore a line. I'll bet he could come up with something.

[Inquartata]

Originally Posted by misha

that means you need to make more chunks

.

And then more, and more...until hey presto! We're back to "a fine rating system that needlessly distinguishes between the 1500th best men's epeeist and the 1501st best".

Or, you know, we could just stop trying to fix something that isn't broken...

[JMcC]

Originally Posted by Inquartata

Or, you know, we could just stop trying to fix something that isn't broken...

I agree.
Just personal observations only, not a thorough statistical analysis; but initial event seeding at national events has, on average, been fairly accurate in predicting the final standings.
Some fencers in the overall field have a great day and finish higher than originally seeded; others fence poorly and finish lower, but most finish around where their current skills and current seeding method predicted.

[darius]

What exactly is the problem we're trying to solve?

If it's granularity between ratings, it would be easy to add a few more letters, or rank by the number of times that a fencer has earned a specific classification, which would provide some extra measure of differentiation between fencers with the same classification.

I don't necessarily have problems with a points or chess-style system, if implemented on the National level.

What I don't like is the idea of having it be affected by local events. Local organizers kvetch all the time about how the good fencers who happen to be in their division don't show up to tournaments. Currently, those fencers have nothing to lose except their entry fees.

Some of us travel heavily, because we use those events to train. However, we're not always fencing to win -- we're fencing to work on stuff. If this had an implication in our national ranking, that would be a major disincentive to fence in these local events. Also, if this were analagous to the USCF system, every match played would affect our score. So you could potentially win the tournament and still lose ranking, if you lost to a sufficiently low-rated fencer in the pools.

darius

[Inquartata]

Ah, the law of unintended consequences...

But never mind that; full steam ahead!

[epeemike81]

Originally Posted by misha

That is NOT the way to approach building of any system....

sure it is.

do you try to anticipate and prevent problems? absolutely.

To think that there won't be any unanticipated ones, though, is hubris. It's much better to have the attitude that there WILL be problems so that you're ready to handle them. if there aren't, so much the better, but I have yet to see it.

-m

[epeemike81]

Originally Posted by JMcC

I agree.
Just personal observations only, not a thorough statistical analysis; but initial event seeding at national events has, on average, been fairly accurate in predicting the final standings.
Some fencers in the overall field have a great day and finish higher than originally seeded; others fence poorly and finish lower, but most finish around where their current skills and current seeding method predicted.

yeah, but at those events, the top seeding IS done by a points system...

-m

[mrbiggs]

Originally Posted by KD5MDK

Then you get to an event where the top 100 people are in one chunk, the 101-180 are in the second chunk, and the 181-220 are in the third...

Any high-level event with that many fencers will almost certainly have a bunch on the points list, who will be seeded higher than everyone else. Then you have, maybe:

1-10 on the points list
90-100 in the first chunk
101-180 in the second chunk
181-200 in the third chunk

So each pool will have either 1 on the points list and two in the first chunk or three in the first chunk, about two in the second chunk, and maybe one in the third. Sounds fine to me.

[Preacher]

I'm not to educated on matters like this. I read the article, the OP, and the connecting thread that Brad put up and I still can't find why the system of rating needs to be changed.


To me both the Letter and the Numerical system have just as many problems as one another. (in theory...the numerical system is not in place yet)


I also don't know who really cares about being seeded 4th instead of 6th at a tournament. In a Nac maybe but then again the people that are on the points standings are seeded accourding to them anyway. So what if a few A's not on the points list get jostled around during seeding. If you fence well and are a good fencers you'll make it farther.


-P

[oiuyt]

Originally Posted by mrbiggs

Any high-level event with that many fencers will almost certainly have a bunch on the points list, who will be seeded higher than everyone else. Then you have, maybe:

1-10 on the points list
90-100 in the first chunk
101-180 in the second chunk
181-200 in the third chunk

So each pool will have either 1 on the points list and two in the first chunk or three in the first chunk, about two in the second chunk, and maybe one in the third. Sounds fine to me.

Until the two in the first chunk with the poor points-list fencer are in actually both in the top 20 of the event, something that was masked because the top 75 fencers all have current-year first-chunk classifications.

-B

[Sciurus-Rex]

Originally Posted by misha ((responding to comment, "we deal with them as we find them."

That is NOT the way to approach building of any system....

No. You're right. In a perfect world, everything is planned out to such a degree that nothing could ever possibly go wrong once we start up the system.

And yet society still stumbles along -- somewhat productively -- still fixing previously unforseen problems as they appear.

It's not an either-or scenario. You don't stop progress entirely because the plan hasn't been given the "Fool Proof(tm)" stamp of approval.

[oiuyt]

Originally Posted by Preacher

I also don't know who really cares about being seeded 4th instead of 6th at a tournament. In a Nac maybe but then again the people that are on the points standings are seeded accourding to them anyway. So what if a few A's not on the points list get jostled around during seeding. If you fence well and are a good fencers you'll make it farther.

