Sweepstakes Log Checking

[km9p at aol.com]

In this year's Sweepstakes I lost about 20 QSO's on CW due to various
mis-copied information.  Of these 20 QSO's KR0Y lost 7 of the exact same
calls for the exact same reason.  

I find it hard to believe that the two of us are copying the same station as
NH, when we lost it because it was supposed to be EMA.  

Now, one of two things are happening...

#1:  The guy sent in his check sheet with the incorrect information or sent
the incorrect information during the contest.

#2:  The ARRL typed in the information incorrectly.  

One might argue that these types of mistakes will be evenly distributed among
participants.  Three years ago I analyzed four large logs for Sweepstakes CW
and found that of these four logs, about 10% (if memory serves) of the QSO's
were found to be unique in each log as compared with the other three.  Very
few were found to be busted calls.  Based on my study, I think that even
distribution of these mistakes is something that should not be considered
guaranteed or assumed.

This year on SSB, Rich & I were seperated by 3 QSO's after working 2400 Q's. 
SS in the past has had some extremely close finishes (NIN & RZ seperated by 1
QSO).  With this in mind, I'm not sure it is fair to the competitors to let
typo errors and their assumed even distribution potentially decide the
outcome of the contest.  

I would propose that the ARRL adopt a check in their software to flag
potential problems.  It would be immediately obvious if the computer found
that W4NT was taken out of 250 logs because they all copied AL instead of GA.
 

By the way, this is by no means a slam of the ARRL's log checking job.  I
think they do a great job.  I just think there is room for improvement in
this area.

73

Bill Fisher, KM9P


>From k2mm at MasPar.COM (John Zapisek)  Thu May  5 00:07:49 1994
From: k2mm at MasPar.COM (John Zapisek) (John Zapisek)
Date: Wed, 4 May 94 16:07:49 PDT
Subject: Sweepstakes Log Checking
Message-ID: <9405042307.AA07987 at greylock.local>

Bill/KM9P noted a problem with the ARRL's SS log checking:

> one of two things are happening...
> 
> #1:  The guy sent in his check sheet with the incorrect information
>      or sent the incorrect information during the contest.
> 
> #2:  The ARRL typed in the information incorrectly.  

The Internet SprINT that runs quarterly here on cq-contest addresses both
these problems.

Problem #2 is trivial.  Logs are e-mailed, so hardly any typing is involved.

Problem #1 is also addressed (though not _solved_).  Busted QSOs are
deducted from both the TX and RX stations' scores.  This provides strong
incentive for an operator to accurately log exactly what's been sent.

Also, the small total number of busted Qs makes it possible to review them
all in search of systematic errors.  And the availability of your graded log
via return e-mail is an effective vaccine against log-checker complacency
disease ;-)

Look for Tree/N6TR's announcement of the next Internet SprINT, probably
sometime in July.  73.  --John/K2MM

>From Robert A. Wilson" <n6tv at VNET.IBM.COM  Thu May  5 03:34:20 1994
From: Robert A. Wilson" <n6tv at VNET.IBM.COM (Robert A. Wilson)
Date: Wed, 4 May 94 19:34:20 PDT
Subject: Sweepstakes Log Checking

> I find it hard to believe that the two of us are copying the same station as
> NH, when we lost it because it was supposed to be EMA.

I came to the same conclusion after analyzing my errors in the 1992
SS.  Of the 20-odd QSOs that I lost, I verified using WB5VZL's (?) log
database that roughly half of those "mistakes" were in fact NOT
mistakes; the data I copied was good, but the data entered from the
other guy's summary sheet, or the summary sheet itself, was bad.

I believe the log checkers need to eyeball all the computer
generated error reports BEFORE they decide to reduce a score.  If 90%
are getting busted for the same error (say, copying W6XYZ's check as
55), then NO ONE should lose for copying check 55, no matter what it
says on W6XYZ's summary sheet.  He obviously made a mistake.
This is just common sense.

Based on the few data points that I have, I conclude that this "sanity
check" was not done for either the 1992 or 1993 SS.  One might think
that since mistakes like this are applied equally to everyone's log,
it makes no difference.  But it does!

The reason is this:  suppose 9 guys get busted for copying "W6XYZ 55."
The guy who copies "W6XYY 55" will NOT lose any QSOs.  Why not?  Well,
W6XYY is not in the database, and the log checking program
currently being used by The League is not sophisticated enough to
recognize busted callsigns, only busted CK, PREC, and SECTION.  The
guy who miscopies 1 out of 10 callsigns will get penalized LESS, on
average, then the guy who miscopies 1 out of 100 callsigns, because
the poor operator will have far fewer chances of being "caught."  For
tight finishes, this is unacceptable.

