Six 40-year-olds and One Gets an "E"!

[jesup]

The best way to deal with the huge number of minor to major problems with the A-E rating system is to chuck it.

All rating systems have issues. However, a better overall solution (modulo the work required) is to drop A-E ratings, and use a system more like what chess uses.

Chess assigns points based on if you win or lose, and the relative ratings of the two people.

If for example Johnny and Billy and their friends are all 1200's and fence each other, then winners will gain the same number of points the losers lose. Say Billy and Johnny fence, and Billy wins. He goes to say 1220, Johnny goes to 1180.

If Billy then fences Joe Serious, age 24, rating 2000 and loses, Joe gains say 1 point, and Billy loses 1 (since this is no surprise). If Billy wins because he's the next Aldo Nadi, he might gain 100 pts and Joe loses 100. But that's not likely to happen.


In fencing, it might make more sense to base it on wins/loses/pts versus the average rating of the oponnents, to reduce the calculations/paperwork required. Statistically, that's pretty close. You may want to list DE's individually while grouping pools.

To deal with the "new player/fencer" you can give them a prospective rating until they've had a few results. Or just put them in at say 1000, and if it's too high they'll give away points, and if too low they'll gain.

Someone without enough results would have their rating move but not affect their opponent's (they'd have a "provisional" rating). That helps get them "in" at the right level, even if they were a top college fencer who hasn't fenced in 20 years. (They'd rapidly get to a fairly good rating without stealing points from others, while newbie beginners will probably have their ratings stick near the bottom until their rating is no longer provisional.)

Note that this is basically a zero-sum game except for people leaving and entering. Also note that it depends on a certain amount of "mixing" between groups to keep the ratings from drifting out of sync.

Also note that this gives you a FAR better way to rank and seed fencers, and removes most any reason for a fencer to give up a single bout (or perhaps even point) to a teammate/etc so the teammate can get a better seeding.

Also note that you don't really need major tweaking to handle different ages - the winner of a Y10 will gain points, sure - but if they enter an open local tournament they'll lose them. Mis-ratings (say from a kid who has competed in a dozen Y10 and Y12's, and is now entering his first open) correct themselves rapidly. He may be mis-seeded for one or two tournaments, and then all of his extra points will have been redistributed to the better fencers in the open (regardless of age).

Downside: confusion. Extra paperwork -- though not a huge amount if it were to be all enterable on a webpage that does all the calculations for you, and which could suck the data out of the various programs for managing tournaments. People who like to coast on an old rating would be upset.

[Inquartata]

Yes, this has been talked to death here from time to time. Personally I feel that as long as the classification system works adequately ( if not superbly ) for seeding, which it does, there is no crying need to discard it in favor of a more complicated one. We are not all comfortable doing calculations and fiddling about with numbers.

[jesup]

I'm sure it has been discussed (in fact I remember this going around about 8 years ago). Accurate seeding makes more difference in chess, since a tournament has many more people generally, and many fewer matches - typically 4 to 6.

However, it does address the issues that apparently are causing a lot of consternation, and apparently affecting the outcome of local tournaments due to pool seedings.

As for the numbers issue: combine a web form, some simple computer calculators, and some tables, and little or no hand calculations. No one hand-calculates chess ratings. And updates don't become official until the next set of ratings come out (once a month or so). The tournament official just has to submit the results after each event.

I realize this will probably not be done - but it would give a much better way to compare fencers, locally, regionally, or nationally. On the other hand, outside of nationals, this only matters for seeding. But that's what people are complaining about.

[veeco]

I am not going to address the shortcomings of ELO type classifications, as I have done so in a lengthy discussion before. Feel free to look it up if you will. Some of them you pointed them yourself originally.

Originally posted by jesup

As for the numbers issue: combine a web form, some simple computer calculators, and some tables, and little or no hand calculations. No one hand-calculates chess ratings. And updates don't become official until the next set of ratings come out (once a month or so). The tournament official just has to submit the results after each event.

As I am working on something that is not dedicated to ratings, but that tries to solve some of the similar issues than such a system would have to, let me point out some of the difficulties in doing such a computer system. I am not saying this can't be done, just that in order to do this effectively, it requires more than just a simple web form and a couple of routines. Combine this difficulty with the fact that it would not necessarily address some of the issues than people are having with the current classification system, and it seems like the return of investment of doing this might not be so great.

1- Entering tournament results supposes that you have a common format to return these results. That is not the case so far. Each tournament software currently uses their own format. Engarde has one way or reporting results, XSeed another, Point Control yet another, and Athos again another. Most of these tournament software don't even include the results of individual bouts in their final results, just the placings of all the fencers. This means that all the software used now would have to find a common ground to give those results, or that the person writing such a classification tool would have to create a set of filters to load all the results in native format. And even then, the tournament software makers would have to include all results from all bouts in their tournament results to begin with.


