[NSRCA-discussion] Scoring Process Question

Tue Jun 26 10:30:38 AKDT 2007

Mike, 

Take some time and read it with a glass of wine tonight<G>...My point is exactly that we are creating an ILLUSION of accuracy which is not statistically present.  If my statistics are correct, scores are only accurate to about 100 points of the 1000 point scale.  We are deciding most of our contests on the statistical "noise".  

I haven't proposed any change, I'm just asking for ideas......If I had a better solution, I'd offer it.  I think that you are right in that expanding the judges score to more digits won't help because it is an inherently subjective number that can't be quantified more accurately than "about a half a point" on a ten point scale.

George  

---- Michael Wickizer <mwickizer at msn.com> wrote: 
> My head hurts after trying to read and follow that.
> 
> However, it strikes me that you are trying to attach mathmatical and 
> statisical validation to something that only has two numbers and that each 
> contain a varying amount of subjectivity.  I am not sure that using a 1000 
> point per manuver system or even greater, would make it more valid but only 
> an illusion.
> 
> 
> >From: <glmiller3 at suddenlink.net>
> >Reply-To: NSRCA Mailing List <nsrca-discussion at lists.nsrca.org>
> >To: NSRCA List <nsrca-discussion at lists.nsrca.org>
> >Subject: [NSRCA-discussion] Scoring Process Question
> >Date: Tue, 26 Jun 2007 12:50:48 -0500
> >
> >I'm going to open a can of worms here in hopes of coming up with a better 
> >system out of the discussion.  Perhaps this has been discussed before and 
> >I'm not aware of it.  Let me preface this by saying I am not a 
> >mathematician or statistician, but I have some familiarity with both 
> >subjects and the following question has been growing in my mind for some 
> >time.
> >
> >It seems to me that we are judging our maneuvers with limited accuracy 
> >(within 1 point in FAI and X.5 points in AMA classes) we are then creating 
> >the ILLUSION of accuracy by multiplying that score by a K factor and then 
> >normalizing to a 1000 point scale.  Here is a fairly brief explanation of 
> >"Significant Digits" that I've copied from the web which will introduce you 
> >to this thought if you haven't seen it before:
> >
> >****"SIGNIFICANT DIGITS
> >
> >The number of significant digits in an answer to a calculation will depend 
> >on the number of significant digits in the given data, as discussed in the 
> >rules below. Approximate calculations (order-of-magnitude estimates) always 
> >result in answers with only one or two significant digits.
> >
> >When are Digits Significant?
> >
> >Non-zero digits are always significant. Thus, 22 has two significant 
> >digits, and 22.3 has three significant digits.
> >
> >With zeroes, the situation is more complicated:
> >
> >Zeroes placed before other digits are not significant; 0.046 has two 
> >significant digits.
> >Zeroes placed between other digits are always significant; 4009 kg has four 
> >significant digits.
> >Zeroes placed after other digits but behind a decimal point are 
> >significant; 7.90 has three significant digits.
> >Zeroes at the end of a number are significant only if they are behind a 
> >decimal point as in (c). Otherwise, it is impossible to tell if they are 
> >significant. For example, in the number 8200, it is not clear if the zeroes 
> >are significant or not. The number of significant digits in 8200 is at 
> >least two, but could be three or four. To avoid uncertainty, use scientific 
> >notation to place significant zeroes behind a decimal point:
> >8.200 Â´  has four significant digits
> >8.20 Â´  has three significant digits
> >
> >8.2 Â´  has two significant digits
> >
> >Significant Digits in Multiplication, Division, Trig. functions, etc.
> >
> >In a calculation involving multiplication, division, trigonometric 
> >functions, etc., the number of significant digits in an answer should equal 
> >the least number of significant digits in any one of the numbers being 
> >multiplied, divided etc.
> >
> >Thus in evaluating sin(kx), where k = 0.097 m-1 (two significant digits) 
> >and x = 4.73 m (three significant digits), the answer should have two 
> >significant digits.
> >
> >Note that whole numbers have essentially an unlimited number of significant 
> >digits. As an example, if a hair dryer uses 1.2 kW of power, then 2 
> >identical hairdryers use 2.4 kW:
> >
> >1.2 kW {2 sig. dig.} X 2 {unlimited sig. dig.} = 2.4 kW {2 sig. dig.} 
> >"******
> >
> >My Point is this:
> >
> >I've seen many contests decided by less than 10 points on a scale of 4000 
> >which has been expanded from (at most) 2 significant digits.  As a matter 
> >of "statistics" I think that any separation of less than 100 points (two 
> >significant digits, ie,  3X00 points) is "artificial accuracy".  
> >Unfortunately, I don't have any great ideas about how to improve upon the 
> >current system, I'm just pointing out what I think is a scientifically 
> >valid problem with it.
> >
> >I smile when I see round scores posted to ten thousanths of a point on a 
> >scale that has been expanded from two significant digit accuracy to a 1000 
> >point scale.  This turns a two significant digit answer into eight 
> >significant digits!  (ie, 1234.5678)    I think that scientifically, the 
> >scores would be more accurately posted as in scientific notation at   x.x  
> >* 10 to the second power.  Most of the contests that I've been to this year 
> >have been decided essentially by random statistical "noise" rather than 
> >actual scoring decisions.
> >
> >
> >Has anyone ever thought/talked about this before ?
> >
> >Let me add, that despite what I think are statistically invalid methods, in 
> >most cases the system seems to work pretty well.  In general the superior 
> >pilots get enough better scores to overcome the "noise" but it sure would 
> >be nice to come up with a more mathematically valid solution, IMO.
> >
> >George
> >
> >
> >
> >
> >_______________________________________________
> >NSRCA-discussion mailing list
> >NSRCA-discussion at lists.nsrca.org
> >http://lists.nsrca.org/mailman/listinfo/nsrca-discussion
> 
> 