Rating system - yet again

[PeterGustafsson]

Hi!


As we are reading yet another set of rating system threads, I thought that I may offer my latest rework suggestion. How about this:

1. In order to advance provisionally one level, a fencer must earn 5 rating points at any one competition in one weapon.
2. If one wins over a competitor of the same rating level as oneself, one gets 1 rating point. Winning over a fencer one level above oneself gives 2 rating points, and so on in an arithmetical progression.
3. All rating points earned in one competition, for one weapon and age group, are added together to see whether they reach the 5 point level.
4. If one is to advance provisionally more than one level, rating points should be counted from the level immediately below the one that one wishes to advance to.
5. If a fencer has earned a provisional level three times, he will advance to that level, and stay there until three years have passed from the earning of the 1st provisional earning of that level. If that 1st earning is more than 3 years ago, he will not earn an advance, but will retain his provisional advance for more tries later on.
6. All USFA fencers will start out the system with the ratings they currently have.
7. When U-rated fencers earn an advancement, they should be advanced to a new level, H. From there, they should advance via H->G->F->E
8. A few high-level competitions (regional, sectional, natl championships) should give out instant ratings as the present system of Ax charts.

Example: Fencer A is in a 2-poule, straight DE competition. He is rated E, and his results are:
Poule 1: Win over 3 E´s, Win over D, loss against C
Poule 2: Win over 2 E´s, Win over 2 D´s, loss against C
DE 1: Win over C
DE 2: Loss against D

Poule 1 gives him 3*1+1*2+0=5 points
Poule 2 gives him 2*1+2*2+0=6 points
DE 1 gives him 3 points
Total rating points: 5+6+3=14 rating points, easily giving him an advancement to provisional D.

Will he advance to C rating?
Poule 1: 3*0+1*1+0=1 point
Poule 2: 2*0+2*1+0=2 points
DE 1 gives him 2 points
Total rating points: 1+2+2=5 rating points, just barely giving him an advancement to provisional C.

If he then posts two provisional D results the following weekends, he will be upgraded to D counting from the 1st competition, and a provisional C to boot.

What are the thoughts behind this system?
1. Mathematically simple - just add integers.
2. Only your own bouts matter - no motive to ask a teammate to throw a bout against someone else in a DE so that enough high-rated people end up high enough.
3. No cut-off effects - not important to get warm bodies for rating purposes.
4. Flukes will not mess up the system - if they can get 3 provisional advances, they are not flukes.
5. More rating levels - seeding ties can be broken with level of, and # of provisional ratings.
6. More rating levels - ratings H,G,F will help at the low level.
7. Easier to reach a 1st goal for newbie - clubs can host a simple intramural for beginner->H.
8. Feature #4 makes it impossible to advance several levels by beating scores of weak fencers.
9. Feature #6 ensures a simple transition from current system.
10. Feature #8 ensures a steady flow of high ratings to keep the system going.
11. A fencer is never penalised for competing - losses do not enter into the system.

So, what do you think? Any glaring omissions, glitches, problems?


Have a nice time!

Peter Gustafsson

[downunder]

my head hurts.

Ratings are simply not that important.

[CvilleFencer]

The main problem that I see, and that I have mentioned with any of the points based systems that I have talked to people about, is logistical. Every division has it's own policies and handles things differently. Often the way things are handled varies a good amount with who is elected or appointed to their division boards every year or so. At its simplest, divisional results (DE slips and an overall tableau for each weapon in a tourney) are sent in to the main office (where I am pretty sure they are stored in the little brown filing cabinet, or maybe the little blue one if they are eco friendly) along with a ratings change form for the people whom that is applicable and the person who does such things notates/updates the ratings change and sends out a new card to the fencer in a few weeks/months. Until then the fencer has his old card signed and has his new rating penciled in.

As simple as this system is, it still does not always work as it should. Division officers may forget to mail things in, the office makes typo's/errors in the recording of the ratings, etc.

The system you propose would mean that the division (and more importantly the division officers) would have a huge amount of data crunching after every tourney in the hopes of getting things correct. The divisions would also have to store all this stuff in hard copy backups somewhere in case there was and error/problem. Even assuming the already busy division officers spent the time to do all this correct, made the space money needed to store the data (after a few years this gets to be a problem, and how long do fencers stay in a database?) then the results have to get to the USFA office.

