Improved Routing Information

[tarbat]

Of 1239 ADS-B-equipped scheduled LHR flights automatically detected by the EGLLADSB website, 125 (10.1%) were listed by RadarBox with either no route at all, or with a route that didn't have EGLL as a From/To/Via.

Seems quite consistent results day-to-day.  Radarbox manages around 90% success on route lookup, with a 10% failure rate (for ADS-B aircraft).  Dave, can you do a similar test for non-ADS-B aircraft?

In my opinion, getting correct route information for non-ADS-B aircraft is MUCH more important, to help "visualise" where the aircraft is.  When you've got 5 Loganair aircraft in the list without route data, it makes it impossible to "guess" where they're flying from/to.

[DaveReid]

Seems quite consistent results day-to-day.  Radarbox manages around 90% success on route lookup, with a 10% failure rate (for ADS-B aircraft).  Dave, can you do a similar test for non-ADS-B aircraft?

Interesting idea.

I'll give it a go, although it's not a check that's easy to automate since, by definition, the EGLLADSB website doesn't list non-ADS-B movements (although I do have a complete list of Heathrow Flight IDs which I can use to identify the non-plotters).

If we believe what AirNav tell us, then we should be seeing more routes for ADS-B than for non-ADS-B, since obviously their algorithm won't work for the latter.

Having said that, there is no evidence of improvement in the Heathrow stats to suggest that the algorithm is working for any aircraft.

Watch this space.

[bearcat]

I have BEE8ET as EGFF - EGPH

[Runway 31]

Cheers Bearcat.

Alan

[bratters]

Of 1239 ADS-B-equipped scheduled LHR flights automatically detected by the EGLLADSB website, 125 (10.1%) were listed by RadarBox with either no route at all, or with a route that didn't have EGLL as a From/To/Via.

Seems quite consistent results day-to-day.  Radarbox manages around 90% success on route lookup, with a 10% failure rate (for ADS-B aircraft).  Dave, can you do a similar test for non-ADS-B aircraft?

In my opinion, getting correct route information for non-ADS-B aircraft is MUCH more important, to help "visualise" where the aircraft is.  When you've got 5 Loganair aircraft in the list without route data, it makes it impossible to "guess" where they're flying from/to.

Point of order here, Tarbat. Your statement  "Radarbox manages around 90% success on route lookup, with a 10% failure rate (for ADS-B aircraft)" is based on figures obtained solely at EGLL.

I'll bet my bottom dollar that a similar exercise at Stansted, Birmingham, Manchester etc would produce significantly different results because of the incidence of lo-cost carriers. Certainly based on my observations here in the centre of England a 10% failure rate is wishful thinking.

[tarbat]

Point of order here, Tarbat. Your statement  "Radarbox manages around 90% success on route lookup, with a 10% failure rate (for ADS-B aircraft)" is based on figures obtained solely at EGLL.

Agreed.  My comment was simply an observation on Dave's results.  Success rate at Inverness also feels like a lot less than 90%, and I'm about to analyse the results from yesterday.

Yesterday's log at http://www.tarbat.gofreeserve.com/data.htm

[DaveReid]

I'll bet my bottom dollar that a similar exercise at Stansted, Birmingham, Manchester etc would produce significantly different results because of the incidence of lo-cost carriers. Certainly based on my observations here in the centre of England a 10% failure rate is wishful thinking.

I agree too.

Heathrow is just a handy benchmark, because I know my box picks up 100% of movements and it's easy to run the analysis daily.

That's why I resisted AirNav's recent repeated requests (6 at the last count) for me to add routes to my EGLLADSB website, thereby enabling them to hoover them all up, add them to the database and then claim that all was well in Routeland.

[tarbat]

Agreed.  My comment was simply an observation on Dave's results.  Success rate at Inverness also feels like a lot less than 90%, and I'm about to analyse the results from yesterday.

Yesterday's log at http://www.tarbat.gofreeserve.com/data.htm

Out of 501 flights yesterday, 329 have a route listed (65%).  I don't know what percentage of these have the correct route, and some of them you would not expect to see a route listed (eg TARTN51).  And also note, I have added a lot of routes (eg BEE*) manually to my database, so the results would have been worse than this without the manual updates.

EDIT: To put this in perspective, I ran FD7.5 for some of the day, clicking on each aircraft to force a route lookup in FD7.5, and it found a route for 246 out of 305 flights, which is better at 80%.

[DaveReid]

Watch this space.

OK, results are as follows:

Of the 82 non-ADS-B LHR movements yesterday (identified by callsign), 6 (7.3%) had no route showing.

