Who would support a change to the USFA classification table?

Discussion in 'Fencing Discussion' started by malediction, Nov 18, 2018.

  1. malediction

    malediction Made the Cut

    Joined:
    Feb 20, 2010
    Messages:
    73
    Likes Received:
    26
    The USFA Classification table: https://askfred.net/Info/eventClass.php

    I think the USFA classification table needs a new type of event. Currently, the "1" and "2" class tournaments, which requires 15 or 25 people (with the exception of an E1), fill a pretty good niche. Tournaments of 15-25 people are not uncommon or hard to find, and the sufficient number of As, Bs , and Cs in many locations means that even relatively small tournaments can be fairly strong.

    However, the jump from 25 person "2" class events to 64 person "3" class events is quite large. Given regular registration fall off (rule of thumb is 10% won't show), you generally need to attract somewhat more than 64 fencers to have a chance at a "3" or "4" class event. If an event registration hovers around 64 people, there is almost always be a massive drop off as people decide it isn't worth it to show up to a 55 person event. A 55 person event, of course, requires more pools and an additional round of fencing over a 32 person or smaller event.

    For example, an A4 with 65 preregistrations is almost certainly drop to an A2, but even worse, the number of attendees often drops well more than 10% as fencers decide it isn't worth it to go to a large, long event.

    I think the USFA should adopt a new group of events that sits between the current "2" class events and "3" class events. For example:

    Group C2.5: Requires 50 fencers, including at least 4 Cs and 8 Ds, awards 2 Cs, 4 Ds, 6 Es.

    Or another example:

    Group A3.5: Requires 50 fencers, including at least 8 As and 8 Bs, rewards given as an A3.

    These are hypothetical examples, of course. But having an intermediate event between "2" and "3" class events would help draw fencers to larger events that don't make the very large 64 person mark. The reduction in tournament registration drop off alone would be worth it.
     
  2. jdude97

    jdude97 Podium

    Joined:
    Apr 10, 2013
    Messages:
    1,921
    Likes Received:
    228
    Preaching to the choir I think. I've proposed over the years of a number of alternatives ranging from slight updates to radical redesign. I have proposed a system where there are no buckets/cliffs. It's dumb that a 63 person event of all A's awards fewer A's than a 64 person event with 32 U's. Similarly dumb that a 64 person event with 32 U's and a 300 person div1 NAC event with 150 A's and 0 U's both only award 8 A's and 24 ratings. As such my proposed system involves assigning each rating a numerical value and then a fencer's earned rating depends on the sum of the ratings of competitors finishing lower in the final standings. I will have to go dig up my precise proposal since I haven't considered it in awhile but the gist is something along the lines of: if an E is worth 3 and a U is worth 1, you have to beat 2*3 = 6 worth of points which can be done by finishing above 6 U's or 3 E's or 4 U's and 1 E, etc (don't get stuck on these precise numbers...just a quick invented number)
     
    Privateer likes this.
  3. K O'N

    K O'N Podium

    Joined:
    Aug 14, 2006
    Messages:
    3,798
    Likes Received:
    602
    I agree. The general evolved nature of the letter ratings as a high water mark seeding structure is fine, better in many ways than reformers give it credit for, but the current chart is pretty imperfect in its details. I agree with jdude97 that the cliffs and buckets could be eliminated, but we'd have to look at a bunch of askfred data and make sure that the new structure would not generate a lot more or a lot fewer ratings than the current structure. Still, good project. It's the kind of thing US Fencing would be very interested in if US Fencing gave a crap about sub-elite fencing and its growth in the US. Sadly they are not interested, because they do not give a crap. But still. Good idea.
     
    Privateer, SenorEquis and jdude97 like this.
  4. jkormann

    jkormann Podium

    Joined:
    Nov 8, 2002
    Messages:
    3,029
    Likes Received:
    206
    Best do that while we still have AskFred. Once sanctioned tournaments are only on the USFencing web site, I doubt we'll ever get access.
     
  5. mfp

    mfp Podium

    Joined:
    May 10, 2002
    Messages:
    2,089
    Likes Received:
    368
    Disliking some aspects of the current scheme and wanting to address them is fine, but replacing it with something worse isn't. Your proposed system causes far more problems than it purports to solve.
     
  6. Ancientepee

    Ancientepee Podium

    Joined:
    Jun 12, 2010
    Messages:
    1,170
    Likes Received:
    353
    Using letter classifications is outdated. Some 20 years ago I wrote a long article in American Fencing proposing that we adopt an Elo type numeric system like chess has been using successfully for about 70 years. It's success over that time period has caused it to be adopted by numerous other organizations. I believe that there's currently a web site for fencing which is trying to get such a system adopted and they're running current competition results through the computations to show how it would work. Don't waste you time trying to "fix" a letter-based system which is inherently too crude and inaccurate to give a meaningful means of ranking fencers.
     
