USACFC Competition Schedule and Participants List

[noodle]

Originally Posted by epeemike81

still fails to deal with Rutgers, but a reasonably good plan going forward, IMO.

Should get more accurate as time goes on and also rewards teams which attend every year.

-m

could combine the two proposals, USACFC and USFA based. initially seed teams based on USACFC results, and if you have a new team, insert them approximately into the spectrum where their USFA ratings would have them fall, with perhaps a cutoff at #5 or something. that would require to rank teams both based on USACFC and USFA metrics, and it would still be an approximate science, but it would be a lot better than throwing a Rutgers in at the bottom.

[Wafath]

Originally Posted by TooLoftheDeviL

I'd like to point out that "declaring your affiliation" roughly equates in these cases to "picking your best result" and if you happen to be a school competing in MACFA, and this MFC, and maybe the SAC while you're at it... just picking your best result is a little ridiculous.

For the men, it actually is fairly simple. There is very little overlap between the three major sets of participants to USACFC: MFC, the various NE conferences & the Mid-Atlantic & southern conferences.

There is excessive overlap between the Mid-Atlantic & Southern Conferences. But I don't know of any school that is a member of more than one conference of those three groups above.

Women are more complicated, because in the NE many teams are members of NE conferences and NIWFA. You can't entirely ignore NIWFA because in the mid-Atlantic, that is the only conference until you get down to SAC.

Now, creating your own conference and seeding yourself #1 in it doesn't do you much good because it protects you from no-one. You just get assigned randomly to a pool.

A seeding mechanism should accomplish two things. First, it should make near-equal strength pools, and second, it should protect you from teams you compete with regularly during your season.

W

[Wafath]

Originally Posted by noodle

could combine the two proposals, USACFC and USFA based. initially seed teams based on USACFC results, and if you have a new team, insert them approximately into the spectrum where their USFA ratings would have them fall, with perhaps a cutoff at #5 or something. that would require to rank teams both based on USACFC and USFA metrics, and it would still be an approximate science, but it would be a lot better than throwing a Rutgers in at the bottom.

Keep in mind that 14 of the 37 schools did not participate in last years meet. That's over 1/3 of the field that we would be seeding randomly.

I postulate that for women's events you would be randomly seeding a significant part of the field if you do it on USFA metrics.

Neither of which deal with the protecting by region/conference affiliation. (but such a protection could be combined with either of the above.)

W

[oiuyt]

Originally Posted by Wafath

Now, creating your own conference and seeding yourself #1 in it doesn't do you much good because it protects you from no-one. You just get assigned randomly to a pool.

Being #1 in a conference protects you from other #1 schools. To the extent possible. And the likely correction to make up for there being more #1 schools than pools will result in protection from a #2 school being in the pool.

-B

[noodle]

Originally Posted by Wafath

Keep in mind that 14 of the 37 schools did not participate in last years meet. That's over 1/3 of the field that we would be seeding randomly.

1) a decent number of schools, although they didn't participate last year, did participate further back in the past.
2) basing their seeding on their combined/weighted USFA ratings isn't that random. it might be somewhat inaccurate but it isn't random.

Neither of which deal with the protecting by region/conference affiliation. (but such a protection could be combined with either of the above.)

then attach a conference affiliation to it as well, and have seeding also consider conference protection just like USFA seeding considers teammate protection, seeking to optimize club affiliation dispersion as well as ratings distribution.

[oiuyt]

Originally Posted by Wafath

Neither of which deal with the protecting by region/conference affiliation. (but such a protection could be combined with either of the above.)

Only if the person/people doing the seeding are willing to adjust it to take regional proximity into account. While that was the #1 seeding criteria last year, it was completely disregarded.

The argument was "I can't disadvantage a team by lowering them in the seeding." Which, of course, meant that no adjustments for regional balance could be made at all.

-B

[seak]

Actually if you look at the way wafath is snaking it and there are more the three conferences, you will still have another #1 seed in your pool (there are only 3 pools).

USFA doesn't help with the vast majority of the women's fencers. Also it requires that to a certain extent accurate rosters are sent in ahead of time.

Having a semi-knowledgeable, impartial person running the tournament, so designated ahead of time, and being paid to do so, and not coaching any of the teams fencing fixes a lot of problems.

[TooLoftheDeviL]

Originally Posted by Wafath

Keep in mind that 14 of the 37 schools did not participate in last years meet. That's over 1/3 of the field that we would be seeding randomly.

I postulate that for women's events you would be randomly seeding a significant part of the field if you do it on USFA metrics.

Neither of which deal with the protecting by region/conference affiliation. (but such a protection could be combined with either of the above.)

W

Just off the top of my head, the 14 "new" schools include Boston U, Tufts, Vermont, and Temple - all of which have competed before. Their results within the past 4 years could still be weighed in, but not as strongly as a team who consistently travels to the tournament. So not exactly "random" placing.

It does still fail to capture Rutgers, whom we can all assume will do exceptionally well in the tournament. But as new schools continue to enter the tournament, it should be rare to see exceptions like schools of Rutger's quality enter the mix.

Not terribly unlike the unrated foreign fencers that might roll up into a USFA event. They all go in as unrated, but not all of them are world-cup champions.

[noodle]

Originally Posted by seak

USFA doesn't help with the vast majority of the women's fencers. Also it requires that to a certain extent accurate rosters are sent in ahead of time.

i disagree and would postulate that it could possibly be a more accurate predictor of success than the men's ratings.

