inflated ratings

[Chris]

Originally posted by a517dogg
five reasons i think my rating is inflated:

A's should make at least 32 in a Junior if they are under 20. i have not yet done this.

Well, first of all, there's no real justification for that statement, besides, noone said the system was perfect, and since you have NOT been to very many JNAC's since you were a 'D', you shouldn't really say that now, you won't: you were operating at a disadvantage: How much better would you have done (and who would you have fenced DE) if you had one more victory in every pool? This is how well you will probably find yourself doing, since you, in effect, have kicked an 'A' out of your pool, because that 'A' is now YOU.

There are two (now four) but two strong As in my division. i beat one of them at this tournement, first time ive ever beaten him, but i basically only beat him because he was confident enough that he was going to win to kind of goof around in the first period, after which i had three touches lead. the other A in the division, the only time i fenced him in a tournement he also goofed around and after the first period i had two touches lead. he still ended up winning 13-12. if i fenced them again in competition they would both beat my *** senseless, just like they do in practice.

Just because you are an 'A' doesn't mean you should beat every 'A', even if you fence them a couple times. Obviously, you are improving: don't let your performance be limited by previous events...

I earned my A in a mixed open. one of the two As that made top 8 was a female. hence, i have a girly A.

Well, certainly, there are 'A's that are stronger/weaker than other 'A's. You feel that you aren't as good as other 'A's because you went to a weak tournament, because of 1 person. MAYBE they are a weaker 'A', but MAYBE this is because she won a weaker tournament, too, and NOT because she is a woman! Secondly, familiarity CAN be a great equalizer, so maybe she beat some better fencers that she knows really well, and still doesn't have the most complete game. It doesn't have to be because she is female.

I beat the Bs 6-5 and 15-14, hardly convincing wins.

Well, 14-15 certainly ins't a convincing win; it's not a win, at all! It all depends on how things go: I recently won a bout, against a 'B' (who I had never beaten before) 15-14 the was very convincing, since the score HAD been 9-13!

Obviously, the 'B'you fenced was convinced he was in trouble, or the score would have been a lot higher!

i went from a D02 to a C02 to an A03 in two months and two tournements. my C02 was not inflated. My A03 i think is.

So what you are saying is the your classification as a D02 was DE-flated...sometimes our perception of how good people with a certain classification should be is warped by the fact there are people with LOWER classification than maybe they really could have.

'A' fencers in my opinion shoud be akin to black belts in tae kwon do or something: master of that weapon, including solid tactics, strategy, footwork, bladework, offensive game, defensive game, speed, endurace, etc etc etc... which i deffinitely do not have.

Well, obviously, that's your opinion, but I am not really so sure that's an approppriate comparison.
(how much do you really know about some of these belt systems?)

yet here i am stuck with this rating i dont really want (it doesn't let me fence div II events), in addition to which, expectations will be a little bit higher: i'll be seeded higher in san diego junior NAC than any junior result i've ever pulled. there should be some way to cancel your rating or something.

Well, I can understand that it will mean an unexpected, and perhaps unwanted, change in your competitive schedule, and don't get me wrong, maybe you really aren't so good as that 'A" would seem to indicate...but try to take advantage of it, if you can.

This does serve to indicate why I liked that feature of the Canadian system, wherein, if you did not repeat this feat in the remainder of the season, you would quite likely be demoted to a 'B', which would seem to be more appropriate, given what you have said.

Well, keep on pluggin!
It's still fun, right!?!

Well, gotta go; I am late for the salle!

[edew]

Originally posted by epeemike81
Let me revise. Have you ever seen somebody earn a rating once and never earn it again, despite having sufficient opportunities to do so? I have. many times. I do call those ratings flukey. almost as soon as a rating is RE-earned, you can't possibly call it flukey anymore. does this mean that all ratings are flukey until re-earned? No, of course not. and that portion of the debate is certainly not quantifiable. my coach is fond of telling a story about a friend of his who was a solid D, and who everybody knew was a D (in caliber). He ran into this friend at a NAC and the following conversation ensued:

Him: "I recently earned a B."
My Coach: "how'd you do that?"
Him: "why does everybody keep asking that???"

