small, hour anomaly

Peter has noticed a minor problem with the way hours are calculated by current firmware. (not sure about Real Speed)
They can't start the calculation at 0kts, most of the time this won't be a problem, especially if there's no minimum speed set anywhere, (GPS unit or evaluating software).

Here's an example of where it can be a problem.

This hour starts when there is a speed recorded and gives a result of 9.588, including the big gap near the beginning and ignoring the smaller gap at the end.

And here's where the hour should be. It starts at 0 and includes the smaller gap at the end



This gives a result of 10.284.

Unfortunately GPSAR seems to have a min speed of 5kts permanently set.
GPSResults has a 5kt min speed by default in the filter settings. For the best hr and distance results, change this to 0.
I'm fairly sure KA72 has no min speed filter, but still needs to start the calculation above 0kts.

For more detail, here's Peter's blog
boardsurfr.blogspot.com/2019/04/hour-analysis-for-gpstc.html

Also, I had a look at Peter's blog post, and he contends that there is a problem with this file because it has large regions of missing data (due to the min speed setting on the gps device probably being set to a non-zero value.)

He gets a different "best hour" result by adding interpolated points to the file with 0 speeds to fill in the gaps.

I'm wary of an approach that involves adding more data to a file than was originally there, even if that data is 0 speed points. You are making decisions about data that may not be valid, in my opinion. You may infer that the speed for these points is zero, but what about the coordinates? There are lots of shaky assumptions that can be made by adding interpolations to the data.

ka72.com does not add fake data to the tracks, so if data is missing it can't be used to calculate divisions based on time intervals like this. However, if there is zero speed data point in the file, ka72.com will use it.

Best approach is, as Peter said:
1. Make sure any min-speed recording setting in your GPS is disabled or set to 0 kts.
2. Make sure to adjust min speed settings in your software (GPSResults) to 0 or use software that already assumes that (ka72.com)

Select to expand quote

Dylan72 said..
I'm wary of an approach that involves adding more data to a file than was originally there, even if that data is 0 speed points. You are making decisions about data that may not be valid, in my opinion. You may infer that the speed for these points is zero, but what about the coordinates? There are lots of shaky assumptions that can be made by adding interpolations to the data.

For the GPS Team Challenge, all speed calculations should be based on doppler speeds. Adding points with speed 0 will not change anything about these calculations. Even right now, ka72.com will calculate 1-hour results that start at a valid point and end in a region with missing points. Mathematically, this is exactly the same as adding the missing points. You are ending the calculation at what you call a "fake point".

Select to expand quote

Dylan72 said..
Best approach is, as Peter said:
1. Make sure any min-speed recording setting in your GPS is disabled or set to 0 kts.
2. Make sure to adjust min speed settings in your software (GPSResults) to 0 or use software that already assumes that (ka72.com)

I suggested this approach to get around the bugs that are present in GPSResults, GPSAR, and ka72.com.
The actual "best approach" would be if the bugs would be fixed in these programs. I have described a rather straightforward way to do this in my blog post, and provided links to data that can be used for validation. At you need to do to find the correct hour in tracks with missing points is to look backwards, too, not just forward.

Note that with the current state of the analysis software (including ka72.com), the problem can occur even is the minimum speed is set to zero. For example, say you have a session with lousy wind where you mostly slog around for a couple of hours. Then you stop to take a break and turn off the GPS. After 30 minutes, the wind finally comes up, and you go out again for a 45 minute session before you have to stop. When you upload the file ka72.com or look at it in GPSAR/GPSResults, you'll have a 3+ hour session, but none of the programs will correctly calculate the 1-hour result for the last 45 minutes plus the preceding 15 minutes of zero speed.

There are other things that can and do cause missing data, even with minimum speed settings of 0 and without turning the watch off. That includes crashes, swimming after your gear, or cleaning weeds of your fin under water - anything were the GPS looses reception. The GW60 does not log points when it looses satellite reception under water. All that can lead to incorrect 1-hour calculations.

Select to expand quote

Dylan72 said..
Best approach for improving your 1hr score is actually to sail for 1 hour, not sit on the beach with your GPS turned off>>>>

Well that's rather obvious, and not what we are talking about.

