When Did US Fencing Begin Using Ratings?

Discussion in 'Fencing Discussion' started by sdubinsky, Mar 2, 2015.

  1. Steve Khinoy

    Steve Khinoy DE Bracket

    Joined:
    Dec 30, 2005
    Messages:
    621
    Likes Received:
    68
    Some posters on this discussion are much more sophisticated about rating systems than I am. But I'd like to make two points.
    First of all, ratings are not just about seeding -- in the minds of most fencers C and under, who are the majority of our fencers. Ratings motivate them. We should not pretend that this is not the case.
    Second, it would be relatively easy to evaluate any proposed rating system (existing, ELO-like, ELO-with-a-floor, or whatever) using the very extensive amount of available data--AskFred and USFA reporting. (And for that matter, the FIE system.) Does the present system predict first-found results in a NAC, a nationals? Does it predict first-found results in a local open? How about DE's? Would an ELO (or any other) system be more predictive?
    It's like asking whether the seedings for the NCAA DivI Basketball Championship are more or less accurate than the Vegas odds, or the RPI, or some other index, or the indices combined. Could we be doing better than we are doing?
    The data is (or, Inq, are) there.
     
  2. DangerMouse

    DangerMouse Podium

    Joined:
    Mar 13, 2003
    Messages:
    3,184
    Likes Received:
    196
    The issue with this is that we are not all agreed that predicting outcomes is a useful measure of a rating system.
     
  3. Inquartata

    Inquartata Podium

    Joined:
    Jul 12, 2001
    Messages:
    36,189
    Likes Received:
    1,183
    If only that were the only issue.
     
  4. NGV

    NGV Rookie

    Joined:
    Feb 6, 2000
    Messages:
    317
    Likes Received:
    25
    By definition, any rating system that effectively seeds according to ability MUST also (at least to some extent) predict outcomes.

    The logic is pretty straightforward. Ability predicts outcomes; so, if the seeding reflects ability, then seeding should also predict outcomes.
     
  5. NGV

    NGV Rookie

    Joined:
    Feb 6, 2000
    Messages:
    317
    Likes Received:
    25
    I agree -and I don't think it only applies to C and under only. Do B fencers really not care about getting the A (or only care because it will lead to easier pools?).

    It doesn't matter how many times people say that ratings aren't a measure of accomplishment or advancement. They will commonly be viewed as such, because for the great majority of fencers, there isn't much else that serves that purpose.
     
  6. DangerMouse

    DangerMouse Podium

    Joined:
    Mar 13, 2003
    Messages:
    3,184
    Likes Received:
    196
    Of course a rating/seeding system at least somewhat predicts outcomes. The point is that the accuracy of outcome prediction is not necessarily a good measure of the quality of a rating/seeding system.
     
  7. mfp

    mfp Podium

    Joined:
    May 10, 2002
    Messages:
    1,828
    Likes Received:
    221
    Perhaps a way to move forward is to put aside ratings/rankings for a bit and see if there's any agreement on ways to characterize and measure competitions.

    Say for example, what does the USFA want to get out of Div 1 National Championship events? Or Div 1 NACs?
     
  8. jkormann

    jkormann Podium

    Joined:
    Nov 8, 2002
    Messages:
    2,733
    Likes Received:
    148
    Perhaps the question isn't about applying the ratings to the events, but how the ratings are awarded.

    To get an E, you need to be first of at least six others who are not rated.

    To get a D, picking on the D1 group, you need to be first of 15 and two of them in the top eight need to be rated E. Why should there be 15 fencers in the tournament. If I have a group of eight fencers and two of them are E-rated, why should it matter if I had another seven fencers in the tournament. As it stands, if I had only eight fencers in the scenario, the highest awarded rating would be an E.
     
  9. DavidA

    DavidA Made the Cut

    Joined:
    Feb 3, 2015
    Messages:
    233
    Likes Received:
    4
    aha! Quantifying the three discordant Muses of luck, talent, and deftness of blade in the face of competition - all who demand different sacrifices - into a prairie school-house grading system.

    What could go wrong..

    The only way to get a ranking should be to fence the necromanced shade of Aldo Nadi on the ethereal planes and then be bequeathed your ranking through the number of scratches on your weapon hand that you have after you wake up.
     
  10. fdad

    fdad Podium

    Joined:
    Jul 9, 2008
    Messages:
    4,129
    Likes Received:
    214
    To get a D, you need 4 Es in the event, 2 of which have to finish in the top 8.
     