Why? because everybody's first reaction was shock, and their second was that he would never re-earn it, no matter how many opportunities he had. They were right. he had plenty of opportunity but has never re-earned the B, or for that matter placed highly enough to get a C. THAT is a flukey rating.

-m

I wouldn't call these flukes. I'd call them outliers within the expectations of what the classification system allows. A "B01," say, that your friend got means that he either earned a B in 2001 calendar year, or he had an A96 and it dropped to a B01 starting January of 2001. That's all that classification means. Because that's all it means, it's not a fluke. He might never attain anything close to that in the future, but it's still not a fluke.

[epeemike81]

Originally posted by edew
I wouldn't call these flukes. I'd call them outliers within the expectations of what the classification system allows.

generally, I have heard "outliers within the expectations of what the classification system allows" referred to as flukes. What, out of curiosity, do you think of as a fluke?

-m

[edew]

Given the loose nature of the classification system, it is not possible to have flukes. Since you're the one offering the possibility, perhaps you should define the term, instead of just labelling some people's classifications as such.

[oiuyt]

MAYBE they are a weaker 'A', but MAYBE this is because she won a weaker tournament, too, and NOT because she is a woman!

She's earned the A twice that I know of, finishing top 4 in div IA nationals one year and top 8 in div I nationals the following year. She is a very good fencer, but is noticibly weaker than all of the other A's in the area. Would she have an A if she were male? Highly unlikely.

Eric- From www.dictionary.com

fluke3 ( P ) Pronunciation Key (flook) n.
1. A stroke of good luck.
2. A chance occurrence; an accident.
3. Games. An accidentally good or successful stroke in billiards or pool.

In this case epeemike is clearly using definition 2. An outlier (A value far from most others in a set of data) IS a chance occurance. The ratings system is meant to approximate the abilities of fencers. When a result far from expectations is achieved, it's an outlier. Outliers (in either direction) are flukes.

Maybe this'll help. Assume that all fencers in a tournament are exactly equivalent and we have 250 fencers. We fence a straight DE table (whether or not we seed first makes no difference, these are exactly equivalent fencers each with 50% of winning each bout). After 9 rounds someone is now 9-0 in first place. We give this person the top seed in our next hypothetic tournament because they won the first one. Are they likely to win the next? Why not, they have the top rating, which they earned. Yet given the exact same pool of fencers for our second tournament they're just as likely to finish dead last as first (all of our fencers are still equivalent ability). Having finished first was a fluke. An accidental occurance that is unlikely to be repeated. An outlier in their expected performance (which should resemble a bell curve centered around 125th place).

Okay, re-reading that, maybe it won't help. I think it explains the situation, but if you don't get epeemike's original point then mine probably didn't make sense either.

The rating system is NOT meant to merely be a record of what our best (fairly) recent result was. It is meant to be a convenient and (fairly) easy way to roughly estimate the relative abilities of different fencers. It is IMPLEMENTED as a record of our best (fairly) recent result, but that is NOT its purpose. If all it were intended to mean were a shorthand (and loosely accurate) record of our top result, flukes would still be possible (due to the loosely accurate part of the shorthand), but would be less common and less important. That's NOT the purpose of the ratings system. To claim that it is, and therefore flukes are not possible is, at best, disingenuous.

-B :)

-B :)

[epeemike81]

umm.... I DID define the term. or rather, I accepted your description: "outliers within the expectations of what the classification system allows". So, I AGAIN ask, if you don't define those as flukes, what DO you define as flukes?

-m

[edew]

Brad, I think the effective definition for the term has to be interpreted within the context of this discussion, rather than opening a dictionary to page 320.

I didn't define fluke as an outlier within the expected allowances. I merely suggested that the person who accomplished the improbable B that you illustrated resulted in an outlier within the expected allowance. I claimed that since it's within the expected allowance, it wouldn't be considered a fluke in my book, although I do not have a hard-and-fast definition of flukiness to go by.

