For those of you who have operated in a multi-op you know that chasing spots
is an important way to add to your mult and QSO totals. You also know how
frustrating it can be to chase down busted and erroneous spots.
During this year's ARRL SSB at NK7U contest I decided I was going to see
what I could learn about bad spots that might help us in the future. Here
are the results of my findings. (I know ARRL SSB was a while ago. This
analysis takes some time. It is quite tedious.)
As a forward it is worth describing how I identified busted spots. I started
with a full list of spots made in the contest that K1TTT captured. I then
reduced this down to just US/VE spots of DX stations as that is what is of
interest. Then I started examining the file in several ways to identify
spots that might be broken. I looked for spotted callsigns that showed up
only 1 or 2 times. The idea being that a busted callsign is not going to
appear as often as a good callsign. I am sure I missed some by limiting the
examination to only 1 or 2, but I have probably gotten most of them. If I
found a callsign this way I then would look at it to see if it was busted.
Sometimes it was obvious based on known good callsigns being actively
spotted in the contest. If not I would then look at it three ways to see if
it might be busted: 1.) Was the callsign in our NK7U log, 2.) Was it in the
supercheck partial master file, 3.) I would look it up on-line to see if it
was a good callsign. Another method I used was to look for comments in spots
that identified prior spots as being busted. There were several operators
who played the role of "spot police" during the contest noting busted spots
for others to see.
With this background here are my findings:
Overall Totals
==============
The spot database I received from K1TTT had a total of 29,485 spots in it.
Of those, 20,202 or 68% were of US/VE stations spotting DX. A total of 1,139
different callsigns made spots for an average of about 18 spots per
callsign. The ranges were pretty broad with AA3B making 992 spots, W3LPL
362, and K3LR 314 then down to 373 stations that only made one spot in the
whole contest. The top 1% of spotters made 3,527 spots or 17% of the overall
total. And, true to the old rule of thumb, the top 20% of spotters in fact
did make exactly 80.2% of the total number of spots.
Busted Spot Numbers
===================
Within the 20,202 spots I was able to identify 467 busted spots, or 2.3% of
the total. So, something like 1 in 44 spots made is busted.
Of the 1,139 operators who made spots, 896 or 79% of the total had "golden
logs" so to speak and made no busted spots. These operators made 7,695 spots
or 38% of the total. Notable operators in this space are K3CT who made 310
spots, KA4RRU who made 155, and K0RC who made 137.
The 467 busted spots were spread across 243 operators for an average of 1.9
busted spots per operator. For these operators their busted spot rate was
3.7% or 1 in 27 spots.
One thing I then looked at was if there were specific spotters who posted a
substantially high number of busted spots. If so, it might be possible to
filter their spots out of the spot stream being fed to the computers. The
idea being you could reduce the number of bad spots while not missing good
spots. What I found was a clear pattern, but it might not be that useful in
a contest.
To make this work you need to identify spotters who are both making a high
number of spots and who have a high busted spot %. There were 49 operators
who made 3 or more busted spots. Of these 21 had a busted spot rate of 5% or
higher and they represented a total of 118 busted spots or 25% of the total.
They had an average busted spot rate of 8.9% -- which is getting up pretty
high. Almost 4X the overall average busted spot rate. So, in theory, if you
filtered out spots from these 21 operators you could reduce the total number
of busted spots by 25% while reducing the overall spot count by just 6.5%.
This may or may not be that useful. It does not seem to make that big a dent
in the overall busted spot count.
What we have is one of the "long tail" situations. For example, if we wanted
to reduce the overall busted spot count by 50% we would have to filter out
spots from 51 operators who represented 31% of the total spots. Which is now
probably overkill -- throwing out too many good spots to limit the bad ones.
And of course these operators may or may not the same ones making busted
spots in the next contest.
Patterns of Busted Spots
========================
Another goal in this investigation is to see if I could identify causes of
busted spots, or at least if there were certain patterns of spots that were
busted.
It was interesting looking through the spots for unique callsigns. I found a
few operators who were posting the callsigns of stations they worked while
running. This generated a high number of uniques which of course were good
callsigns. But, the spot is more or less useless for others than except
maybe to say the band is open to some part of the world. However, most
operators know that already. And, if there are operators out there who grab
spots, QSY, and dump their call into the airwaves, these operators posting
their running QSO's are just generating QRM for themselves. And, they could
also be guilty of self-spotting.
Examining the log file found 539 unique callsigns spotted of which my
estimation was 337 were busted or 62%. So there is a pretty good chance that
a unique callsign was a busted one. Of the 38% that were good callsigns, 17%
of them appear to be generated by operators posting their own run QSO's.
Looking at callsigns that were in the log file twice, there were 140 of them
and my estimate is that 47 of them were busted or 33%. So, the chance that a
callsign that was spotted only twice is busted is still pretty high, but it
is much less than those that were unique. I did not do an in depth study
beyond this point. But it appears the trend continues. There were 89 calls
spotted three times, 60 four times, and 45 five times. My incomplete
examination showed 8 busted callsigns in the spotted three times group for a
9% rate. And, I only ran across 1 example in the spotted four and spotted
five times group -- a HZ1KEF spot for actual OZ1KEF that took a long time to
die out.
These patterns while interesting are not that useful in preparing for and
during a contest. If a new unique callsign gets spotted, it may have a
nearly 2/3 chance of being busted, but you probably can't afford to not
check it out. What is needed is a way of examining the spotted callsign in
real time to see if it has a high likelihood of being busted. I will leave
this idea for others to think about for now.
Causes of Busted Spots
======================
I did finally spend some time looking at the busted spots themselves to
investigate what causes the spot to be busted. The major buckets seem to be:
Typos -- about 25% seem to be what I would call a typo. The spotted callsign
was only 1 character off the correct call and the mistyped character was
right next to the right one on the keyboard. In absolute terms number errors
and letter errors are about the same. Though since there are far more
letters than numbers in calls, the error rate on typing numbers is much
greater than letters.
Transpositions -- about 5% of the time all the correct characters are in the
spot, but the operator transposed a couple of letters.
Busts & Miscopies -- about 65% seem to be total busts. The operator heard
and/or typed something different than the correct callsign.
Dropped last character -- about 5% of the busted spots are because the last
character of the call was missing. This was the cause of all the busted
spots from NK7U. This is a known problem with Writelog. Sometimes in doing a
"CTRL-T Enter" sequence to spot a call the program drops the last letter.
I hope found this interesting.
See you in the next contest.
Scott/K7ZO
_______________________________________________
CQ-Contest mailing list
CQ-Contest@contesting.com
http://lists.contesting.com/mailman/listinfo/cq-contest
|