  7. Inquartata

    Inquartata Podium

    Joined:
    Jul 12, 2001
    Messages:
    36,867
    Likes Received:
    1,339
    Here we go again.
     
    jjefferies, K O'N and Strytllr like this.
  8. Strytllr

    Strytllr DE Bracket

    Joined:
    Nov 6, 2003
    Messages:
    676
    Likes Received:
    118
    What website is that?
     
  9. jdude97

    jdude97 Podium

    Joined:
    Apr 10, 2013
    Messages:
    1,921
    Likes Received:
    228
    Criticizing my proposal without articulating any of that criticism is the epitome of unhelpful.
     
  10. jdude97

    jdude97 Podium

    Joined:
    Apr 10, 2013
    Messages:
    1,921
    Likes Received:
    228
    14meters.com I believe is to what ancientepee refers.
     
  11. jkormann

    jkormann Podium

    Joined:
    Nov 8, 2002
    Messages:
    3,029
    Likes Received:
    206
  12. Ancientepee

    Ancientepee Podium

    Joined:
    Jun 12, 2010
    Messages:
    1,170
    Likes Received:
    353
    For those interested in the history of the classification systems that have been used by the USFA since it was founded, attached is a brief history that I researched about two years ago.
     

    Attached Files:

    K O'N and Strytllr like this.
  13. jdude97

    jdude97 Podium

    Joined:
    Apr 10, 2013
    Messages:
    1,921
    Likes Received:
    228
    Idk which one AE intended to refer to, but 14meters does use "an emulation of the World Chess Federation (FIDE) Elo system (including K-factor considerations)" per their website header.
     
  14. malediction

    malediction Made the Cut

    Joined:
    Feb 20, 2010
    Messages:
    73
    Likes Received:
    26
    I just read Ancientepee's very interesting history of the US Fencing rating system. But the point of my post isn't to criticize the use of letter classifications instead of statistical models.

    Rather, I am concerned with tournament drop off. I was at the Champagne Challenge yesterday in Washington, DC. The event had hovered around 64 preregistrants for most of the past two weeks, before dropping down to 59. Actual attendance was 40. This reduction (about 1/3), I theorize, is due to uncertainty that the event will maintain its rating. I don't know a single fencer who prefers low level, small events to large, high level events. But I do know a number of fencers who won't bother competing unless an event gives them the highest chance of success.

    Ancientepee suggests, in the history he linked, that US swimmers judge success on personal best times, and that a statistical model (ELO is suggested) would better allow fencers to judge their progress. So too, I think, would additional tournament classes which offer more rewards and opportunities to advance and maintain ratings. More event classifications would fill gaps and open up new opportunities to progress. Besides which, I know many fencers who compete for the prospect of earning ratings, which testify to their skill and success.

    As for statistical models, I have two gripes with them. ELO ratings can fluctuate in a depressing manner, and a numerical score is more abstract than a letter rating. I think the virtues of the ease of understanding what a letter means and how to earn it are self-evident. The problem with ELO - at least 14meters implementation of ELO, are less evident.

    Sometimes, I go to 14meters to check out where I stand. One thing that strikes me each time is that, if I lose a bout at all, the loss is usually weighted far more heavily than preceding victories. This creates a real time fluctuating record where I can see preceding point gains wiped out. This fluctuating record is depressing because it always ends up on a negative note unless you win everything. Secondly, it's aggravating to see 5 points from a pool victory and 9 points from a DE victory mostly wiped out by one 12 point DE loss to someone who, statistically and by the existing rating model, should have won the bout anyway. A better presentation would be to show net change based on overall performance at the tournament. Hide the math, in other words.

    Secondly, many fencers get "worse" before they get "better." A novice fencer, just starting out, is likely to suffer multiple losses at the beginning of their career. ELO reflects this with a declining rating. You go from a 1000 to something less. This is true whether you show ratings based on net gain/loss or per bout. I think this could be discouraging, especially to fencers in strong regions where it takes time to get good enough to really challenge the existing good fencers. The letter rating system has the virtue of being slowly regressive. Yes, you might do poorly or lose at an event, but at least it doesn't hurt you for a while.
     
    K O'N likes this.
  15. Strytllr

    Strytllr DE Bracket

    Joined:
    Nov 6, 2003
    Messages:
    676
    Likes Received:
    118
    I love these nuggets of history that you've accumulated. Do you have a central repository for all of these stories? I would love to spend a week or two reading through your memories. :)
     
    jdude97 likes this.
  16. Privateer

    Privateer Podium

    Joined:
    Feb 8, 2009
    Messages:
    1,119
    Likes Received:
    141
    honest question - what do other countries use for ranking and seeding purposes?
     