Here's my offer for a hard definition of fluke for fencing. A person's new classification is a fluke if the odds of the person obtaining the classification at ANY possible tournament within that person's lifetime is greater than 5,000 to 1.

So, the odds of a person obtaining a D in his or her lifetime is pretty low, say, 12 to 1 (I'm figuring, after twelve tournaments, the person is going to come out with a D or higher). Getting a C may be 20 to 1 at the outside, and dropping as the person improves. Getting a B will of course be higher, but the odds will drop over time. I doubt any person, even one who never touched a weapon in his or her life, will have preset odds of obtaining a B to be greater than 5,000 to 1. Maybe if the person is blind, crippled, hard of hearing, and devoid of all sense of touch, would the odds be higher than 5,000 to 1. Add to the list the person's innate dislike for the sport.

Why 5,000 to 1? It's arbitrary, I'll grant you, but I'm making the definition.

Now, in the case of your friend who unexpected earned a B. I doubt you would bet against 5,000 to 1 odds of him never getting it. And furthermore, given the fact that he has it, the odds are substantially lower than what you previously set, no matter how highly or lowly you think of his fencing skills.

So no, I don't think most, if any, high classification obtainment is a fluke.

[oiuyt]

Eric- Are you intentionally obfuscating the conversation or do you actually believe your comment has content? You asked for a definition of fluke, I looked up a fairly easy to find source for one. You then complained that I looked up a definition and rather than accept that I'd found what the word meant, you invented your own arbitrary definition that, through convoluted reasoning that had at BEST a tenuous connection to the preceeding discussion, helped support your point while being confused enough to make disputing it hard.

Okay, without resorting to references that state how the word is commonly used, here's a definition: A fluke is an unlikely result that occurs through some combination of random chance, luck, and exceptional circumstances (which includes people under/over-performing on a given day, either the person who achieves the fluke or other involved people).

Requiring 5000 to 1 odds is a bit excessive for something to be a flukey result. If I flip a coin 5 times and get heads every time I'd be surprised. That's an unlikely result. I'd comment on it to a friend who was sitting next to me. Heck, I'd probably comment after just 4 flips. 16 to 1 against is therefore within a reasonable definition of fluke for me. After 3 flips? I'd be surprised, but probably not enough to bother making a comment. So the boundary for a fluke is somewhere near there. Let's split the difference and call it 12 to 1 against for it to be a fluke. My definition makes a lot more sense than yours (in that I actually have a basis for my numbers and my numbers are much more likely to be close to those chosen by the average person). Actually I've just noticed that I've overestimated what it takes to be a fluke. I'd be equally surprised by a string of EITHER all heads or all tails. Since we now have 2 outcomes that hit that level of flukiness the odds of a fluke now only require about 6 to 1 against.

I'll put it another way. If I went down to the corner convenience store and bought a ticket for the "daily 3" lotto drawing I'd have approximately a .001 chance of winning. By your definition my winning wouldn't be a fluke. Just something within the boundaries of expectation. I know I'd be happy though. Why? Because I'd just won ~$500 that I wasn't expecting to win. My winning would be a fluke. I would not expect to be able to replicate the performance. Yet with 5000 tries I WOULD expect to replicate (five times over if probability holds which I expect it to with large enough sample sizes).

Now we go into the part of your argument where you seem to be intentionally muddying the water. The whole 12 to 1 of earning a D, 20 to 1 of a C, etc. bit. WTF? Pulling numbers out of air? What are they even supposed to mean? Is this a brand new novice fencer? An A-level fencer? A fencer of the level that s/he is attempting to earn? Okay, the only time you actually state what the given are you have:

I doubt any person, even one who never touched a weapon in his or her life, will have preset odds of obtaining a B to be greater [sic] than 5,000 to 1.

Are you kidding? If a novice came into a Bay Area Cup event having never fenced or practiced, etc. and borrows equipment you'd bet on them to walk out of there with a B if I gave you 5000 to 1 odds? (Okay, maybe DitD, let's move it to a state where betting on sports is legal). The average American? If you actually believe that statement I feel sorry for you. I also have a bridge for sale.

