Contesting - Deterministic Methods
Richard (G3CWI)
on
March 22, 2004
View comments about this article!
Modern contesting software offers the user a huge amount of information both on the user's own performance and, through bandmapping, on what others are doing. I wonder if it might be possible to harness this information to allow contest software to suggest tactics in real-time? Here in the UK we have a series of contest running that may be suitable for such an approach. They operate over a limited band and only last for 90 minutes; perhaps a standard queueing theory model Erlang B or C might be adapted to suggest if "run" or S&P is most appropriate at any moment in time. These UK contest take place three out of four weeks each month so an adaptive algorithm could be devised that learnt each time from your performance.
Pie-in-the-sky or is someone out there already doing this?
73
Richard
G3CWI
|
Contesting - Deterministic Methods
|
Reply
|
|
by VE5ZX on March 26, 2004
|
Mail this to a friend!
|
Richard
You have asked an interesting question! However, I suspect that contesting involves too many non-deterministic processes to make such predictions reliable. For example, whether an operator should be in run or S&P mode depends on what mode the majority of other contestants in the contest are operating in. If most contestants are in run mode then your optimum mode would likely be S&P and vice versa. A couple of years ago I wrote a short article called "A Novel Perspective of Amateur Radio Contesting" available from http://www.sussex.ac.uk/Users/sylvank/VE5ZX/Contest.pdf. It discusses some of the non-deterministic and hence relatively unpredictable processes in an amateur radio contest that might make it difficult for a contestant to gain much information from real time software analysis techniques.
|
| |
|
RE: Contesting - Deterministic Methods
|
Reply
|
|
by AB5XZ on March 29, 2004
|
Mail this to a friend!
|
Sylvan,
Thanks for pointing us to your paper. Interesting, to say the least!
In last weekend's CQ WPX SSB contest, I watched my rate pretty carefully, and I found, late in the contest, that I was getting about 10 contacts per band (10, 15, 20) as I swept through the bands. So my yield was diminishing in S&P mode.
I picked a quiet spot on 15M, at the low end of the phone portion, and started calling CQ (very unusual for me in a contest). The results were amazing. My rate went above 60/hr for the rest of the contest. Most of the contacts had serial numbers in the 30's, so these were very casual participants. A few had serial numbers in the 300's, and one had a serial number over 1000. There were only two instances of any "piling up", when two stations called at once.
With this shift in strategy, I was able to easily meet my goal of 300 contacts.
As you mentioned, there are many non-deterministic factors in the contesting system, and each contest has its own dynamics.
There's a new booklet on tips from top DXers that may give some more insight into these complex systems.
73TomAB5XZ
|
| |
|
RE: Contesting - Deterministic Methods
|
Reply
|
|
by kb3kaq on March 29, 2004
|
Mail this to a friend!
|
software like N1MM provides the ability to monitor log activity via the internet. this opens a host of possibilities if the majority of contesters move to this type of software. having a server with the real-time logs would enable a prediction software to calculate where the progation paths are working best and also be able to manage band load.
the management of the band load would allow more room for operators at a given time. here's the scenario as i see it: the contest starts and in the first 30 minutes, a window opens that suggests a tactic - S&P or RUN. if you select RUN, it suggests a set of frequencies based on the last 30 minutes of propagtion. if you select to RUN on one of these frequencies, the server assigns you to this and based on the last 30 minutes of propagation, shifts that frequency suggestion to an area that is not in your passband.
2 hours into your RUN, a window opens that suggests you move to S&P on 20m from your run on 15m and it also suggests a beam heading to maximize your S&P. this is possible, you just need to have enough data input into the system and enough servers to share the load of the incoming data.
if you had selected S&P, it would suggest a band and beam heading and those mults and stations that were in that beam headings workable path would be different colors on the bandmap than those not in the propagtion path.
not sure how you would allow for workinf off the back of the beam, but i'm sure it could be figured out... maybe thru a several contest learning mode. once those stations on the current band are exhausted, it might suggest another band or new beam heading.
if a frequency in your propagation path opens and the time is right to RUN, the system would suggest the change.
the problem - not enough data from the many stations not wired to the internet running the software - maybe packet could be used to move the data, but i don't think it has the bandwidth for the potentially high volume of incoming data - the lag time would be too long to make real-time possible.
as the contest progresses, the software guides your decesions, but you are given a choice at all times. if you want to RUN the entire contest on 7230, have at it. the software will work around you. at some point it the system will breakdown, but i'm sure the number of ops not supplying data would be more of a factor to overcome than ones supplying data and not following the suggestions.
the ultimate in push button contesting ;)
-kb3kaq
|
| |
|
Contesting - Deterministic Methods
|
Reply
|
|
by vk2cz on March 29, 2004
|
Mail this to a friend!
|
Using erlang tables (or spreadsheet formulae) could assist in determining how many QSO's are possible in saythe 40m window during a nightime window. The sprints would translate to the erlang non-blocking, whereas stations parked on a freq calling would be blocking. Not sure where it would go from here - just presenting an observation.
Back to being deterministic, I have enough historical data for WPX SSB and CQWW SSB showing 'what band @ what time' actually worked for me from here in VK2 and VK8.. granted conditions (and my antennas) vary, but given the data is for the same weekend each year, it has been useful to a limited degree - and in 2004, it will affect my antenna choices for the CQWW SSB effort later this year.
|
| |
|
Email Subscription
You are not subscribed to discussions on this article.
Subscribe!
My Subscriptions
Subscriptions Help
Other Misc Articles
DXLog Contest Software
|