  11. mfp

    mfp Podium

    Joined:
    May 10, 2002
    Messages:
    1,828
    Likes Received:
    221
    It's a minimum of 5 others, not 6, and it doesn't matter if the others have classifications or not. A "U" who wins a 6 person sanctioned competition placing ahead of 5 As gets an E.
     
  12. jkormann

    jkormann Podium

    Joined:
    Nov 8, 2002
    Messages:
    2,733
    Likes Received:
    148
    You and fdad are correct on the numbers, I was typing too fast. My point is still there, perhaps we are awarding the ratings wrong.

    In your example, a competition with 5 A-ratings and 1 U-rating would only allow a single E to be awarded. Logically the event should go to one of the A-rated fencers because they are seeded higher. If the U-rated fencer is a national champion or OLY fencer from another country, they should be better than E. But that's what is awarded.

    Continuing the example, if the competition was an A1 to start (15 fencers) and the final 8 contained 5 A-ratings and the 1 mystery U-rating, then the U would be awarded the A. Why? Because the beat the As? He would have done that in the original 6-person competition where he was awarded an E.
     
  13. DavidA

    DavidA Made the Cut

    Joined:
    Feb 3, 2015
    Messages:
    233
    Likes Received:
    4
    What if instead of the decay system you could drop rating by being beaten too many times by ranks lower than yours.

    Like if you were in school and got an A on the big first exam - technically youre an A student but that doesnt mean you can just coast for the next two years - subsequent opponents are like subsequent exams - with b eing beaten by lower ranked opponents lowers your Rank "average" I guess by the corresponding amount - having an and getting beaten by the appropriate amount of b's lowers your average - being beaten by a e lowers your average at a correspondingly faster rate - that way the problem of in-bred and Hail Mary A's (a term I just came up with right now) resolves itself through the crucible of a variety of other opponents - and a more flexible rating system allows for accuracy to be pinpointed by finer degrees. That way nobody stumbles into a NAC - has the luckiest string of seizures ever seen on the strip - and gets an A that messes up seeding for the next two years. High water mark is good only if the ratings are more to make kids feel good about moving up - and compels them to work harder to maintain the new "level" that they are at - otherwise - with todays ability to instantly correlate data in the most nuanced of ways - there is no excuse to not have a system at least as fluid and dynamic as a public school's grading system
     
  14. DangerMouse

    DangerMouse Podium

    Joined:
    Mar 13, 2003
    Messages:
    3,184
    Likes Received:
    196
    This disincentivises higher rated fencers from fencing local tournaments or anything that is outside their primary competition goals for the season.

    I still don't understand what is actually wrong with the current system. All I have heard is that it either doesn't predict the outcome of tournaments perfectly or whining that some people have classifications that other people don't think they deserve.
     
  15. DavidA

    DavidA Made the Cut

    Joined:
    Feb 3, 2015
    Messages:
    233
    Likes Received:
    4
    yeah i guess you're right

    Wanna fence?
     
  16. jkormann

    jkormann Podium

    Joined:
    Nov 8, 2002
    Messages:
    2,733
    Likes Received:
    148
    The main gripe with the current system is it doesn't reward the strongest fencer in the group based on who you beat. It requires too many variables (tournament size, minimum number of a certain rating) before awarding a rating. You cannot predict the outcome because there will always be spoilers or people having a good (or bad) day. If a fencer wins against six A-rated fencers, they shouldn't get an E. That's an example of what I'm poking at.

    Sure!
     
  17. K O'N

    K O'N Podium

    Joined:
    Aug 14, 2006
    Messages:
    3,593
    Likes Received:
    458
    There are as many "main gripes" about the current rating system as there are gripers. There are too many A's, it's too hard to get an A, too hard to get a C, too easy to get a B and rate yourself out of Div II, fencers don't get punished for bad days, fencers don't get rewarded and punished fro each bout, the current system doesn't have an equation associated with it so it feels less precise than an Elo system, on and on and on.

    Yours is sort of based in "fairness", I would say. If someone beats six A's, they deserve an A. Ok. But if you reward everyone who "deserves" an A or a B or whatever on a given day you're going to have a lot of ratings inflation. Lots of people have a good day every once in a while. You would either have to move the requirements way, way up or you would have to also drop people in ratings who "deserve" to be dropped. And when you do that, when you start to drop ratings for bad performances, you end up with the effects DM was describing; why would an A fence locally against U's and E's and D's, if the only possible result is a loss in rating? It would be perfectly reasonable for an A to say, well, I'm an A now, I fence nationally. That would be a perfectly reasonable thing to do. And poof! We lose a lot of A's fencing locally.

