Men's epee seeding close to breakdown?

[counterattack]

In Men's Epee at the Pacific Coast Section Championships this year the seeding was getting pretty rough. There were 60 fencers in 9 pools, there were more than nine of several ratings, though if you take into account national points it wasn't quite as bad:

A06 = 7 with national points
A06 = 4 without NPs
A05 = 1 with NPs (plus one if Pearce does, but USFencing site is down so I can't check)
A05 = 9 without national points
B06 = 11
B05 = 9
C06 = 6
C05 = 2
D0x = 4
E0x = 1
U = 2

As it was, the pools were a bit of a crapshoot, but not too bad. It seems like with one more year of ratings increasing like this, though, and it will be a complete logjam at the top.

Compare to the 2005 numbers, with 91 fencers and 14 pools:

A05 = 13 (several with NPs, but don't know exactly)
A04 = 6
A03 = 6
A01 = 1
B05 = 15 (probably at least one with NPs)
B04 = 11
B03 = 2
B02 = 1
C05 = 7
C04 = 3
D05 = 8
D04 = 2
D03 = 2
E0x = 4
U = 10

Basically in 2005 we had the fencers spread over 17 ratings (I collapsed the low end ratings if there were only a couple) into 14 pools, with a good tail into the lower ratings. In 2006 we had 11 ratings into 9 pools, with very few fencers at all in the low ratings (2006 has 7 D, E, and U compared to 2005 with 26 D, E, and U).

It seems the problem is as much with more frequently refreshed ratings as it does with inflated ratings since there were no ratings older than the last two years in 2006, yet in 2005 there were lots of them.

Anyway, the question is at what point do the ratings get broken. I don't think we are quite there yet in the PCS, but we are getting close. I think that if there are more people in a given rating than there are pools then that begins to be broken. Since we were right around that level this year, I suspect we may get there next year. But then maybe we just crossed a threshhold that allowed a rapid increase in ratings, and possibly we have now maxed that out and without adding a bunch more tournies within the PCS it is going to be harder to get any more bunched up.

I know we have had plenty of conversation on FNet about the uses of ratings and ratings inflation, but I would like to see some discussion in terms of the math of it or the actual results of it. For example, if there are large buckets of fencers that are equally rated but you assume there is actually a decent spread of ability, then you could be much more likely to get a pool with two very tough fencers of the same rating as those in another pool that are much weaker. I know some of this will go on no matter what, but it seems more likely the more congested the ratings are.

Also, any similar situations in other areas?

[notalent]

Greetings counterattack,
We had 30 less fencers this year with 20 of those coming from the bottom of the heap. It doesn't take a genius to figure out that without some lower ratings to flesh out the pools you are going have tougher pools. I imagine this is a fluke and next year we will have the normal 80-90 fencers with lower rated fencers so you have some easy bouts to help your seeding.

[seak]

The problem is their is no incentive for lower rated fencers to fence sectionals. for what is, at least in the mid - atlantic, a fairly expensive tournament that can be relativly far away.

[counterattack]

Greetings counterattack,
We had 30 less fencers this year with 20 of those coming from the bottom of the heap. It doesn't take a genius to figure out that without some lower ratings to flesh out the pools you are going have tougher pools.

Terry,
I don't think that is true. If you look at the 2004 results (72 fencers) you will see the trend is even worse. There are only a handful of A04 fencers and much much less congestion at the top.

http://pcsfencing.com/senior/srpccs/...ccsresult.html

I don't have time now to do the similar ratings comparison, but a superficial glance shows that they are much more spread out even than in 2005. It seems like 2005 was the beginning of the avalanche, which is probably slowing down now, but the question is will is slow down enough before the ratings are really very little use for seeding at sectionals.

[counterattack]

The problem is their is no incentive for lower rated fencers to fence sectionals. for what is, at least in the mid - atlantic, a fairly expensive tournament that can be relativly far away.

This is true for the PCS also, but I don't think the lack of the lower rated fencers at sectionals is so much a problem. The lower rated fencers get their kicks at their divisional D2/D3 qualifiers.

[Ordway]

This is true for the PCS also, but I don't think the lack of the lower rated fencers at sectionals is so much a problem. The lower rated fencers get their kicks at their divisional D2/D3 qualifiers.