With all else being equal, the difference created by seeding ranging from massively overseeded to massively underseeded can change the expected number of pool victories by as much as 1 (the wider the spread in abilities in the tournament the closer this value gets to 1).

A difference in expected result of 1 victory -- or any significant portion thereof -- is huge in seeding into the next round, especially if it's a DE format. This results in significantly higher expected finishing placement.

Seeding does matter.

Then there's the point of wanting even-strength pools -- the entire purpose behind the seeding in the first place. The more quirks/flukes/noise in the seeding process the less even the pools will end up. Rarely updated, high-water mark, systems with relatively low granularity are a good way of ensuring lots of quirks. There's a reason that at higher levels -- whether the top 50ish people in the country in a given weapon/age or the top 300ish people in the world -- such a system is not used. Will there still be quirks in any system? Yes, of course. Is it worthwhile to switch to a system with fewer of them? That depends. It depends on the one-time switching cost, it depends on the on-going marginal increased cost, and it depends on the benefits -- both in improvements to how the system does its job and in secondary benefits such as appearence of objectivity, accessibility to those from outside the sport (who can understand "I'm the 305th ranked sabre fencer in the US" but don't know what "I'm a B" means), and flexibility to be used for other pruposes.

Specifically what prompted this current incarnation of the discussion?

The Tournament Task Force is looking at ways of reorganizing the national tournaments. One of the things being looked at is changes in how people qualify for entry into them. One of the ways being examined of how to do such qualification is having a numeric ranking of people that covers all/nearly all of the USFA rather than just the top ~50. With such a ranking there are a lot more possibilities for how to do such qualification. As such a discussion piece was generated to start exploring what would be needed in order to have such a system, what the various options might be, what the benefits and drawbacks of each are, and to solicit input from the general membership.

As a discussion piece it seems to have accomplished at least a part of its job. :)

-B

[epeemike81]

Originally Posted by Inquartata

Or, you know, we could just stop trying to fix something that isn't broken...

Inq, I'm surprised at you!

Don't you realize that using a points list for the elite while only using a generic bucket system for the rest of us is unfair and elitest???



-m

[Sciurus-Rex]

Originally Posted by darius

What exactly is the problem we're trying to solve? ...

I don't necessarily have problems with a points or chess-style system, if implemented on the National level.

What I don't like is the idea of having it be affected by local events. Local organizers kvetch all the time about how the good fencers who happen to be in their division don't show up to tournaments. Currently, those fencers have nothing to lose except their entry fees.

Some of us travel heavily, because we use those events to train. However, we're not always fencing to win -- we're fencing to work on stuff. If this had an implication in our national ranking, that would be a major disincentive to fence in these local events.

OK, so consistency nationwide isn't something worth striving for. We don't need any sort of sound basis for intra-section comparison so that an A or B from, say, Texas can be seeded fairly against the same rating of a fencer from New York or Idaho.

Personally, though, I want my competitive sport to be comprised of people who are not afraid of losing status. People who strive to perform at their best, not avoid challenges because they feel uncertain about the legitimacy of their accomplishments. Competition, after all, implies that someone is trying to attain something or take it away from at least one other person. If a national ranking system is a disincentive -- and I don't believe it is -- for some fencers, then those fencers don't deserve to keep their rating to begin with.

The assumption that a national ranking is a disincentive is flawed anyway. A fencer may find himself losing or gaining some small amount of status in the middle of a list of thousands, but the stay-near-home competitor can still boast of being at the top of his region.

[Preacher]

As the system stands now.....you can be on the points list and still have a rating of a B. In the number system would you get bracketed into a lower slot number during the switch over or would you be ranked as you currently are in the points standings?

[oiuyt]

Originally Posted by Preacher

As the system stands now.....you can be on the points list and still have a rating of a B. In the number system would you get bracketed into a lower slot number during the switch over or would you be ranked as you currently are in the points standings?

The answer to this question obviously depends on the eventual implementation of the swtich-over. Given that there BEING a switchover, much less what form it will take has not been decided this seems somewhat premature.

That said, I find it hard to believe that the eventual mechanism (if and when) wouldn't try to take into account such obvious information as current senior points standings.

-B

[mrbiggs]

Originally Posted by oiuyt

Until the two in the first chunk with the poor points-list fencer are in actually both in the top 20 of the event, something that was masked because the top 75 fencers all have current-year first-chunk classifications.

-B

Well, there is of course going to be some error. But the really big discrepancies that mess up a tournament (people who fence at a level 2 or 3 letters off their current classification) aren't going to change just by assigning numbers to the letters. Most of the ones I've seen in those circumstances are coming out of retirement, are not from the US, or don't compete enough to have earned the classification that most with their skill level would have earned.

[telkanuru]

It seems that a lot of headache could be avoided by just making ratings degrade every 2 years instead of 4...