A modest proposal

[tchwojko]

Wait, this isn't satire??

I think it's about as satirical as the GOP primary has been so far.

[mrbiggs]

We go down in rating with the current system. Don't re-earn your rating in four years, go down. If you want to go down faster make it three years, that would be ok.

:eyeroll: that basically never happens. Ratings inflation is a real thing. Making it three years won't make the number of As in men's epee any less ridiculous.

What if they stay home rather than get punished?

What if they don't? The Elo system isn't some made-up theory, it's been put into practice for many different applications, and this was not a major issue for any of them. You're basing this on nothing.

What problems has rating inflation caused?

The current distribution is decent and I agree it's better than it was 10 years ago. But there are too many As to accurately seed large tournaments, especially in men's epee. This will be a much bigger problem in another 10-20 years unless something changes. Why not start thinking about a system that doesn't have to be reset every 40 years?

Wait, what? All a seeding into pools has to do for top fencers is keep them apart. If you think it's failing on this, go find FRED data to show us, that would be interesting. How many tournaments can you find where the 1 and 2 finishing fencers were in the same pool?

This is what I mean by incredibly low standards. You could seed randomly and the chances of the top 2 ending up in the same pool is quite low. Here, I'll take the most recent A4 men's epee event on askfred (haven't looked at what it is yet). Here's the initial seeding in order of final placement for the top 8:
15, 5, 3, 1, 10, 16, 11, 47

This isn't catastrophic; tournaments still work. But why not strive to do better?

Ok, another testable claim. If for example Bs are hard to get somewhere, that somewhere should dominate Div II on a national level, since they have a bunch of "B" fencers who haven't been able to get Bs. If Cs are hard to get in a region, that region should dominate Div III on a national level. Please demonstrate.

To demonstrate this would require looking at large regional successes at lower-level national tournaments over large areas and large amounts of time and you know I have no interest in doing so. You're also assuming that these fencers don't eventually get ratings before going to national tournaments. I can tell you from personal experience that there's a wide skill variation even within ratings at a Div3 NAC. I can also tell you that newer fencers from New York City go to Connecticut, Long Island, New Jersey, and southern upstate to fence in tournaments where it's easier to get ratings, because they told me so.

This is kind of like the chest protector debate in the other thread in that it's not really worth my time to quantify something that the vast majority of fencers can see around them every day. If you really believe that a WS E from Portland is going to be the same skill as one from South Dakota, that's up to you.

Or 'soft' ratings from other category fencing; Y14, or Cadet, or Vet. Probably a lot of what's driving rating inflation. So what? And, more to the point, what do you think an Elo system would do with category fencing? Here's a hint: It would be worse than this.

Another example of 'the ratings system has this disadvantage but I don't care'. Again, you could seed tournaments randomly and it would work; that's not a reason to ignore better solutions. Also another example of where you come up with a disadvantage of the elo system and conclude that it would be worse. It's true that youth and vet fencers would be proportionately overrated, as in the current system. But with today's system, there's no way to adjust for that. If you get a B, you're a B, period, no matter what. Under a system where you can actually go down in rating, a fencer will be rated relative to the tournaments they fence. So the more a youth fencer does open tournaments, the more their rating will be correct for open tournaments. Sure you can game this system if you try hard enough, but again, extensive testing in other implementations shows that it doesn't become that much of an issue. I'd rather have fencers theoretically being able to game the system for small advantages rather than larger advantages and disadvantages being given out randomly.

And as I said before, do you really want the USFA to spend scarce resources on this?

No.

That doesn't mean the current system is good though. And if a fencer comes up with an easily understood, more accurate, and easily implemented system, I think the USFA should consider it. That hasn't happened in this thread but I still don't think we should be satisfied with the status quo.

[tchwojko]

Seeding, metrics and qualification are three different things. The current system is ok for the first, not so good at the rest.

[mrbiggs]

On second thought I'm overcomplicating.

Basically, most tournaments I go to (maybe 50-80% depending on type, but I'm making that number up) have some pools that are significantly easier or harder than others. If we could bring this percentage (whatever it actually is) down, and simultaneously provide a relatively accurate ratings system for fencers, I think that would be a positive step for the sport. This will never happen if we insist that there's nothing wrong with the current system.

[telkanuru]

I don't think anyone's insisting that there's nothing wrong with the system, but rather that there's not enough wrong with it.

[Jason]

Seeding, metrics and qualification are three different things. The current system is ok for the first, not so good at the rest.

This seems to me what causes so much of the confusion in ratings discussions.
I suspect that adding more granularity to ratings wouldn't really have a significant effect on competition results. We don't just use a single-elimination format, after all; there is a seeding round (pools). The "an E from South Dakota is weaker than an E from Portland" argument is not a seeding argument. It's a "but how will people know that I'm better than he is??!!" argument.