  17. Ancientepee

    Ancientepee Podium

    Joined:
    Jun 12, 2010
    Messages:
    1,170
    Likes Received:
    353
    The Elo system is based on the probability of beating an opponent. So if an opponent's rating indicates that you have a 60% probability of beating that opponent and you fence that opponent 10 times, you should expect 6 victories and 4 losses. This means that 6 times your rating will increase and 4 times it will decrease. Having a higher rating does not guarantee that you will beat that opponent every single time that you fence him/her. As a result, your rating will fluctuate within a certain range even if your fencing prowess remains the same. So you should not consider suicide if your rating decreases by a small number of points nor should you post in Facebook if it increases by a small number of points.

    The Elo formulas contain a number that can be considered an "uncertainty factor". Different groups can use a different number in their formula. I believe that at one time chess used 16 in their formula. This means that a chess player's rating can fluctuate to be either 16 points higher or lower than the rating that they currently have without indicating any change in the player's actual ability. So expressing concern that beating or losing to a particular opponent in a single particular bout might result in an increase or decrease of 5 or 10 points is meaningless if the group's uncertainty factor is 10 or less. You should expect that change to be "corrected" over the course of multiple bouts if your rating is accurate and your fencing abilities have not changed. Much greater changes in both point values and relative ranking can be seen in our team selection point standings and I am not aware of massive resignations from USFA membership caused by this.

    As to concerns that there are some fencers who might become obsessed with minor changes and place too much emphasis on them, fencing could (and probably should) do the same as chess and assign titles or categories for certain rating ranges. For example, I believe that USCF uses the following:
    Senior Master Over 2400
    National Master Over 2200
    Expert 2000–2199
    Class A 1800–1999
    Class B 1600–1799
    Class C 1400–1599
    Class D 1200–1399
    Class E 1000–1199
    Class F 800–999
    Class G 600–799
    Class H 400–599
    Class I 200–399
    Class J under 200
    So a fencer could say: "OK, my rating has gone done 100 points during the last season but at least I'm still a Class B".

    Keep in mind that an Elo system has been used by chess for almost 70 years, has been used by other organizations for decades, and its use has been increasing. If numeric rating systems are inherently faulty or harmful to organizations that adopt them, where's the proof?
     
  18. bbower

    bbower DE Bracket

    Joined:
    Feb 20, 2003
    Messages:
    519
    Likes Received:
    141
    I can count on zero hands the number of chess tournaments I've seen in the wild.

    If you have a rating system where it can be to your detriment to participate in a competition, there will be hundreds of incidents yearly of fencer A walking into a tournament, seeing that B isn't fencing so there's no point in A competing, which causes C to drop out ....

    Letter ratings just aren't that important. They provide a very rough and imperfect way to rank fencers who aren't on the national point standings. They also provide a way to segment the fencing population into the groups of "Can fence in Div I NAC, Can't fence in Div I NAC". Again, imperfect, but good enough for what it needs to do. Especially with the introduction of regional tournaments and point standings (and apparent destruction of local tournaments), the need for anything better than A,B,C,D,E,U is arguably less today than it was in 2000.
     
    mfp, Inquartata, Peach and 1 other person like this.
  19. keropie

    keropie Podium

    Joined:
    Sep 10, 2005
    Messages:
    1,478
    Likes Received:
    178
    Ultimately, it doesn't bother me. I think it works well enough for seeding most events, and I haven't been chasing ratings in years. But I do understand the lack of granularity bothering people.
     
  20. Ancientepee

    Ancientepee Podium

    Joined:
    Jun 12, 2010
    Messages:
    1,170
    Likes Received:
    353
    With the Elo formulas, you cannot lose points except by losing. If you beat all of your opponents, you will either gain points or (if all of the opponents are phenomenally weaker than you are) neither gain nor lose points. If you fence against ten fencers who are so weak that you have a 90% chance of beating them and you beat 9 but lose to 1, the formula is designed so that the total amount of points you gain from the nine victories will be roughly the same as the points you lose from the one loss.

    If fencer A wants to compete in a tournament only if fencer B is also but doesn't enter because fencer B is not competing, then what difference does it make what rating/classification system is used?

    I would encourage you to read the last sentence in my post #17 and, instead of thinking up imaginary scenarios and proposing what you think will happen in those scenarios, that you try to explain why so many organizations are using an Elo system when it is so detrimental as you claim.
     
    jdude97 likes this.

Share This Page