Software does exist that can do a better job of log checking.  So I
ask KR1R, if K2MM is willing to modify his log checking software to
handle SS logs, and if he is willing to give it to you for FREE, would
you consider trying it out for the 1994 SS?  Please?

73,
Bob, N6TV

>From blunt at arrl.org (Billy Lunt KR1R)  Thu May  5 14:28:08 1994
From: blunt at arrl.org (Billy Lunt KR1R) (Billy Lunt KR1R)
Date: Thu, 05 May 94 08:28:08 EST
Subject: Sweepstakes Log Checking
Message-ID: <17213 at bl>

                       Log Checking
                          in the
                    ARRL Contest Branch 

Questions have been raised about accuracy in the cross-checking 
procedures of ARRL contest participants. We are concerned about 
this perception on the part of some members.

We are committed to treating all contest participants fairly. We 
want our members to know that we do our best to treat them with 
the respect and fair play that they deserve. 

PROCEDURE

The ARRL Contest Branch uses two different computer programs for 
checking contest logs. One (Check) is for paper logs and the 
other (Cross) is for disk entries. Both of these programs perform 
the same functions. The difference between them is the way that 
data is entered into the programs. In our "Check" program, the 
data is keystroked in manually, and our "Cross" program uses data 
files stored on disks.

Simply stated, our procedure is to start checking those contest 
entries with the highest claimed score (in each category), and 
work our way to the lowest score. We do this until we run out of 
time (limited by QST publication deadlines). 

Because we use computers, errors are more easily found than by 
the old hand checking method. However, the major benefit is time 
efficiency. The computer is quicker than  hand checking, giving 
us more time to check logs both paper and disk logs.

As for the 1993 ARRL November Sweepstakes, the two-man Contest 
Branch spent approximately 100 man hours, cross checking both 
paper and disk entries. We wish that we could have checked more, 
but our major enemy, time, was the limiting factor. 

DATABASES

The ARRL Contest Branch does not use an ongoing database of call 
signs or exchange information. We only use data from the entries 
for the particular contest we are checking. (Nor is data carried 
on from year to year.) We do not care to build a "master 
database" because data is constantly changing from one contest to 
the next and such a database would become dated. 

SCORE REDUCTIONS

In the League's log checking software, before QSO credits are 
taken away for any reason--and claimed scores are reduced, there 
is human intervention. The program requires us to manually check 
any QSO before credit is removed. We have the option of taking 
away or giving back credit for any QSO the computer flags. This 
prohibits computers from reducing scores automatically and 
unchecked by human supervision. This was one of our major 
concerns when the software was written, so safeguards were 
inserted into the programs to prohibit such things. This also 
allows us to double check for inconsistencies or recurring 
patterns. If an inconsistency or recurring pattern develops, we 
give the entrant, who's log is being checked, the benifit of the 
doubt and do not remove QSO or multiplier credit.

PERFECT?

We have missed logs that probably should have been checked. 
That's a result of time constraints placed upon us by QST 
publication deadlines or by an occasional miscalculation of man-
hours needed to complete certain tasks. We think that this is the 
exception rather than the rule. We take pride in striving to be 
the best. 

Besides QST publication deadlines, a number of other factors 
limit time that we can allocate for cross checking contest logs. 
Our year is divided into 12 months (12 issues of QST). Within 
these 12 issues, we report on 16 major contests. (Each one being 
as important as the others.) Our time is divided between:

  o Cross checking logs for the contest being reported on in that 
    particular month (four of the months have two contests to report 
    on the results). 
  o Keystroking of all the data that is used for generating the 
    contest score listings and winner boxes.
  o Analyzing scores for records.
  o Pulling soapbox comments from the logs.
  o Choosing photos for the article and writing photo captions.
  o Writing lead material for the results.
  o Processing incoming entries for different contests.
  o Answering written correspondence.
  o Answering e-mail correspondence.
  o Answering telephone inquiries. 
  o Printing, applying stickers, stuffing, addressing and mailing 
    of contest certificates and plaques.
  o Compiling data for the Special Events and Contest Corral 
    monthly columns for QST.
  o Various other day to day tasks within the Contest Branch. 
  
ARRL SOFTWARE ENHANCEMENT

With some recent developments and some improvements in progress, 
we hope to have the most thorough and accurate contest log-
checking software available. This new software should help the 
Contest Branch meet its continuing mission to treat all contest 
entries (both disk and paper entries) equally and fairly.