It's seems that the issue is whether or not ratings should be something other than a seeding guide.

[DangerMouse]

On second thought I'm overcomplicating.

Basically, most tournaments I go to (maybe 50-80% depending on type, but I'm making that number up) have some pools that are significantly easier or harder than others. If we could bring this percentage (whatever it actually is) down, and simultaneously provide a relatively accurate ratings system for fencers, I think that would be a positive step for the sport. This will never happen if we insist that there's nothing wrong with the current system.

I can agree with this in principle, but execution is much harder for men's epee than it sounds. The other weapons (and genders) tend to have more consistent results in general, but men's epee results can be highly erratic. I don't see this as a failure of the classification and seeding system, just the nature of the sport.

In your example A4 tournament, the C fencer who made the 8 obviously has either not been fencing tournaments recently, or had the day of his life. No system will fix that issue, especially if he doesn't compete again for a while (how does Elo handle changes from not competing?) .

Similarly, my experience with some pools being significantly harder than others often has to do with people that are better than their classification, but don't go to enough or the right tournaments to re-up. The other times I see pools that skewed that way is when the A fencer(s) in the pool are from a different gender/age group. It used to be women's classifications were reduced by two levels when fencing mixed competition and that a classification in one weapon gave you two steps lower in each of the other weapons. This got scrapped a while ago for a number of reasons.

[qatet]

This got scrapped a while ago for a number of reasons.

Several of which were named Ann Marsh, eh?

[mrbiggs]

I don't think anyone's insisting that there's nothing wrong with the system, but rather that there's not enough wrong with it.

Enough wrong for an immediate overhaul or large resources dedicated to a substitution? No.

Enough wrong for suggestions from fencers to be regarded as possibly beneficial? Yes. In other words, whenever someone suggests a rating system, there are cries of "this one's fine why are you bothering." My point is that it's not fine, and if someone was to suggest an adequate replacement it would be a good thing for fencing. Especially if it circumvented large resource expenditure.

I mean, historically, this system was created incrementally and reactively. It started out very simple, and changes were added as problems came up, without too much regard for the big picture. This is, of course, going to eventually result in a system that's sufficient, but shouldn't we someday take a step back and look at the big picture of what we really want out of our ratings system? And if someone attempts to do this, shouldn't we at least look at their proposal?

In other words, I'm fine with ratings proposals being dismissed as bad, as they mostly are. I take issue with them being dismissed as unnecessary or redundant.

[NGV]

I don't think anyone's insisting that there's nothing wrong with the system, but rather that there's not enough wrong with it.

Where there's "enough wrong" to attempt a fix depends on the cost and difficulty of the fix, as well as the size of the problem. If a minor problem would require a big and costly fix, then it's not worth it. But if a minor problem can be somewhat addressed by a smallish and easy fix, then why not?

I don't think that going to an Elo style system would be worth the trouble. To justify adding that much complexity, the current rating system would have to be a festering sore that was eating away at US fencing, and it's not actually that bad. On the other hand, it seems like it would be pretty easy to modify the classification categories so that competitions of drastically different strength would be less likely to award the same ratings. So why not do that?

[Inquartata]

By the same logic, (and by observing subsequent results) does winning one tournament translate to "consistently"? Both events say one thing: "At this point in time a fencer is "better than" some small group of other fencers."

True enough.

I thought he was purposely proposing a really bad system in order for us to learn to love Elo.

If so, he failed.

Wait, this isn't satire??

Have you met Mike? It's more like satyr.

In other words, whenever someone suggests a rating system, there are cries of "this one's fine why are you bothering." My point is that it's not fine, and if someone was to suggest an adequate replacement it would be a good thing for fencing.

No, it's still fine. Sorry if this causes you additional frustration, but alleviating your frustration is not a valid reason to change things. ( I am sure you realize this, but you keep bringing it up anyway. )

[gillaspy]

how is your book being received?

Since you asked, I've sold a few hundred copies, but certainly not a best seller. Those that have bothered to review it on Amazon say they liked it though.

[NGV]

Seeding, metrics and qualification are three different things. The current system is ok for the first, not so good at the rest.

I think that this is correct. That's why I'd keep the current rating system for seeding purposes, with some tweaks (modifying the categories to more precisely reflect competition strength and dropping the decay period to three years would be a start).

For metrics and qualification, I'd instead create an expanded national points ranking. This could be done in large part by awarding a small number of national points for high finishers at the ROC or whatever the ROC turns into ( the amount of points awarded for regional events would need to vary based on some strength factor calculation, given the wide disparity in ROC competitiveness). There would be some potential issues and snags that might arise from such a change, but I think that they could be dealt with - and having a less exclusive national points list would have some pretty clear benefits.