Some analysis...

2009 IARU HF Championship controversy

The ARRL Award Committee's decision to end the IARU member-society headquarter station category (effectively multi-multi-multi: multi-operator, multi-transmitter, multi-location) competition in the IARU HF Championship contest is perhaps unprecedented in the history of the IARU as something it has done that amateurs around the world may not agree with judging from the excitement it has created.

The initially released results of the contest appeared to have mistakenly not counted uniques.  Uniques are call signs in a participant's log that are unique to that log - no other participant in the contest logged a contact with that call sign.  They may truly be unique, or they may be bad calls in some way not found to be bad by adjudication.  Everybody has uniques in their log, the total number depends on a number of factors.  Those factors & their influence vary between participants, so at best when the unique rate of a participant is significantly higher than others it might be an indication that further attention may be warranted.

Since these uniques are the reason for the controversy, it seems appropriate to see if they make it any easier to understand what happened, as well as try to see if there is anything else the available data might be saying but might not be seen.  Only when you understand what happened, if something were then to be done in hopes of improving matters would there be any chance of achieving the maximum optimal result.  Or in other words, if you just do something without considering what is wrong, what you do might not help at all.

To start, we look at the difference between each of the Top 10 HQ stations' average unique rates as compared against themselves & all HQs, their average points-per-QSO & the differences between their average points-per-QSO compared against themselves & all HQs.

It can be seen that DA0HQ has over four times the unique rate of the Top 10 HQs, though slightly less than average for all HQs.  DA0HQ's sub-two point-per-QSO average is not only lowest of its peers, it is the lowest of all HQs.  DA0HQ's log really stands out compared to the rest of the Top 10 HQs, who are in general reasonably closely located to each other, with SN0HQ, TM0HQ & AO8HQ having significantly above peer average uniques.  Point-per-QSO averages of the rest of the Top 10 are in 2-point-something to 3 range, with AO8HQ of course standing out in this regard due to being in AF.  Is interesting that compared to all HQs, AO8HQ has much lower than average number of uniques, though a bit higher compared to rest of the Top 10.  This could be the result of the obvious additional attraction an EA8 has due to its relative "DXness".

Overall, DA0HQ & AO8HQ are on either extreme: far more uniques, lower points-per-Q & below on both points-per-Q averages for the DLs - less uniques, far higher points-per-Q & above on both points-per-Q averages for the EAs.  This looks to be one of those situations where uniques might be suggesting there is reason to look closer.

In DL TV coverage of the DA0HQ operation a close in shot of the logging computer screen showed the 160m op was knocking them off at two per minute - admittedly six minutes after the start of the contest, but still an envious rate for early afternoon in northern hemisphere summer & just about 13% shy of the average rate for the 2700 QSOs they made on topband - the next thing that might be looked at is the distribution of QSOs by bands.

It can be seen that DA0HQ, out of a group of HQ stations located fairly close together (other than AO8HQ), worked quite a bit more (roughly double) 160m QSOs than the other Top 10 HQs except SN0HQ, OL9HQ & GB7HQ.  DA0HQ made as many (actually, a few more) QSOs on 10m than 15m, which also significantly stands out.  Likewise it can be seen 20m was not the workhorse band for DA0HQ as it was for the others.  As a result, the distribution of DA0HQ's QSOs across the bands is quite a bit more uniform than the rest.  One possible explanation for this is a higher proportion of own-country QSOs.

Another way to look at the HQs is based on the number of participants in the contest from each HQ's territory, the unique rate & the average points-per-QSO for each HQ.

Only the 30 highest are shown, which does not include all the Top 10 HQs.  Since there is a connection between IARU's own HQ entry & USA, for this chart we can consider NU1AW/KH6 to be like a second ARRL HQ entry.  The usual continental influence can be seen in the points-per-QSO averages, highest being from AF, OC & SA, with the most grouping around three.  Because their own-country QSOs are worth three or five points, we really need to see the logs of NU1AW/KH6, W1AW/KL7 & AO8HQ in order to get any idea of how many own-country Qs they made.  For the rest, the closer an HQ is to two points-per-QSO average, the more own-country QSOs there are in its log.   The scaling here was chosen for convenience, but DA0HQ & SN0HQ again stand out as having the highest unique rate of those with the lowest points-per-QSO average.  At the other extreme is 8NxHQ, who clearly works a lot of own-country QSOs in a high amateur population country, but a lot of those stations also work somebody else other than their HQ during the contest.

And finally, here is a look at the change in HQs' average points-per-QSO from the initially excluded unique QSOs compared to the average for all HQs as well as the Top 10.

Here we can clearly see the effect of unique, own-country QSOs.  The HQs' unique rate is included for reference.  DA0HQ's average points-per-QSO reduced 152 times that of all HQs when uniques are included.  SN0HQ & TM0HQ's reductions are also noticeable when viewed this way, as are their lower unique rates.  Outside the Top 12, including uniques almost universally increased the point-per-QSO average, suggesting own-country QSOs are quite a higher proportion for those at the top of the HQ game.  HQs where "home" territory contacts were with a different continent or entity (AO8HQ, NU1AW/KH6 & W1AW/KL7), do not stand out significantly from the rest when seen this way, considering for them these QSOs are worth up to five times more than for the other HQs.