I don't agree. To say "there is too big a proportion of highly rated fencers" is exactly the same thing as saying "there is too small a proportion of lower rated fencers." If there had been plenty of U, E, and D rated fencers to fill out the pools, you would never have ended up with multiple As and Bs and have the problem you are describing.

In women's sabre at PCCs, we also had a fairly top-heavy tournament (by our standards - out of 23 fencers, no As, but 5 Bs and a lot of Cs and Ds; only 1 E and 2 Us. That made for every pool being hard-fought.

Ironically, by having ratings be much harder to get in women's sabre*, we have the same kind of compression issues as in men's epee - instead of having fencers spread out over A, B, and C ratings, we have a lot of compression in B and C, with resulting overlaps in the pools of people who really ought to be in different pools.

* As in, Jen Ngolab and I each won one of the PCS circuit events in women's sabre - tough events -, and not only did we not earn our Bs, we didn't even renew our Cs. Sigh.

I do think it would help matters quite a bit to have more lower-rated fencers at Sectionals; how to get them there is a different matter.

[Redblade]

Too much emphasis gets placed on initial seeding and pools. At the end of the day, the only thing that matters is whether the fencer is good enough to make it through his DE bouts ... regardless of where he was seeded coming out of pool bouts.

[bigdawg2121]

Too much emphasis gets placed on initial seeding and pools. At the end of the day, the only thing that matters is whether the fencer is good enough to make it through his DE bouts ... regardless of where he was seeded coming out of pool bouts.

Unless its a qualifier and you're kept out of the running b.c. you got two As and a B that all have national points...

[larkmaj]

Condensed numbers for quick comparison:
2006........2005
11 A06.....13 A05
10 A05.....13 <A05
11 B06.....15 B05
9 B05.......14 <B05
6 C06.......7 C05
2 C05.......3 <C05
4 D..........12 D
1 E...........4 E
2 U...........10 U

There were actually less fencers of each rating that attended the tournament according to these numbers. Though the majority of the dropoff was in the lower ratings the higher ones saw dropoff as well. I would expect this trend as the lower rated fencers would see little point in attending a probably expensive tournament with lots of high level fencers with little chance of success and even getting meaningful bouts out of it. At least to me it wouldn't be worth paying all that money to lose most of my pool bouts and losing in the first round of DEs. Rating inflation does not appear to be the issue here in my opinion. It seems to be more of a snowball effect with less and less lower level fencers attending making the tournament more and more difficult for those that do.

[counterattack]

Rating inflation does not appear to be the issue here in my opinion. It seems to be more of a snowball effect with less and less lower level fencers attending making the tournament more and more difficult for those that do.

Well, that is why I mentioned the total number of pools, but I can see your point. I actually looked later on at some of the 1998-2004 results and realized that perhaps the real anomaly was in my first few years on the circuit (2002-2004) where the numbers were a little out of balance with the 1998-2001 and 2005-2006 numbers.

All of this, though doesn't really answer the question of "at what point would the ratings system be considered broken"? If its purpose is to get roughly accurate initial seeding, then say you had a 60 person tourney with 15 A06, 15 A05, 15 B06, 15 B05. The initial seeding could vary anywhere from "pretty accurate" to "very unbalanced". And anyone who says initial seeding doesn't count for anything is either WAY better than most people or has some very strange fencing experience. Every fencer wants a reasonable (or weak) pool. I have never heard anyone say "I sure hope my pool strong! It really gets my blood pumping."

I certainly don't think ratings are broken yet, but is there a reason to think they can't become broken for larger events (like sectionals or NACs) if there get to be a preponderance of higher ratings?

[PeterGustafsson]

Hi!


There are mathematics covering the topic on how to get as even pools as possible given a set of fencers with known ratings (or other ability measures), but it is not an easy topic.

One way to alleviate your particular problem is to flush the really good guys out of the system first, and then have more balanced pools.