2- A lot of people still run local tournaments by hand. This places the burden on the organizers to type all the results after the event is over, in a format known by the software. Because it is a manual process, there is a potential for human error, which would mean that someone would have to write some bullet-proof code to ensure fairness. What if that code doesn't handle that human error, and you get credited with a loss when you should have been credited with a win? In the current system this is still possible, however there is someone at the USFA office and someone at the tournament who looks over the results. So this makes 2 pairs of eyes to check for the result, instead of one pair and a computer program. Also since the current system is quite simple, if you don't get your classification you know there is something wrong, whereas unless you do all the calculations yourself in the ELO case you don't know if the calculations were correct.
Furthermore, forcing the organizers to retype all the results by hand might deter people to organize tournaments if they don't have a computer to run it, resulting in less tournaments to compete in in the end. I for one don't want that to happen.

Anyway, as you can see, it's not a clear win-win situation to move to that type of classification system. So personally, I am happy with what we have now. Sure as it's been proposed it could be further enhanced to provide a category next to the classification. For example an A02 would have to be associated to the event it was won in. If it was an open it would become an A02(o), if a women's it would become an A02(w), and so forth for all youth, cadet categories. One could then devise some guidelines for open events organizers to map from one category to the other in order to seed the open events properly, so for example an A02(y14) could be seen as a B02(o) or something. Pretty soon the people competing in opens would get their "open" ranking and we would know what they are worth in that category. Or it could be left to the discretion of the organizer to seed with those classification, with the added knowledge of the category they would be able to do a better job seeding.

[jesup]

Thanks for the well-reasoned reply.

Many of your concerns are valid, though the software people use would almost certainly be updated by the time any such system went into effect (and others would write new software). And to ease hand-done local tournaments, I'm sure people (or even I) would create a Palm/PocketPC app to record the information (note: for what I'm proposing, you don't need to run the tourney on a computer, you just need the results).

Also, you can probably make a system where all you need to enter is the list of entrants and their results, either numerical place (including ties/undetermined) or by-pool results plus DE results if any. Your rating would go up or down relative to the strength of the total competition (much as you can only earn a good rating today in a strong competition).


Your system would be easier to implement, though still confusing in ways. And it doesn't solve the problem of ratings lasting 4 years, or the problem of valid ratings being tough to get without a large pool of fencers, or local ratings inflation if a club/coach holds lots of U tournaments to get the local area seeded with tons of E's, etc.

Not that an alternative points-based rating system would solve all of those, or that it wouldn't have it's own problems.


Fundamentally, I simply really like the idea of making each bout or each tournament affect one's ranking/rating. Effectively that's done in the National events - you get points which are cumulative, but it resets every year, and while it serves the purpose it's meant for, it doesn't necessarily tell you who is better because it's not really meant to.

FYI, I've competed in a lot of events, fenced for 23 years on and off, fenced competitively regularly for around 3 to 4 years. I've come in second in events with B's, C's/D's and various E's. There are D's out there I can beat regularly, and B's I can hold my own with (some much more easily than others, given the rating system). I've beaten (good) B's in large tournaments. But I've never gotten an E rating.

I'm starting to fence competitively again this season after not doing so since around 98-99 or so, so perhaps I'll finally earn a rating. (I may try to go to some larger regional/sectional tournaments.) But I'm not holding my breath. :-)

[oiuyt]

The ELO system has a significant overhead. The USCF charges tournament organizers for every game played. When results are submitted electronically in the proper format that charge gets as low as something like $0.50 (from memory, I'd have to go wade through the USCF site to find the exact amount currently) per player per game. Now when you're playing a 7 round tournament with a $100 entry fee the $3.50 really isn't all that much. Tack that onto a local fencing event with a $10 entry fee and it's much more noticible. Just because it's all done by computer doesn't mean the cost is low, just that it's lower. The improvement in the system ISN'T worth this extra cost, OR the added costs because people don't understand the system (I'm a math geek, when I played competitive chess I also cranked the numbers to see how each tournament affected my rating. In my experience most people not only aren't willing to do this (not that it's required), but dislike no knowing how the system works and feel that it's too complicated to learn).

Jesup- before we go over too much old ground can you look up the previous threads on this topic? We haven't had one in a while, so perhaps it's time for a new one or the resurrection of an old one, but it'll help the discussion if you've seen what has gone before. Thanks.

Many of your concerns are valid, though the software people use would almost certainly be updated by the time any such system went into effect (and others would write new software).

Heh. I bet you even believe that. :)

The lack of standardized means of conveying every bout result is presumably why in the system that was proposed to the USFA about 6-7 years ago that ratings were only affected by final placement (and the ratings of the other people in the tournament) rather than by actual bouts fenced.

Jesup- Welcome to the board.

-B :)

[veeco]

http://www.fencing101.com//vb/showth...g&pagenumber=3 The previous thread...

I linked to the 3rd page because this is where the chess rating discussion starts. Feel free to read from the beginning to see the development of the story ;-)