[Skimmertalk] FW: Skimmer Tests during SS

[Jim Baremore]

Hi Rick

Thanks for taking the time to read my very long experiment.  I appreciate
your comments and questions.

I shortened your response to several key points I wanted to address but I
hope I have not taken anything out of context.

Yes,  I used 1.2 and only when I noticed the Skimmer:DXSpots page was
showing S/N ratio and Speed did I tumble to the new version being out.

I wonder about that also.  I had a query if I had the "CQ'ing only box
checked."  I was not sure exactly what algorithm Alex is using if that box
is checked.  Pete just sent a note earlier showing some of the cases where
it might not spot the running station.  I commented on that in my recent
reply to his post.

The accurate decoding would seem to fit my numbers.  Note in the beginning I
calculated that of the 12,679 apparent US call signs, 12,255 were actually
call signs currently valid and issued per the FCC data base.  Most of the DX
reported was likely broken calls and so overall I estimated CW Skimmer
copied 94% accurately while outputting a bit over 1000 calls an hour to the
Linux Processing box.  The Linux box checked every call against the FCC data
base and only currently valid US licensed calls were then forwarded to the
logging program by the Linux Telnet Server.  The 30% I refer to in my report
means that 30% of the time, the spot I clicked on found that station on the
frequency.  Most of the time I clicked on a spot that had just left the run
frequency and the running station was the one that was still there.  Both
accurate calls.  In summary, my accuracy observations of 94% are actually
higher than your friends observations of 80% to 90%.  My 'being the station
on frequency' percentage is a different calculation.

Regarding spots, your position is much like mine.  CW Skimmer by itself just
provides spots.  To become a useful system, it needs a lot more software to
be built.  Typically that will become contest specific.  In a DX contest,
the callsign tells you a lot.  In SS, mostly nothing and even if you have
the state, some of those have multiple section multipliers as you mention.
If you will, some of the intelligence needed by CW Skimmer is already being
used by virtue of the logging programs that look at prior contest data to
prefill in the reports.  

I also support technology.  As I commented in my report, the young people we
want to attract to our hobby are technology based and barriers to technology
advances, perceived or real, will only stagnate the hobby. 

Thanks for your observations.  I have played with it in several contests now
and since I chose not to enter this year's SS, I decided maybe I would try
to gather some data to start the basis for discussions on this group.  There
have been several anecdotal reports and a constant request for data.  Your
observations fall in line with mine.  CW Skimmer is neat, but it is not the
killer application that many thought when it was first introduced.  It is a
real Q improver but so are the fancy software loggers, so are the SO2R
setups and perhaps other things I might describe.  But this is beginning to
sound like my soapbox so I'll stop.

Thanks again for your comments Rick and good luck in the contests.


73

Jim  K5QQ



-----Original Message-----


I'm assuming that you performed your analysis using version 1.2.  I wasn't
aware that version 1.3 had come out until after the SS had concluded.  


I wonder if the SS exchange gave the decoding algorithm some
type of unexpected trouble.

When the contest concluded, he commented to me that in his opinion my
skimmers were producing correctly decoded call signs at an 80-90 percent
success rate.  Interesting that there could be such a difference in accuracy
from contest-to-contest.

I also agree that the lack of meaningful multiplier information was a bit of
a disappointment.  I feed the output of my six skimmers into a local DX
Spider cluster, which I configured to show the US state of the spotted
station in the "Remarks" field.  The hope was that Win-Test would pick up on
this and flag some of the spots as multipliers.  I obviously realize this
wouldn't have helped to identify all possible mults (such as those in CA,
FL, MA, NJ, PA, TX, and WA), but I was interested in seeing what the results
might be.  Unfortunately, that portion of my experiment was a complete bust.

I am a big supporter of this technology and use it on a daily basis.  

I say all that as a prelude to my statement that the competative edge we
derived from skimmer in our 2008 SS CW multi-op effort at K4TD was
essentially non-existant.  

73,

Rick
K4TD