You could have a system where the competition starts with a straight DE - with no ranking - consisting of 5-point bouts, and run the DE until there are 8 survivors left. Those are then directly qualified to a later stage of the competition. This stage would go fast, and only the reasonably good would make it to the top-8. Based on their performance in the starting DE and their ratings, all the others - 100% - are placed in pools. Their pool results then rank them for an ordinary DE with 15-point matches, which is fought until the last 8. Those 8 then go into a 16-man DE of 15 point matches against those that made through the first DE.

Another partial solution would be to start a regional points system, in order to break ties among those who have no national points.

Another solution would be to have more levels in the A->E,U national rating system. I have outlined the algorithm of a simple such system in a post in one of the many previous rating/ranking threads.

Yet another solution would be to institute local ratings based on the A->E,U national rating system, but make good regional competitions an always A4 event. Fencers would then have both a national and a regional letter, and more ties could be broken.


Have a nice time!

Peter Gustafsson

[sabreur]

Part of the problem is that the chum don't want to come out and swim with the sharks, because it usually isn't much fun.

If you're a lower ranking fencer, unless you are ambitious, masochistic or way better than your ranking shows, you are probably going to be somewhat loth to enter a tournament that will have over 40 As and Bs.

[Poulet]

Every fencer wants a reasonable (or weak) pool. I have never heard anyone say "I sure hope my pool strong! It really gets my blood pumping."

In local events, once one already has an A or whatever the rating potentially being given away that day, many would prefer a harder pool. It means more experience.

[Redblade]

Unless its a qualifier and you're kept out of the running b.c. you got two As and a B that all have national points...

You're applying the assumption to the wrong circumstances. Most of the complaining about seeding has been due to local events -- in this case, too, the issue is a sectional championship. And at those events, I have yet to hear of pool results determining qualifiers.

[Craig]

In local events, once one already has an A or whatever the rating potentially being given away that day, many would prefer a harder pool. It means more experience.

True. Easy pools don't prepare you for NACs. Especially those pools on day 2.

[academe]

In local events, once one already has an A or whatever the rating potentially being given away that day, many would prefer a harder pool. It means more experience.

There is also a better chance for upset in a pool than a DE. I've seen kids beat A's with National Points and this provides confidence that winning an easy set of pool bouts does not. In my mind, you've got to fence the best in order to beat the best to go on to be the best.

As competitors we should embrace the difficult, not the easy path.

[fencerontheline]

True. Easy pools don't prepare you for NACs. Especially those pools on day 2.

Day 2 was over a long long time ago Craig...

[oso97]

Another partial solution would be to start a regional points system, in order to break ties among those who have no national points.

And this is exactly why we have started using such a system in the Southwest at our SSCC events. Even though Mr. Epee loves to bash it, its worked very very well for helping to seed our high level competitions. (We also take into consideration national points as well).

[teacup]

[QUOTE=counterattack]
A05 = 1 with NPs (plus one if Pearce does, but USFencing site is down so I can't check)

The points lists are on the international website:
http://www.international.usfencing.org/

[oiuyt]

And this is exactly why we have started using such a system in the Southwest at our SSCC events. Even though Mr. Epee loves to bash it, its worked very very well for helping to seed our high level competitions. (We also take into consideration national points as well).

What MrE is pointing out is that you have a seeding mechanism that contravenes the written national policies (whether or not doing so is allowed is a separate discussion). If the national policies were rewritten to reflect PeterG's suggestion this would not be an issue (I also don't think it would be particularly helpful, but that's yet another debate).

And at those events, I have yet to hear of pool results determining qualifiers.

They ABSOLUTELY do!

Say you have a sectional championships with upper-40's entries (or any number higher than that). You go out in the bracket where the cut-line occurs (in this case the L16). Your first-round pool results is what determines whether or not you've just qualified. The kid who got lucky by having an easy pool and winning that bout-he-really-shouldn't-have had an easy path to the 16, finally hit a decent fencer and bounced 3-15. Of course his seed means he just placed 10th and he qualified. The regular decent-level fencer who drew the pool from h-e-double-hockeysticks (darn U from overseas who hasn't been competing AND the badly under-rated C in addition to the anticipated strength) scored a well-deserved upset in the L32, also bounced in the L16 (losing 13-15 to one of the top regional fencers), finished 14th, and failed to qualify.*

-B
*note: the situation is a hypothetical, not an actual occurance