[CQ-Contest] Kudos to CQ WW for Busting Packet Cheaters!

[Ward Silver]

So - interesting discussion, including several off-line posts.

We agree that there are three "domains"...

<-- legitimate --> <--possible cheats--> * <--obvious cheats-->

* - DQ threshold

I also agree that the margins of winning or losing are sufficiently narrow
that IF the middle domain is broad enough, the clever cheat could get enough
advantage to make a difference.  The goal of the log checkers is to reduce
the middle domain to the point where it is narrow enough (either
mathematically or psychologically) that cheaters find it difficult to hit.

You also need to compare operators with the same capabilities for the
argument to be "tractable".  Given a certain DQ threshold, does the effort
required to cheat, yet remain below the detection threshold, equal or exceed
the effort put forth by the legit operator?  That would be the goal of the
checkers - to set the threshold such that the risk/benefit ratio is low
enough to make significant cheating a non-worthwhile activity.  Kind of like
counterfeiting $1 bills or pirating $5 paperbacks.

By keeping the threshold and detection techniques unknown, the checkers can
impose extra caution on the part of the cheater, lowering the effective
threshold even further.  I'm sure there are a number of doctoral theses on
this exact problem in Game Theory.

We're still in the early stages of being able to detect and penalize
cheaters, so it remains to be seen how much of a deterrent or behavior
modifier this will be.  Anyone care to compare the reported scores in the
archives against the published results to see who is missing?

73, Ward N0AX

----------------------------
>
> I would agree with Bill. Thirty "bonus" mults in the ARRL DX contest can
> make a big difference in score (~10% for SOAB). That is less than one ill
> gotten mult per hour. Since a legitimate single op will inevitably work
> some new multipliers just after they are spotted by coincidence (false
> detects), there will be a region of statistical overlap between the
legitimate
> operator and the cheater which will force the log checkers to set the
> threshold of disqualification high enough so as minimize the chances
> of disqualifying a legitimate competitor who by sheer chance has
> accumulated a statistically high number of packet spot correlations. This
> is the region of uncertainty that a shrewd cheater could use to pad his
> or her score by hedging against the committees desire not to DQ false
> positives. Of course, one could argue that the average cheater is lazy,
> so they probably wouldn't want to go to the trouble of picking and
> choosing when to make the cheats (too much like work). This is
> especially true when you consider the prize money at stake :) :).
>
> It would be interesting to compare the packet correlation statistics of
> known packet users (M/S, M/M, and SOAB(A) stations) with those of
> known legitimate single op unassisted stations over a large number of
> contests to see how much "space" there is between the packet users
> and the non-packet users.
>
> 73 de Mike, W4EF...........
>
> >
> > >Do it cleverly enough to avoid detection and the advantage conferred by
> > >the proscribed behavior, by definition, vanishes :-)
> > >
> > >Said in less turgid prose, if nobody can tell the difference between
the
> > >cheater's log and a legit log of a similarly equipped station, then the
> > >cheater hasn't gained an unfair advantage.  All the effort that went
into
> > >cheating balances a legitimate effort.
> >
> >  N0AX
> >
> > _________________________________________________________
> >
> > I don't see how you figure.  If the cheater is clever enough to
> > evade detection and manages to pick up some mults he wouldn't
> > have otherwise gotten, he's ahead of where he would have been had
> > he gone legit.  If those mults are enough to beat out a legit
> > station, he wins by cheating.  Had he not cheated, the legit
> > station would be hanging the certificate.
> >
> > Right?  Tell me where I'm wrong.
> >
> > Bill, W7TI


--
CQ-Contest on WWW: http://lists.contesting.com/_cq-contest/
Administrative requests: cq-contest-REQUEST at contesting.com