Is it easier to cheat in 6.0 or CoP?

[Mathman]

Finland rallies its natural allies, Mexico and India, to promote the fortunes of Kiira Korpi against her rival Miki Ando (equally stunning and a better skater.) It is a close contest, but with fair judging Ando wins on every judge’s card.

Is it easier for the three-person conspiracy to affect the outcome under ordinal judging or under point-total judging?

6.0 There are nine judges. Six judges vote for Ando, three for Korpi. Ando wins.

CoP The conspirators decide to cheat by elevating Korpi’s PCSs by one point and diminishing Ando’s by one point. They also decide to give Korpi one extra GOE point on half of her elements (say, seven out of the thirteen scored elements) and to take away one GOE point from Ando on half of her elements. Under CoP two of the judges’ scores are eliminated by the random draw, then highest and lowest are thrown out, and the remaining five scores are averaged.

Case I. The conspirators have bad luck. Two of them are eliminated in the random draw and the remaining one has all of his scores for both Korpi (highest on the panel) and Ando (lowest on the panel) thrown out. Ando wins.

The probability of this happening is 3/36 = .083.

Case II. The conspirators have medium luck. One of the three is eliminated in the random draw, one set of scores is thrown out for highest/lowest, and one conspirator survives to influence the outcome. Korpi (vis-à-vis Ando) gains an extra 2 points across the board for program components, which translates into an extra 3.2 points in total PCSs.

In GOEs the relative gain in TES is 2.8. This is a total relative gain of 6.0 points between the two skaters (long program only).

The probability of this happening is is 18/36 = .500.

Case III. The conspirators have good luck. All three survive the random draw. One is thrown out for highest/lowest, two conspirators’ votes count. Korpi gets an extra 6.4 points in PCS and an extra 5.6 in TES, for a total of 12 extra points. (Assuming more good luck, she could also pick up an extra 6 points from the SP.)

Probability of this happening: 15/36 = .417.

 

[janetfan]

Just wondering - where the tech caller fits into your equation?

I recall Morozov saying it is easier than ever to "change" outcomes under CoP based on calls by the the tech caller.
I also recall Orser had a few choice words about this.

The potential for tech callers to determine outcomes has even a skater as technically sound as Yuna showing doubts about her flip since gettting called on it last season.

Mirai at SC seemed to only be skating for the tech caller so shouldn't the role of tech callers be considered in any fair comparison of the two systems?

 

[fairly4]

they cheat the same.
the reason--goes -up or down/ missed calls for ur's/
easier to down pc's scores under cop.

easier to keep them bumped up with mistakes under cop.
than 6.0's without figures. why falls and mistakes kept people off podium/
gold medals without a perfectly--clean skate.
under cop--messy-urs/ high pcs-overmarked goe's keep skaters on top over skaters who skate clean--i don't mean skaters who ur/ mean skaters who skate clean and are overlooked .
example--rachel skate america=fall on triple/triple-counted all
but all other skaters on fall downgraded.
6.0 fall would have knocked her off.
but in reality- easy both ways.
the pc's scores and underlook/overlook ur's on certain skaters.
JDon't make any difference Cheating is cheating under both systems.
and unfair to SPIRIT OF OLYMPICS And fans, atheltes, and sponsors. just the same.

 

[cornell08]

On the one hand, you've got crazy program components that prop up skaters despite splatfests. Spectators scratch their heads at standings.

On the other, you've got no replay and majorly cheated triple jumps. Tara's Nagano freeskate's first 3-3 would have been marked UR on protocols, and her second 3-3 would have been downgraded.

6.0 was easy on jumps, but hard on the actual performance on ice. CoP is easy on actual performance, hard on details. I think its easier to cheat in 6.0 in terms of technical details....but usually the most enjoyed programs wins at least (so usually the cleanest to the naked eye). Any chance for hybrid system?

 

[Blades of Passion]

It's easier to cheat in CoP, but also easier to point out where the cheating happened (except that we don't know which judge entered each score - a HUGE problem).

 

[DesertRoad]

It is absolutely easier to cheat under the COP. Jump downgrades, with their double-whammy in point deduction, decide high level competitions all the time. And this all depends on one tech caller (though there are two reviewers who can but never do challenge the caller) deciding if a jump is landed 90 degrees short vs 89 degrees. The tech caller is a single point of failure, a very unreliable way to build something. Bribe the TC and you're all set.