    Forget "fairness" on a short timeline, the rating system doesn't need to reward you every time you have a good day. Think of it over the course of a year. If I'm fencing at an A level every time I go out, maybe this time or next time I go fence I don't get my A, but over the course of a year I'll get it. And in the meantime there's no dis-incentive to fence, there's no reason to stay home, and we don't have wild ratings inflation.

    I mean, if you want US Fencing to tweak the requirements for an A, ok. I think it should not in general get a lot easier to earn a rating, on average, but if you want to tweak the chart I wouldn't mind, I guess.

    Personally, I'd like them to tweak the cutoff levels for Div 1A/Div II/Div III events. Right now there's no reason for a 28 year old B to ever go to a NAC. But a 28 year old B is a very good fencer who probably has years and years of dedicated practice behind him. He's put a lot of time and effort into fencing, and US Fencing's reply to him is to bugger off, there's no Div N event for him. That's a problem, IMO. Not the ratings thing, that's for seeding, seeding seems fine to me so the rating system seems fine to me.
     
  18. Ancientepee

    Ancientepee DE Bracket

    Joined:
    Jun 12, 2010
    Messages:
    951
    Likes Received:
    181
    If all you've heard is that it "doesn't predict the outcome of tournaments perfectly", then you haven't followed many of the postings. It's been pointed out that NO classification system can possibly do this. If you don't understand what is actually wrong with the current system, then I have a question for you: "What do YOU think the purpose of a classification system is?" I'm curious as to what you think the current classification is doing perfectly.
     
  19. DangerMouse

    DangerMouse Podium

    Joined:
    Mar 13, 2003
    Messages:
    3,184
    Likes Received:
    196
    I have been paying attention and, with the exception of a few people who really don't seem to represent the majority of the gripers, most of what I see is a lack of measurable goals for a new rating system.

    I see 2 main purposes for a classification system. First, to seed tournaments. Second to provide a criteria for deciding eligibility for different tournament levels.

    I think the current system does a fine job of seeding. The only consistent issues I see with seeding have to do with foreign fencers, boundary cases between age groups, and, at the very top, the discrepancies in fencers with the means to chase international points vs those that just do well in national tournaments. None of these issues are really direct results of the classification system, but are instead related to other policy issues, such as how to weight international competition as compared to domestic. Or how to handle unknown international competitors. The only one of these that I think could use real improvement is handling competitors as they age into new categories. The other issues are outliers that can easily be handled with small policy changes. I don't know how to handle the age group issue, but I have yet to see a compelling argument or even a decent suggestion of how a new rating system would fix the issue. I think people moving from one competitive category to a new one in which they have not previously competed will always result in some short term seeding issues that will settle out over the long run. Maybe making ratings from age groups other than senior/open decay faster would help, but it hardly requires an overhaul of the entire system.

    For the second issue of deciding eligibility for tournaments is entirely a policy issue unrelated to the classification system. I agree with K O'N's assessment that being a B in your late 20s or 30s puts you in a competitive wasteland. Changing the classification system would not have the slightest impact on this issue. Instead we need to look at changing the arbitrary cut offs for eligibility. I would propose different cut offs based on weapon and the apparent unwillingness of the NO to add more events. Div I ME might be better limited to A and maybe B fencers with B fencers also being allowed in Div II. For WS, Div I should probably include all the way down to C fencers and the current structure is not as much of an issue. None of this requires a new classification system as the current one provides just as useful of arbitrary boundaries as any other system.

    The other role for a classification system that I see a lot of people put forth is to provide goals for competitors to use to measure progress. I don't consider this a primary objective for classification systems as I see this as part of the coach's job. However, the current system does this just fine. I also think most other systems would do this just fine as long as competitors knew their current classification and had a good idea and immediate feedback on how their competition results would affect this. The current system is great for this. People know that going to a certain tournament and placing in a certain round will earn them the next level of classification. I would argue that any proposed change to this system needs to maintain this feature. If calculating classifications happens after the tournament at the NO with resultant changes posted a few days later, this loses the excitement and momentum generated by leveling up.

    And FWIW, I don't think that the current system does any of these things perfectly, but I don't see perfection as the goal. I also don't see any system that does anything perfectly. I think the onus is on those proposing significant changes to the system to show how the proposed system would meet the primary goals better than the current system. So far, I haven't even seen a plan for how to appropriately measure those goals. Without that plan, then all arguments boil down to how someone FEELS about the system.
     
  20. Ancientepee

    Ancientepee DE Bracket

    Joined:
    Jun 12, 2010
    Messages:
    951
    Likes Received:
    181
    What measurable are you using to determine this?
     

Share This Page