For the fencer in question, referenced in epeemike's anecdote, would I say that he has a less than 1 in 6 chance of re-earning the B (or more to the point, at the time would I have said that, his skill level has since changed). Yes. Conveniently, in the slightly over 2 years since, although his ability has gone up, he has never come close to reearning.

I still don't see what's wrong with using commonly accepted definitions for non-fencing related words when used on this board. Until and unless an alternative definition is agreed to, the default definition used SHOULD be the one in the dictionary. Okay, that means that your statement was wrong. Now WE should accept you modifying the definition of a common word until you're right?

Finally, your final statement.

So no, I don't think most, if any, high classification obtainment is a fluke.

Of course not, none of us claimed that the majority of high results are flukes. I think that MOST high classifications are earned and deserved. In my experience they tend to be reearned and the top people (who have said top ratings) tend to finish near the top of most tournaments that they compete in. In a seperate issue I think that the way we calculate ratings helps allow for flukely ratings (by taking just the high outliers and ignoring the rest of the results earned by someone), but that does NOT mean that I think that most high ratings aren't justified, merely that some are. Those few that are caused by flukes, which, by their very nature, are uncommon.

Okay, this post is getting WAY too long. In summary your definition of fluke is crazy, seemingly intentionally distorted enough to make it meaningless but support your point of view. You've denied a common dictionary look-up of a normal english word (without justification other than the fact that I used a reference book to bring FACTS (ohh horrors) into the discussion). I've proposed a definition (with a basis, something you neglected) for a quantifiable definition. Most of the rest of your post I don't really understand, it seems like you're making up numbers for a bunch of unrelated points that don't connect (either to the conversation or each other). Feel free to clarify and let me know what I'm missing.

-B :)

[MikeHarm]

Anyone know what the actual odds are for getting the various ratings within your lifetime? That would be interesting to know as well as if it was different for different weapons.

[epeemike81]

Originally posted by MikeHarm
Anyone know what the actual odds are for getting the various ratings within your lifetime? That would be interesting to know as well as if it was different for different weapons.

It is far too complex a problem to give odds for. Also, the odds are different for each person. I suppose you could say that the odds of getting an A are (Number of A's)/(Total Members), but we all know that they are significantly higher for some and lower for others. After all, at the beginning of the season, none of us would say that the Bengals have a 1/32 chance of winning the super bowl, or that the Devil Rays have a 1/30 chance of winning the World Series. By the same token, I would give the Yankees far higher odds than 1/30 (unfortunately).

-m

[edew]

The reason I made my definition as I did is because it's a testable definition. You can decisively determine whether a person getting a classification is a fluke or not. In your case, you're just substituting the word "fluke" with "unlikely". Neither of which is quantifiable. What exactly does "unlikely" mean? I just arbitrarily placed a number, 5000 to 1.

(As an aside, I have a standing bet with a coach here in the Bay Area for $1000 -- I can't win, he can't lose [stupid me in poorly stating the betting criteria]. The bet is that a certain fencer will never get a B. In his lifetime. If he gets his B, the coach wins $1000. If he doesn't, I don't have to pay. Since there's no loss to the coach, it's essentially an infinity-to-1 odds of him getting the B, which would be considered a fluke, according to my definition.

So far, I'm not too worried, and the coach isn't expecting any financial windfall for a very long time.)

[U.M.Amherst.Sabre95]

I think we should clarify the term "inflation". By the term inflation, I mean that there are many cases where it is easier to earn a rating now than it has been in the past.

I'd have to agree speaking from past experience. Back when I competed for UMass (1991-1994) the number of rated fencers was significantly smaller than those I'm currently aware of. That being said, I don't believe there was quite the saturation of fencers involved on a competative level back then, but there were still a good number.

Our head coach at the time held C- classifications in all three weapons, a source of no small respect considering our better fencers probably held D-ratings at best.