Of course, this is painfully easy to fix in any number of ways. Like increasing the number of tech callers, changing the guidelines for what constitutes a downgrade, and/or removing one of the point penalties for them. So there's hope yet.

 

[enlight78]

The tech caller call can not really cheat that much if he is bribed; the performances are recorded; so doubt he would get away with downgrading an obviously rotated jumps; the rules of levels are clear so it would be too obvious if he threw out undeserve level ones the coaches will catch it; the best way for a tech to cheat is to overlook cheated jumps; wrong edges; and mistakes that will require to reduce a level; fans dont seem to catch that any way; Judges dont have that much room with goe's without risking being the highest or the lowest that will be thrown out; If three judges came together the best way to cheat would be PC's; In any other way it would be obvious; I think it was easier to cheat under 6.0 because it would be eaiser to get away with it; Just tip the vote so a certain skater get the majority of 1st'

 

[prettykeys]

Mathman, your probabilities are off. If you read this before I get through the analysis, then my subsequent post can be ignored. But I'm just letting you know right now.

 

[Blades of Passion]

It's easier to cheat in CoP, but also easier to point out where the cheating happened (except that we don't know which judge entered each score - a HUGE problem).

Actually, now that I think about it, it's not really "easier" to cheat under CoP. Since, in 6.0, the judge could just give whatever mark they felt like.

I do think CoP potentially discourages cheating more because every exact element is being scored.

But the anonymous judging system doesn't help.

 

[Mathman]

Mathman, your probabilities are off. If you read this before I get through the analysis, then my subsequent post can be ignored. But I'm just letting you know right now.

They are? The hypergeometric distribution. For x = 0, 1, 2, the probability that x of the three conspirators are discarded in the random draw, together with 2-x of the honest judges -- to me, it seems like that should be (binomial coefficient 3 choose x) times (binomial coefficient 6 choose 2-x) divided by (binomial coefficient 9 choose 2).

I was more worried that I didn't figure out the points right -- how the scores of one judge affect the averages and totals.

 

[Mathman]

Actually, now that I think about it, it's not really "easier" to cheat under CoP. Since, in 6.0, the judge could just give whatever mark they felt like.

However, a majority of the judges would have to do that. Or at least enough judges to switch the majority from one skater to another. One cheating judge could not prevail over 8 honest ones, except in the case of a perfectly even split among the honest ones, as in the Salt lake City pairs.

With the CoP, though, one or two cheating judges could absolutely drive up the total score of one skater and drive down another's, even if all the other judges are on the up-and-up.

 

[janetfan]

Actually, now that I think about it, it's not really "easier" to cheat under CoP. Since, in 6.0, the judge could just give whatever mark they felt like.


But the anonymous judging system doesn't help.

One of the most disputed skating teams under 6.0 were the Duchesnays, who received scores ranging from 5.1 to 5.9 at the same event their first few years.

It was addressed by officials and their scores did become more consistent after judges were warned about being unfair and biased against a more modern style.

Still, it was Ice Dancing and fixed placements were more or less accepted in those days.

Has CoP had any moments yet - a dispute that was addressed and changed (maybe scores at SC would be fun to analyze )
or has it been decided every carefully counted and reviewed mark has been good under CoP?

BTW, it is my opinion as a failed magician that it is easier to trick people if they are blindfolded.
When they can see what you are doing it is much harder to fool them.

 

[Mathman]

Just wondering - where the tech caller fits into your equation?

In my opinion the technical caller and his panel completely throws a monkey wrench into the analysis.

To my untrained eye, about 90% of all jumps (ladies) are underrotated. Whether they are underrotated by 80 degrees or 100 degrees -- that seems like flipping a coin. Sometimes the skater gets lucky, sometimes not. Whoever gets the fewest downgrades wins.

I agree with Enlight and Fairly4. All the tech specialist would have to do is let a few underrotations slide.

 

[janetfan]

In my opinion the technical caller and his panel completely throws a monkey wrench into the analysis.

To my untrained eye, about 90% of all jumps (ladies) are underrotated. Whether they are underrotated by 80 degrees of 100 degrees -- that seems like flipping a coin. Sometimes the skater gets lucky, sometimes not. Whoever gets the fewest downgrades wins.

I agree with Enlight and Fairly4. All the tech specialist would have to do is let a few underrotations slide.

Yes, I agree as well.
Morozov agrees too and I bet he stays up all night worrying what to do about this

Sometimes as gkelly likes to point out it is not about cheating or about incompetence or even intentional. Judges are human and unfortunately skating is very subjective. But I would hate to lose a major medal because a tech caller was having a bad day.