There they would have to do the same for tens of thousands of fencers with new results to be added/updated coming in every week. This would become an overwhelming amount of data entry, records retention and quality control work for an organization that can't even keep its website up.

I like the system in theory (mainly because I am pretty sure I would be one level higher in each weapon than I am now in this system!) but logistically it just would never happen and would not work IMHO.

Also, the US is a big place with lots of fencers competing. Many of them are
casual fencers that may compete once or twice a year or worse yet, do so and then leave the sport. Others are hard core collegiate fencers (and ratings help add status and may even help with team eligibility, admissions, etc) that are damn good, but only fence in USFA comps a few times to earn/maintain a rating. At least until they are out of school. That would be a lot of effort to spend on people who don't really care about USFA events or are just playing with the sport to "try it on".

I think a much easier thing would be to mandate that all divisions have to update their tourney results to Fred to be combined into a master points list for the country (like the UK) and then take X number of results for a given year and give Peet a percentage to work from. As an example the top 2% get an A, the remaining top 4% get a B and so on. And best of all Peet already has the infrastructure to do it, probably with some tweaking and maybe some added bandwidth! That way we would have a national points list for everyone, a fair system based on results of bouts fenced via placement in a fencers best 6 (as an example) tourneys for the year and a good way to prevent flukes and ratings inflation as the number of top fencers would always be a reasonable percentage.

There are probably some problems with that system, since I just made it up while home sick with the flu and I have not even finished my morning wake up beverage yet, but it would seem to be both fair and workable at my fist addled glance. What do you guys think?

[keith]

I think a much easier thing would be to mandate that all divisions have to update their tourney results to Fred to be combined into a master points list for the country (like the UK) and then take X number of results for a given year and give Peet a percentage to work from.

well the problem is that, as you point out, america is a big country - how do you compare a 60 fencer comp between regions? In smaller countries a national list based on opens works since most of the good fencers attend most opens; so it is easy to work out weightings for each comp.

Only way to work this in the US is to go to a regional rating system. Which might not be a bad idea - leave USFA ratings for the NACs and nationals.

[esskreemr]

well the problem is that, as you point out, america is a big country - how do you compare a 60 fencer comp between regions? In smaller countries a national list based on opens works since most of the good fencers attend most opens; so it is easy to work out weightings for each comp.

Only way to work this in the US is to go to a regional rating system. Which might not be a bad idea - leave USFA ratings for the NACs and nationals.

I proposed this earlier. Your rating would = Letter+Level+Year.

So someone recieving a C in 2006 at a regional level tournament would be classified as a Cr2006 whereas someone recieving a C in 2005 at a divisional tournament would be a Cd2006.

Seeding could then be done by letter, level, year.

[KD5MDK]

That assumes divisional level tournaments are by definition easier than regional ones, which isn't the case.

[glowstix]

my head hurts.

Ratings are simply not that important.

exactly.. they really don't tell you anything. just another opportunity for an "i'm better than you" moment. sure, they give somewhat of an indication but too much emphasis is placed on them. yeah, the system sucks, especially when you can earn a rating without having even beaten a fencer of that rating on one day, and on another day, you can beat many fencers above your rating and not get a new one. the system is flawed but its the best we have right now .

[oiuyt]

Peter- Your system will result in VASTLY different outcomes with different event formats. Last May I spent a 3-day weekend fencing in a series of poule-unique tournaments at a club in the Boston area (1 weapon/day). Mid-20's in ROW weapons, over 40 in epee. MANY more bouts from which to score points than in a standard pool->DE event. In epee the average number of bouts fenced would increase from slightly less than 7 (pools of 5/6 to an average of 2 DE bouts) all the way to 40. These aren't comparable events classification-wise.

Even looking just your example, having a second round of pools, which isn't common in the US, that extra pool ended up representing nearly half of the points earned by your hypothetical fencer at either level.

-B

[Feraud]

Ouch, my brain hurts just looking at it!