Based on such a small sample size, I'm not sure it's possible to draw any conclusions, although it may be significant that more than half of the 82 movements were BA or BMI, and three-quarters of those were non-alphanumerics, i.e. timetabled flight numbers.

But there's certainly no evidence to suggest that the algorithm is leading to a higher proportion of routes being identified for plotting flights.

Did someone mention snake oil ?

[John Racars]

Attached CSV file suitable for import to your routes table.

Hi Mark / Hi Tarbat,

thank you both verry much for all this! I imported the csv-data succesfully into my "NavData.db3" this morning.

[DaveReid]

Results for yesterday (26th September):

Of 1222 ADS-B-equipped scheduled LHR flights automatically detected by the EGLLADSB website, 123 (10.1%) were listed by RadarBox with either no route at all, or with a route that didn't have EGLL as a From/To/Via.

New carriers added to the list of those operating unknown/incorrect routes: EVA, FIN, SVA.

[bratters]

Results for yesterday (26th September):

Of 1222 ADS-B-equipped scheduled LHR flights automatically detected by the EGLLADSB website, 123 (10.1%) were listed by RadarBox with either no route at all, or with a route that didn't have EGLL as a From/To/Via.

New carriers added to the list of those operating unknown/incorrect routes: EVA, FIN, SVA.

Good morning Dave -  I welcome your stats for EGLL and think they would have even more relevance were you to actually split the figures between "no routes" and "no EGLL"s.

We could then say X% were in fact decoded however an absolute minimum of Y% of all decodes are errors. This might give us some indication of the respective proportions of the two different elements of the problem: A/ getting routings decoded and B/ getting routings right.

What do you say?

John

[DaveReid]

We could then say X% were in fact decoded however an absolute minimum of Y% of all decodes are errors. This might give us some indication of the respective proportions of the two different elements of the problem: A/ getting routings decoded and B/ getting routings right.

Happy to oblige.

Of those 123 mis/unidentified flights, 109 had no route at all and 14 showed routes that didn't involve EGLL.

There's no excuse at all for the latter ones (how can UAL948 be KLAX-KDEN when the network sees it landing at Heathrow every day?) and, of the flights with blank routes, many are from/to other airports which also have good network coverage, so they ought to be easily identifiable too.

It's now a month since I ran the first LHR analysis and the failure rate hasn't improved at all since then.

[bratters]

Thanks for that Dave.

As mentioned before, the figures for EGLL are not representative of the country as a whole but they might well represent the best performance and results that Airnav can put up in the UK.

In fact from an Airnav perspective, they’re not too bad at all. We seem to have 91.1% of all flights being decoded and, of these, up to 98.8% might be accurate.

Looking at the two figures separately, over 90% decoded would probably satisfy most of us were it a universal figure. However we know that this figure bears no relation to the other UK airports where decode rates will be considerably lower.

Looking at erroneous routings, a 1.15% failure rate is in fact close to Airnav's own claimed accuracy figure, however as this figure measures only the incidence of the "EGLL" route error, the overall percentage of route errors will probably be higher.

Much of EGLL traffic being scheduled, presumably the same errors are being perpetuated. Your "UAL" exemplifies this but one imagines that this and similar instances could be easily and speedily cleared up. Easily & speedily.............

Do we know the answer to the “UAL948” question?
Of course we do - it’s LAX/DEN/LHR.
So let's immediately correct that entry in our databases and all will be well.
Er.......

I guess your 10% error rate at EGLL is going to remain just that.

[DaveReid]

As mentioned before, the figures for EGLL are not representative of the country as a whole but they might well represent the best performance and results that Airnav can put up in the UK.

Yes, in fact I would expect Heathrow to be better than pretty well any other airport covered by the AirNav network, given that it's surrounded by sharers and a lot of the flights are international ones that FlightStats should have.

Or, to put it the other way round, routes to/from other airports are likely to produce results that are even more adrift from AirNav's claims.

I guess your 10% error rate at EGLL is going to remain just that.

Yes and no. 

Between now and the end of October I don't expect to see much change.

But if AirNav haven't got their solution working in time for the Winter schedule changeover, I expect that the hit rate at Heathrow will plummet.  There will be more flight numbers showing incorrect routes, because the airlines have reassigned them and the algorithm that should pick that up isn't working.  Likewise new flight numbers, even if they are between Heathrow and other airports with good network coverage, will stay blank no matter how many times the network logs those Flight IDs taking off and landing.

Little did I suspect, when I innocently suggested using network movements to deduce routes, that the implementation would turn out to be such a dog's breakfast ...