I'd have to say if those fencers (the ones having earned D's,) were competing today, based on a comparable level of skill, they'd be rated higher.

Just one man's opinion

[oiuyt]

A fluke IS an unlikely event. The fact that they are equally vague is NOT are argument that they aren't the same thing. If anything that would help support the contention that they are synonymous (obviously not sufficient, but it is required). A fluke result and an unlikely result mean exactly the same thing. In this context they are equivalent. Complaining about the fact that I'm using a synonym to help define a word is, well, I can't even come up with a word for what it is. Intentionally obstructionist, pointless, non-sensical all come to mind.

How is 5000 to 1 any more testable than 6 to 1? While 6 to 1 is ALSO somewhat arbitrary, it's less arbitrary than 5000 to 1 (I gavea rational basis for my number). As to testibility they're identical, other than the sample size required to test each reliably. 6 to 1 requires a MUCH smaller sample size (convenient given that a large sample would be hard to obtain).

How do you come up with infinity to 1? The odds that you GAVE were infinity to 1, but that has absolutely no bearing on the odds of the outcome. If I offered you a bet where I'll pay you $1 if the American League win the All-Star Game and you pay me $1000 if the National League does does that mean that the AL has 1000 to 1 odds of winning the series? No, it means that you've made a sucker bet. Clearly the odds are much closer to 1 to 1. You could make the same bet you did about Bissdorf earning a B (assuming he had just come over to the US for whatever reason). Doesn't mean that he won't earn it (or that it would be a fluke).

-B :)

[edew]

Originally posted by oiuyt
A fluke IS an unlikely event. The fact that they are equally vague is NOT are argument that they aren't the same thing. If anything that would help support the contention that they are synonymous (obviously not sufficient, but it is required). A fluke result and an unlikely result mean exactly the same thing. In this context they are equivalent. Complaining about the fact that I'm using a synonym to help define a word is, well, I can't even come up with a word for what it is. Intentionally obstructionist, pointless, non-sensical all come to mind.

How is 5000 to 1 any more testable than 6 to 1? While 6 to 1 is ALSO somewhat arbitrary, it's less arbitrary than 5000 to 1 (I gavea rational basis for my number). As to testibility they're identical, other than the sample size required to test each reliably. 6 to 1 requires a MUCH smaller sample size (convenient given that a large sample would be hard to obtain).

How do you come up with infinity to 1? The odds that you GAVE were infinity to 1, but that has absolutely no bearing on the odds of the outcome. If I offered you a bet where I'll pay you $1 if the American League win the All-Star Game and you pay me $1000 if the National League does does that mean that the AL has 1000 to 1 odds of winning the series? No, it means that you've made a sucker bet. Clearly the odds are much closer to 1 to 1. You could make the same bet you did about Bissdorf earning a B (assuming he had just come over to the US for whatever reason). Doesn't mean that he won't earn it (or that it would be a fluke).

-B

5000 to 1 is just as testable as 6 to 1. I just threw a number out there just as you did. You need to understand probability better. Here's the way you phrase the question: pick a fencer, X.

YOU first take, say, $1000 (must be something substantial, otherwise, you'll get the lottery effect, people willing to put money on money-losing propositions because the amount is small). At what return for the $1000 would you place your bet on fencer X earning a new rating at the NAC in San Diego (for example)? Are you also willing to take the reverse bet? When the return is balanced (i.e., you can't decide on whether to accept the bet or the reverse bet), then the odds that make the return is the odds you choose.

Now, do that with enough interested parties to average out the odds. If that average is greater than 5000 to 1, then, if the person succeeds in obtaining his rating, it's a fluke.

That I chose 5000 to 1 is arbitrary. You can, indeed choose the 6 to 1 odds. But, I don't think 6 to 1 in anyone's eyes is considered a fluke, in most people's colloquial understanding of the term. (I'd consider myself a genuine "A" fencer, but my odds of getting an A in San Diego, say, is far lower than 6 to 1. Maybe 300 to 1.)

[achilleus]

Originally posted by edew
(I'd consider myself a genuine "A" fencer, but my odds of getting an A in San Diego, say, is far lower than 6 to 1. Maybe 300 to 1.)

