I understand that noise is useless. Garbage in Garbage out. I’m running the Fightaware filters to try to eliminate that noise, but negative aircraft count seems a little stiff to deal with my noise. Anyone else getting negative aircraft count in the “other” category?
Was it January 20th only?
That was when the UAT978 stats were added to the website.
It was on the 21 and today the 24th.
Looks like we still have a few glitches to work out…
(we don’t directly store “other”, rather we derive it from the total aircraft less aircraft categorized as ADS-B or mlat; previously each aircraft could only fall into exactly one category, but I suspect that it’s now getting thrown off by aircraft that were heard on both 1090 and 978)
I hope you look into these glitches.
I have specifically build up a feeder, site #103073, to take advantage of combining es1090 and uat978 to produce better statistics. As you can see from the site snippet bellow, most of the increases of adding the uat978 is negated by the “negative other”. The 1/27/20 stats show the first full day of operation for this site with both dongles.
I know. I’am a statistic prostitute.
Yeah, we’re looking at it; but I’d point out that the “total” value should still be correct.
(The fun thing is going to be that once the “other” value is corrected, the “total” value won’t be the sum of the rows above it. that may be confusing enough to warrant rethinking how we display this)
At the end of the day, the choice will have to be what number you want to be accurate. This could be the Mode-S and UAT, though that already leads to double counting (some AC have both after all). Or you choose to have Mode-S as always accurate and UAT only counted when the AC does not have Mode-S. MLAT may be of less interest, and could be third down the line (if no Mode-S AND no UAT then MLAT). Same for Other. Then your Total should be the sum of the 4 lines above.
One other option (which I would not favor but it looks confusing without explanation) would be have accurate Mode-S AND UAT and report UAT as 18 (2) where (2) refers to two AC that also had Mode-S.
Makes me wonder how this was handled before the recent update. Possibly some double counting for Mode-S
I’m up to -1390 on Other Aircraft Reported. Is there something I need to do? Is this affecting my ranking? I put a lot of effort into building a nice system and would like to get the best numbers.
No, it’s a server-side issue.
A short term fix that should work is to run your 978 install as a separate site.
Just a quick update, I have a fix for this in testing now.
If I interpreted this correctly, what’s happening is:
- approx 52 UAT aircraft were also reported (and double-counted) via 1090MHz TIS-B/ADS-R; in the fixed version, these aircraft are counted as UAT only.
- approx 7 Mode-S-ADS-B aircraft were also reported (and double-counted) via UAT TIS-B/ADS-R, or had dual 1090/978 out; in the fixed version, these aircraft are counted as Mode-S-ADS-B only.
- approx 2 UAT aircraft were also reported with MLAT positions (presumably they have a non-ADS-B Mode S transponder + a UAT transponder) but weren’t reported via TIS-B/ADS-R; in the fixed version, these aircraft are counted as UAT only.
- now that nothing is being double-counted, the “other” line correctly reflects aircraft where we never got any sort of position
nb: dual-equipped 1090MHz + 978MHz ADS-B out aircraft (uncommon) are counted as 1090MHz ADS-B out only under this scheme.
(the +/-1 difference in the total is, I think, just that the test system isn’t using exactly the same time window)
Longer term I’d like to break out TIS-B/ADS-R into its own row, but that is a larger change than I can do right now.
The fix is now live so you should see corrected “other” values (and some rearrangement of how aircraft are categorized, see above) from 22 Feb (UTC) onwards. Let me know if you see anything weird!
Thanks for the work you do to keep us entertained.
Now that the fix for “negative other” is working its way through 30 days of stats, how about giving us a method for displaying site ranking by ADS-B UAT aircraft on the flighware.com/adsb/stats page?
… and just removing ‘other’ from the main stats ;). How on earth can the person who’s currently second in the global stats be receiving nearly 3,000 ‘other’ aircraft/day? I’m sure that’s not possible.
The “other” field is basically meaningless at the moment. I’ve been collecting unique icaos of aircraft for a project over the last week or so. The collection samples visible aircraft once per minute and ignores any with less than 250 messages received - real aircraft are almost always going to have more messages than that and be visible for longer than a minute. Any noise/bad decodes that get through the airspy decoder filtering shouldn’t produce enough messages to be recorded.
The results I get for example, yesterday:
I recorded 3020 unique icaos (ignoring FA mlat fakes) which is exactly the ADS-B + MLAT totals recorded by FA. It hasn’t matched exactly every day, but it’s usually very close. The difference is likely to be actual aircraft seen that don’t ever receive an mlat position, and is usually less than 10 per day.
It is valid to have the field because not everywhere has sufficient mlat coverage, but it does need significantly better filtering - seeing as many “other” aircraft as mlat and ads-b combined is pretty unlikely to be a legitimate reflection of what’s going on.
That’s really good work @caius and it demonstrates it nicely.
This is the madness I’m referring to.
That’s from the 6th March. It gets better.
Oh how I laughed when I saw that. I don’t see how anyone can submit this sort of ‘data’ and actually believe they’re providing something useful.
The ‘nearby sites’ section on his profile is handy because the next highest number of aircraft/day is over 3,000 behind. There’s no way this is good, clean data.
I suppose in the long run, this doesn’t affect me so it shouldn’t bother me and really, it’s none of my business what FA do and how they display statistics, it just frustrates me to see this rubbish!
It’s quite likely that they aren’t aware - I suspect many people set up feeders and don’t pay much attention to the details of what it is actually doing. They see numbers on the leaderboard (if they even look at it) and assume that it’s correct. Relying on users to condition their own output is not going to solve the issue since most people won’t realise there’s even a problem. The filtering has to be done on FA’s side and invisible to the end user or it won’t be reliable.
I understand your dismay at the aberrant statistics at the top of the table and the apparent noise included as ‘other’ aircraft.
As I don’t receive in a crowded airspace I have no interest in the merit tables at all however I would strongly disagree that ‘other’ should be removed from the main stats as in some cases it makes up a significant and legitimate proportion of the aircraft received.
At my location about 1% of aircraft are ‘other’ and looking at the stats for similar receiving stations around me I see that my stats almost exactly align. This gives me confidence that my stats are not aberrant.
The site I installed on Cocos Keeling Island which is now run by @ypcc1 shows 20-30% as ‘other’.
I don’t have access to the raw data anymore but when I did it was clear that the ‘other’ aircraft were counted correctly and were made up of Border Force, Military, Coastal Surveillance and other agencies that didn’t want to announce themselves. His other co-located receiver is not doing so well but is still reporting essentially the same number of aircraft per day but a much higher proportion are ‘other’.
If there is a problem with the statics for a very small number of contributors then that problem should be fixed rather than remove that particular count for the tens of thousands of others that contribute correct data.