[oiuyt]

Originally Posted by seak

Actually if you look at the way wafath is snaking it and there are more the three conferences, you will still have another #1 seed in your pool (there are only 3 pools).

Only doesn't help if the number of existing conferences already matches the number of pools and adjustments aren't made at the #2 seed and below level to account for the "imbalance" at the #1 seed tier.

Let's say there are 3 pools and 4 real conferences. Suddenly you've guaranteed that your team won't end up in the pool that has 2 #1's.

Especially if you team would normally be a #5, #6, #7-type team, you effectively get rid of a top-level team and replace it with another team of your level. That should result in a gain of about half a (squad) victory. So figure an extra 1-2 bouts won in the pool round.

Originally Posted by seak

Having a semi-knowledgeable, impartial person running the tournament, so designated ahead of time, and being paid to do so, and not coaching any of the teams fencing fixes a lot of problems.

Absolutely.

It can be tricky finding people that are both semi-knowledgeable on this specific sub-domain and not closely affiliated with one of the participating clubs, however. Only requiring not coaching helps (eg I'm not coaching any participating club, I am, however, clearly affiliated with one), but might not be sufficient guard against an appearance of conflict of interest.

-B

[TooLoftheDeviL]

Originally Posted by oiuyt

Only requiring not coaching helps (eg I'm not coaching any participating club, I am, however, clearly affiliated with one), but might not be sufficient guard against an appearance of conflict of interest.

-B

If it helps any, I'd put my trust in you, Brad :P

[LordShout]

Two pools would also help, much easier to divide into two equal strength pools then three. If this means the individual comp has to get cut way down, so be it.

I can't possibly approve of using conference standings, certain conferences (MFC) are above and beyound insanely harder to compete in and produce good results, even once you elminate the varisty competition and look only at club results its still fairly unbalanced.

[Wafath]

Originally Posted by TooLoftheDeviL

If it helps any, I'd put my trust in you, Brad :P

I'd be amused to see what happens if we did Seed By Brad.... (maybe a little too amused though)

---

BTW, to publicly answer a PM, this conversation is not entirely academic. The algorithm posted earlier is exactly what the software currently does, including the step "the BC will do a sanity check".

I am not convinced that this is the best one, just a workable one that will piss off everyone equally.

Finally, if anyone wants to help test the code, esp if you have experience with Java & Eclipse & SWT, pm me. I hope to get to fully-functional prototype by Friday evening.

W

[Wafath]

Originally Posted by LordShout

I can't possibly approve of using conference standings, certain conferences (MFC) are above and beyound insanely harder to compete in and produce good results, even once you elminate the varisty competition and look only at club results its still fairly unbalanced.

Are you saying that the relative seeding of clubs in the MFC does not reflect their relative strength? I am a little confused as to why. Could you explain more?

W

[seak]

Originally Posted by oiuyt

Only doesn't help if the number of existing conferences already matches the number of pools and adjustments aren't made at the #2 seed and below level to account for the "imbalance" at the #1 seed tier.

Let's say there are 3 pools and 4 real conferences. Suddenly you've guaranteed that your team won't end up in the pool that has 2 #1's.

Especially if you team would normally be a #5, #6, #7-type team, you effectively get rid of a top-level team and replace it with another team of your level. That should result in a gain of about half a (squad) victory. So figure an extra 1-2 bouts won in the pool round.



Absolutely.

It can be tricky finding people that are both semi-knowledgeable on this specific sub-domain and not closely affiliated with one of the participating clubs, however. Only requiring not coaching helps (eg I'm not coaching any participating club, I am, however, clearly affiliated with one), but might not be sufficient guard against an appearance of conflict of interest.

-B

It's true, and if we could figure out relative conference strength it would be awesome, just don't know how one would go about doing it. Regardless of system there's going to some amount of error.

As to USFA how do you see a team with one B and two U's vs a team with 2 E's and a U etc etc. Also large percentages of the women have U's yet some are clearly significantly stronger then others. Yes to a certain extent it would make sorting the very top from the bottom easy. But you have lots of squads with one top and the rest U's etc. And I would argue that the variance between U's is large enough so as to completely distort seedings when you're talking about 75% of the field being U's.

And hey I'm not currently affiliated with/nor coaching any club team snickers

[epeemike81]

Originally Posted by seak

As to USFA how do you see a team with one B and two U's vs a team with 2 E's and a U etc etc.

well, at a USFA team event you simply assign each rating a value:
A = 1000
B = 2000
C = 3000
D = 4000
E = 5000
U = 6000

Thus, the B and 2 Us would be 10,000, the 2 E's and a U would be 16,000

the B and 2 Us are seeded higher.

-m

[epeemike81]

btw, as to the overall seeding algorithm that Wafath is suggesting, I'd just like to echo Brad's point. It gives an advantage to teams from smaller conferences (I'd guess that this year at least it will benefit the western teams).

If conference A has 7 teams and conference B has 3 teams, the last place team in conference B gets seeded above over half the teams from conference A.

This is quite likely to present some absurd seedings.

-m

[KD5MDK]

Originally Posted by Wafath

Are you saying that the relative seeding of clubs in the MFC does not reflect their relative strength? I am a little confused as to why. Could you explain more?

W

I think they're saying MFC is a much tougher conference than the others and deserves higher seeding. This is why they won Men's Foil last year.