I would consider you making the top 8 at San Diego a fluke.

[BugabooX]

fluke fluke fluke.

You'd think the "fluke" ratings being talked about here were actually something special or unusual. Like winning the lottery, perhaps?

The problem with that perspective is that it fails to take into consideration that such ratings are part of the system, not an odd exception to a steadfast rule. But that's simply not true. With all the variables at any competition, combined with fencers' unique attitudes, goals, outlooks, etc., combined with uncertainty about what will come to pass in the future, combined with all sorts of other crap ... Well, give it up already. Claiming a fencing result is a "fluke" is about as valid as being surprised when someone actually wins the lottery.

Just consider some of the advice posted here or heard elsewhere in fencing clubs: Waiting for hours between pools and DE is part of the game, so you'd better learn to relax and focus. You can't control the referee, only yourself. If rotten traffic made you late to arrive for a competition, then maybe next time you'd be smart to leave earlier. Don't trust a borrowed blade. Leave your non-fencing emotions off the strip in your gear bag. Etc.

Now imagine that your doofus victim faces a series of opponents on his way to the finals, and that each person has failed in some minor way to fence at the top of his (or her) game. Is it a fluke that he's successful? No. It's part of the big picture, part of the game.

People in this thread keep putting the entire onus of the fluke outcome on one person. But if that fencer has put himself in the right place at the right time and is performing at the best of his ability, then it's no mistake that he won.

And don't even get started on the nonsense about being able to repeat his performance to earn the rating twice. Retrospection is always 20/20. You can't justly assume you know enough about a fencer's personal life to ever determine that he will or will not improve. I said it earlier: People change. Interests change. Circumstances change. And a letter grade is a vague estimation of ability that implies a range -- a C+ is only one "fluke" point away from a B-. I've heard people refer to a "strong A" vs. a "weak A," so it's clearly been recognized that not all letters are equal.

The so-called "fluke" rating cannot exist within our sport.

[epeemike81]

Originally posted by edew
YOU first take, say, $1000 (must be something substantial, otherwise, you'll get the lottery effect, people willing to put money on money-losing propositions because the amount is small). At what return for the $1000 would you place your bet on fencer X earning a new rating at the NAC in San Diego (for example)? Are you also willing to take the reverse bet? When the return is balanced (i.e., you can't decide on whether to accept the bet or the reverse bet), then the odds that make the return is the odds you choose.

Now, do that with enough interested parties to average out the odds. If that average is greater than 5000 to 1, then, if the person succeeds in obtaining his rating, it's a fluke.

What you are talking about here isn't probability, it's oddsmaking. The two have little to do with eachother. Oddsmaking is about what the public (in this case the fencing public) thinks, whereas probability is about actual likelihood. Oddsmaking is designed to equalize bets on each side. for example, the Patriots were 14 point underdogs in vegas for the superbowl last year. Did the oddsmakers actually think they were likely to lose by 14? who knows. the bottom line is that they arrived at that betting line because about half the people in the country felt that they would get crushed. it has nothing to do with what was likely to happen, and everything to do with how people were betting.

You want a way to measure flukiness? fine, here it is, though its somewhat impractical: Take all of the A's in the country and take their re-earn quotient. that is, take the number of times they have earned their A and divide by the number of times that they have had an opportunity to earn an A since the first time they earned it (inclusive). now, graph the frequency of occurance of the quotients. you now have a bell curve (roughly). You see (in this case, conceive of) those very low re-earn quotients at the bottom (i.e. people with high opportunity, but very low numbers of earning)??? THOSE are flukey A's.

Clearly, the same can be done with ANY rating. As for what odds make something flukey, it is arbitrary, confusing, and irrelevant. the bottom line is that when you look at a bell curve, anything with a frequency of less than 10 percent of the peak height is a fluke. (assuming a decent statistical sample size)

Admittedly, this will not reveal whether it is a fluke immediately, but it certainly can confirm suspicions over time.
-m