Did URE bringing to light not only how ARRL adjudicates the IARU contest but how DARC might have been using this to win the HQ category really justify killing the HQ competition for everyone?  Obviously HQs will tend to attract own-country callers & some of those stations might not work anyone else in the contest.  It is almost a given HQs could have the highest unique rates of all participants.  But all participants work uniques & if uniques are apparently being accepted without question, was this ever really something only to do with HQs?

Being DX can hurt you in RDXC

Russian DX Contest results are based on confirmed contacts.  This is different than most other contests, where unless there is a clear reason (such as a bad call), claimed QSOs count for the final score regardless of whether or not both stations involved in the contact send in entries or a checklog.

Different participants in RDXC may work differing proportions of non-combatants, which may result in some participants taking more of a "hit" in adjudication than others if non-combatants have a tendency to work some participants more than others - such as those who are not-exactly-garden-variety DX.  The desirability to a non-combatant of making contact with a participant understandably increases as a function of the relative "DXness" of the participant.  Can we see any indication of this in RDXC results?

The method adopted for this analysis is to look at participants' deviation as a percentage from the average of a group of participants for the percentage of uniques, the amount of penalty points per confirmed QSO point & the difference between claimed & confirmed QSOs, multipliers & total score (relative to claimed for each).  The data used is from the 2009 RDXC single operator all band mixed mode category.

The graphs are limited to +/- 100% maximum - where the data presented exceeds that amount, the value is given above the bar.  The bars for the participants are in order of descending score from left to right.

The RDXC results show 0% uniques for a number of participants.  Uniques are call signs in a participant's log that are unique to that log - no other participant in the contest logged a contact with that call sign.  They may truly be unique, or they may be bad calls in some way not found to be bad by adjudication.  Not working any uniques in a contest is highly unlikely, so 0% might be taken to have been rounded down from <0.1% (still an enviable figure).

It should be noted from the RDXC rules a QSO does not count nor is penalized for logging errors made by the station worked, the reported times of the QSO being >3 minutes apart, the reported band that the QSO took place differs, or for a duplicate contact that does not appear in the other station's log.  A QSO is penalized three times the QSO points for a busted call, busted exchange, or if not found in the other station's log.  So a participant with a lower confirmed than claimed score (due to fewer QSOs and/or multipliers) who does not have a higher penalty applied will have lost credit for either others' mistakes (suggesting many have problems copying what he sends - not very likely, unless the operator is especially sloppy or has a call that is easily busted), or because of having worked more non-combatants.

Of the six not-garden-variety DX participants, LX1NO is a bit odd - nearly-0% uniques & a much lower than average penalty & reduction in QSOs & multipliers (and therefore score) than the top 50 entrants.  He was the only one of the not-garden-variety DX participants to operate significantly less of the contest period than the rest (only 69%).  It appears quite likely that he worked a very high proportion of stations that submitted entries, which would be expected for the amount of time he operated - a lot of potential run time missed that would have undoubtedly attracted non-combatants.

Of the remaining five participants, all but one have greater than average reductions in claimed-versus-confirmed QSOs.  Four (though not the same four) of the five have less than average penalties.  Two of the four - that do have greater than average reduction in QSOs & less than average penalties - also have less than average unique rates.  The one with a less than average reduction in claimed-versus-confirmed QSOs had a much greater than average unique rate - and one had a unique rate more than six times the average.  Note how the unique rate does not necessarily follow from a higher penalty (the participant with a higher penalty might be expected to have more bad-but-not-detected-as-bad calls in his log) or reduction in QSOs (such as a participant who works more non-combatants).

A sample size of 50 is small, but these five participants do not appear to be overly sloppy operators, so it could be that their bigger hit on confirmed QSOs might be because they worked a higher proportion of non-combatants.  An analysis involving more participants appears to be in order.

Of perhaps greater general interest (since the majority are not garden variety DX) would be how the Top 10 participants in the category compare if looked at in the same way.  Since the winner of the category is not-garden-variety DX & there is one other such station in the Top 10, this further analysis also might help firm up whether those two stations lost more QSOs because the operators were a bit sloppy (which might be seen when compared against others in the Top 10 - a group likely to be the best of operators participating in the contest).

Referenced against his Top10 peers, EA8EW's unique rate of nearly three times the average stands out, though his higher than average penalty is countered by a just above average score reduction & a bit below average reduction in both QSOs & multipliers.  It appears he might be a bit sloppy, though it did not result in a big reduction in QSOs or multipliers as compared to top scorers.  His being the exception to the expectation of not-garden-variety DX participants could suffer more from adjudication remains.

As for 5B4AII, compared to the Top 10 his unique rate is average (no bar visible as a result) & although he has an above average penalty, it is not by much.  He does not appear to be a sloppy operator & the reduction in claimed-to-confirmed QSOs could very well be because of working more non-combatants due to being not-garden-variety DX.  How much his call sign might contribute to others making mistakes would be interesting to know.

Overall it appears that not-garden-variety DX is at a bit of disadvantage in RDXC because of adjudication.

Worthy of mention are LY9A, ES5TV, SP4Z & YP9W, all with lower than average penalties & reductions.  YT1AD's lower than average unique rate rather goes against his higher than average penalty, suggesting logging errors may be involved.  Also not having the best of weekends were UU4JMG & 4O3A.

Finally, the Top 10 referenced to the Top 50 shows most performing better than average (as would be expected).  YT1AD's unique rate is well below average & penalty is just below average, but reduction in QSOs & multipliers (and therefore, score) remains well above average.

2010-06-24