I think mm makes a good point with this 89% vs 90% and for me it is not only baloney but I honestly don't believe tech callers can ever have any hope of being consistent enough to make such close calls either accurately or fairly.

I like the old "if you can't see it Live with your eyes it must have been OK."
Atleast for the Ladies because I can't stand to see so many wonderful skaters so frustrated and scared. No wonder the joy has gone out of skating,

 

[Blades of Passion]

However, a majority of the judges would have to do that. Or at least enough judges to switch the majority from one skater to another. One cheating judge could not prevail over 8 honest ones, except in the case of a perfectly even split among the honest ones, as in the Salt lake City pairs.

With the CoP, though, one or two cheating judges could absolutely drive up the total score of one skater and drive down another's, even if all the other judges are on the up-and-up.

I was just talking about each individual judge.

Remember, though, the highest and lowest scores do get thrown out for CoP (although they shouldn't be; nor should random selection of scores happen). Too much deviation between marks also apparently pulls up a red flag in the computer.

 

[prettykeys]

Mathman, I am rusty as f*ck when it comes to probability.

But I always thought of it like this (and I don't know if the random tossing happening first before highest/lowest scores are dumped, or if the random tossing happening after, matters. But I'm assuming that the high/low are dropped before anything else happens.)

So one of the cheaters is always going to get dropped. An honest judge will also always get dropped. You have now 10 viable judges to do the random dropping from.

In case of having both remaining cheaters booted, it's a conditional probability. Given the 10 judges, there are 2/10 chances of a cheater being randomly booted. The chances of the second cheater being booted is 1/9.

And now since the order of the cheaters being booted (by high and random processes) can be mixed around 3 times, what you have is

P(A and B) = P(A) x P(B|A) = 2/10 x 1/9 x 3 = 0.066667

That is, there is a 6.7% chance that you're going to get all three cheaters out of the game (not very high at all.)

Is this wrong?

 

[prettykeys]

They are? The hypergeometric distribution. For x = 0, 1, 2, the probability that x of the three conspirators are discarded in the random draw, together with 2-x of the honest judges -- to me, it seems like that should be (binomial coefficient 3 choose x) times (binomial coefficient 6 choose 2-x) divided by (binomial coefficient 9 choose 2).

I was more worried that I didn't figure out the points right -- how the scores of one judge affect the averages and totals.

I don't care about the mathematical method too much (really, there are a lot of factors that aren't being considered although it is kind of fun to think of it this way ), but the chances of each judge being booted, or staying on, does not fall under binomial analysis. For one, the identity of who gets booted matters.

It's more like selecting a hand out of a deck of cards rather than getting a certain outcome out of flipping a coin several times.

 

[Mathman]

Remember, though, the highest and lowest scores do get thrown out for CoP (although they shouldn't be...

Actually, the trimming of the mean by throwing out the highest and the lowest is one of the features of the IJS that I do agree with. This does no statistical harm (as reducing the panel by the random draw does), and, as you point out, it diminishes the effect of one biased judge from high-balling his favorite and low-balling his favorite's rival.

The other thing it does is eliminate the score of the hapless judge who simply makes a key-stroke error -- punching in 1.75 instead of 7.75. Like the 1968 Olympics when Tim Woods lost the gold medal to Wolfgang Schwarz because one of the judges fumbled with his cards and accidentally held up the wrong number.

 

[Mathman]

... the chances of each judge being booted, or staying on, does not fall under binomial analysis.

Not the binomial distribution, the hypergeometric distribution, q. v.. (They both use binomial coefficients in the calculation.)

It's more like selecting a hand out of a deck of cards rather than getting a certain outcome out of flipping a coin several times.

Yes, this is exactly right.

There are two "suits" of cards in the deck, conspirators and honest judges. Call them clibs and diamonds respectively. There are three clubs in the deck, the king, queen and jack. There are six diamonds, the 2, 3, 4, 5, 6 and 7.

If you draw two cards at random from the deck, what is the probability that you will get two clubs and zero diamonds? What is the probability of getting one club and one diamond? What is the probabilty of getting zero clubs and two diamonds?

(But you knew all that. )

 

[prettykeys]

Apparently the order of whether the random booting happens first or after the high/low are determined, matters. If I'm not totally messing my probabilities up, the conditional probability is:

3/12 x 2/11 = 0.0454545 or 4.5% chance that you're going to get all three cheaters booted if the random booting happens first.