[esskreemr]

That assumes divisional level tournaments are by definition easier than regional ones, which isn't the case.

Yes. It also assumes that 'division' level events don't draw competitors from outside the division. Another faulty assumption in many cases.

I didn't say it was perfect, just ideas being presented.

[Feraud]

Why does this pointless subject keep coming up again and again? Besides I've already come up with the perfect system if anyone wants to adopt it!

What we need is a ranking system that clearly describes who you beat in the last tournament, how tough it was and what kind of moves you’ve got

So there will be a series letters and numbers

First, three letters:
The first three letters are the people you defeated in your last three bouts.

Followed by three numbers:
Indicating the number of; successful attacks/ number of good parries / number of counterattacks

Then two digits for the year earned

Followed by a –00 for the month

Oh, wait and a letter up front for F E or S

The in parentheses are the numbers of rated fencers in your last tournament to show how tough it was.

Here’s an example:

F-AAB/9/8/12-04-11(12A-12B-6C-4D-8E-22U)

[nyacfencing]

Neh... I have been convinced that the way to fix this system is quite simple. Just change the year from "earned" to "expires", and make all ratings good for one year (cumulative). If I win a B today, I will have a B06 (The B is good through 2006). If I win another B tomorrow, I will have an B07 (The B is good through 2007). etc. This is easier to understand as it is clear when your time runs out, and it differentiates between a fencer that earned one B this year and 3 B's this year (the B08 would be seeded above the B06).

[KD5MDK]

However, the really good fencer who spends most of his time in practice and goes to 3 tournaments a year should not be seeded worse than the decent fencer who goes to a dozen.

[D+F+P=Hadouken!]

Why does this pointless subject keep coming up again and again? Besides I've already come up with the perfect system if anyone wants to adopt it!

What we need is a ranking system that clearly describes who you beat in the last tournament, how tough it was and what kind of moves you’ve got

So there will be a series letters and numbers

First, three letters:
The first three letters are the people you defeated in your last three bouts.

Followed by three numbers:
Indicating the number of; successful attacks/ number of good parries / number of counterattacks

Then two digits for the year earned

Followed by a –00 for the month

Oh, wait and a letter up front for F E or S

The in parentheses are the numbers of rated fencers in your last tournament to show how tough it was.

Here’s an example:

F-AAB/9/8/12-04-11(12A-12B-6C-4D-8E-22U)

Your rating system needs a few more features. Such as who your coach is, how long he's been coaching, where you practice, how many times a week you practice, the size of your legs in comparison to the rest of you, years fencing, how many comic books you read and how many grams of protein you consume per day. Got it?

[zorro19]

It seems that a fencer could become an A rated fencer without ever beating an A. If he/she fenced in alot of competitions it should be easier to become a B or an A then the current system, or am I wrong.

[JEC]

I think the easiest and most helpful change would be to add the number of times it has been re-earned.

i.e.: A05-8 vs A05-1. Clearly, somebody who has re-earned 8 times their A would be better than somebody who has earned only once.

[mrbiggs]

Clubs can barely handle the chart we have now, which is easy, consice, and fits on a page.

Last year, some board member whent through all the division websites and looked for mistakes in some qualifier. He found mistakes in, IIRC, almost half of them. This is again a very simple set of rules.

In other words, I don't think it's practical to make a formula, a more complex set of rules, or any change at all, even. There isn't anything horribly wrong with the system we have now (yet).

[mrbiggs]

It seems that a fencer could become an A rated fencer without ever beating an A. If he/she fenced in alot of competitions it should be easier to become a B or an A then the current system, or am I wrong.

You have to come in before an A in a tournament (usually two or more), and a good deal of the time you have to beat an A in a DE. The only exception is the A3 format, but I think it's very, very rare that there are 24 Bs and 12 Cs without any As in a tournament.

[Inquartata]

I have heard that there exist such things as math puzzles, equations that no one has yet been able to solve. Maybe you math lovers should work on those instead of trying to tinker with our rating system! We don't all love calculation, and some of us think that we should leave well enough alone!

[MikeHarm]

Sounds like the honor system in World of Warcraft a little. I kind of like it.