- Jul 1, 2005Yeah but that's the problem with anything involving people.

What's to stop me from posting an average of 15.00 on UWR list, even

thought I've never been under 16? And it's not that unbelievable

really. But it might be a bit odd, statistically, I might post a

fake average where all the hundredths of a second were under .50 for

example, which is statistically very unlikely. It's also unlikely,

given enough cubes, that you would beat your long term average by

more than 3 SD's.

But yeah, I agree with you, but if people want to cheat, then

they're only really cheating themselves.

Dan :)

--- In speedsolvingrubikscube@yahoogroups.com, Sachin Shirwalkar

<sachin_civilian@y...> wrote:> No Dan but there is a way to beat this. If you are

protection for FREE! http://in.mail.yahoo.com

> considering only the best improvement done by a person

> then I'll do my worst times in the first 30 solves and

> then in the real solves i'll solve normally. So my

> scores come out far better than the 30 previous ones.

>

> Though this looks like a good idea. Keep working on

> this.

>

> Sachin

>

>

> --- Dan <dan_j_harris@n...> wrote:

>

>

> ---------------------------------

> Hi Leyan, Tyson,

>

> The 30 - 40 solves would have to be fairly near the

> time of the

> competition, yes. As you say, if you were using an out

> of date

> statistic, it wouldn't work, because people's long

> term averages

> would slowly increase as they improved.

>

> This, in my opinion, is the main stumbling block of

> this method, how

> would you get all the competitors to bother to time 30

> or 40 solves,

> before the day of the tournament? Perhaps everyone

> should time as

> many cubes as they possibly can, in a practice session

> before the

> competition or so.

>

> Well, I guess we will just have to see what happens on

> July 16th!

>

> Dan :)

>

> --- In speedsolvingrubikscube@yahoogroups.com, Leyan

> Lo

> <leyanlo@g...> wrote:

> > Where would you get the 30 solves?

> >

> > If you're using solves that span a large time frame,

> this is not a

> good

> > idea because most people get faster the more they

> cube. It would

> > probably be more fair to fit their solve times to an

> exponential

> decay

> > curve up to a constant as a function of time, and

> then calculate

> their

> > expected average and standard deviation for the day

> of the

> tournament

> > from the fit.

> >

> > Leyan

> >

> > PS. From the statistics class I took, I remember

> the number 40

> (Law of

> > large numbers).

> >

> >

> > Dan wrote:

> > > Hi Everyone,

> > >

> > > I would like some input on the following method

> for judging a

> > > handicap competition, especially whether it is

> actually feasible

> to

> > > use in a competition!

> > >

> > > 1. From each competitor, have a long term Mean and

> Standard

> > > Deviation Statistic. From my Statistics course a

> long tim ago,

> the

> > > number 30 seems to ring a bell in my head, ie you

> have to make

> 30+

> > > solves before the statistics becomes valid.

> > >

> > > 2. After the competition, calculate each

> competitors average.

> Then

> > > calculate a Normal Probability, with competition

> average as

> the "x"

> > > statistic, and using the long term Mean and SD.

> This will

> calculate

> > > the probability of the competitor achieving this

> competition

> > > average, if they were to make many thousands of

> averages.

> > >

> > > 3. The competitor who has achieved the lowest

> average

> probability

> > > would win, and competitors are ranked in order of

> ascending

> > > probabilities.

> > >

> > > So, all fine in theory?

> > >

> > > I have tried plugging some numbers into a

> spreadsheet, and the

> > > results are quite promising. But one observation I

> have made, is

> > > that in real life, results don't seem to "quite

> fit" the normal

> > > distribution. It is perfectly likely for someone

> who is quite

> > > consistent in practice, to have a wildly varying

> average in

> > > competition, and the probabilities end up being

> almost 0, or

> almost

> > > 1. Is there someway of "stretching" the margins of

> the normal

> > > probability, so it can cover a wider range of

> results? One way

> would

> > > be to multiply the competitors standard deviations

> by a factor,

> but

> > > is this mathematically valid?

> > >

> > > Any help is greatly needed :)

> > >

> > > Dan Harris - www.cubestation.co.uk :)

> > >

> > > P.S. I hated Statistics in college ;)

> > >

> > >

> > >

> > >

> > >

> > > Yahoo! Groups Links

> > >

> > >

> > >

> > >

> > >

> > >

> > >

>

>

>

>

> ---------------------------------

> YAHOO! GROUPS LINKS

>

>

> Visit your group "speedsolvingrubikscube" on the

> web.

>

> To unsubscribe from this group, send an email to:

> speedsolvingrubikscube-unsubscribe@yahoogroups.com

>

> Your use of Yahoo! Groups is subject to the Yahoo!

> Terms of Service.

>

>

> ---------------------------------

>

>

>

>

>

>

> _______________________________________________________

> Too much spam in your inbox? Yahoo! Mail gives you the best spam

- << Previous post in topic Next post in topic >>