The point is that we think you should be able to get a 1 hr average, even if you have not sailed for the whole hour. Of course it will always be slower than if you were able to sail for the full hour, so it is no advantage. It seems that some software will do this now, but some won't.

EG: You go out and sail hard for 50 minutes and then the wind suddenly dies. If you take the distance you sailed and average that over an hour, you would still get a reasonable 1 hr to reward you for the attempt. Is this not reasonable?

Of course, if you just leave your GPS going while you de-rig, and it ends up with another 10 minutes of very slow data, the software programs will still give you a result. This is certainly what happens quite regularly. But if you can't do that becasue yo have to get in the car and go RIGHT NOW, to get the kids from school or meet your partner for their birthday lunch, is that fair?

Bob, this is a TEAM challenge remember, it's not just about your hour.

It could become an issue if your only mate gets a good hour, and you need your not so good hour to gain points for the team.

It comes down to the definition of the hour, is it just your average speed for an hour, or does it have to be a full hour sailed.

The problem with it having to be an hour sailed, as Andrew says, is some programs include the time left on while derigging and some don't.

If you just rely on KA72 to post the numbers, you have no idea if you actually sailed the full hour or not.
Who is going to check everybody's hours to make sure every bit of it was spent sailing? Not me.

Because the hour was never defined in the GTC rules, we all seem to have a different assumption of what it is.

The majority of the GTC advisory group are in favour of just the hour average, regardless of how long sailed.

And it's easy to work out, if the software gives a 0 hour.
If you've sailed under an hour, your distance in nautical miles is the hour average in knots. (but I'm inclined to agree with Bob here), Two short sessions before and after lunch can't be joined together.

Select to expand quote

Swindy said..
I can see there is an issue, but to me seems a bit irrelevant as sessions like these will rarely give you a pb or score for your team.
The hour is the toughest category we sail for, not only physically. Getting the location, gear and consistent conditions for an attempt can be hard. I have given up on attempts on numerous occasions and for many reasons but when you get it right its so rewarding.
So im with Dylan and Bob, sail for an hour.
Looking forward to seeing you running Bob.
Edit and engage brain, distance is probably tougher.

+1 when sailing for a solid 1hr plus normally this is where my best results are, unfortunately I usually cut it short by 5 or 10 minutes but the GPS is on so it gives me a not so good hour. I Got to stop stopping and gasbagging then might get something decent. If you look at my last Coods trip you'll see it's about a 21 hour but you'll also notice the last 5 to 8 was a stop start probably gas bagging to swindy, I back calculated over the proper minutes I sailed and it averaged around 23, so moral to the story less gasbagging and sail the hour to get a good average??

A little reality check, looking at the ranking from March: the team that ranked #21 of 42 had a 1-hour average of 16.4 knots. The next 21 teams had even lower 1-hour averages. So at least half of the hours that earned a jelly bean in March did not include an hour of planing.

I would expect GPS analysis software to pick the 1-hour period that gives the highest average. Does anyone really disagree with that?

It turns out that the current software does not do that. Boz, for example, has used a minimum speed setting for his GPS. That's not unreasonable, and even was recommended in some setup guides in the past.

As Boz said, he usually takes breaks. Of course he should not, but he does, and the majority of speedsurfers does, or simply does not get one hour of reliable wind.

If this break falls somewhere in the middle of an hour, the analysis software will score the hour. Here is an example:


In this example, GPSAR picks a region with a lot of missing points in the middle, even though picking a region towards the end would clearly give a much higher average. If the track just has a few points a couple of minutes later, the reported 1-hour result increases from 8 to 14.7 knots:

It is rather obvious that the first result is dead wrong. GPSAR did not pick the hour period with the highest 1-hour average - it did not even get close. Uploading the track to ka72.com (www.ka72.com/Track/t/382529) gave a very similar result:

Again, the 1-hour result is more than 6 knots too low.

Not a bug? Really?

I am a little puzzled by this. I would have thought this would be the easiest division to calculate. Just largest delta distance over 3600 seconds irrespective of how many points are recorded.

Select to expand quote

decrepit said..
Trouble is David, ages ago it was decided that the hour should start and finish with a valid point, but the middle doesn't matter. So you can have a period of inactivity in the middle but you can't start or finish there.

I don't know what was decided ages ago, but the current algorithms all allow hours to end on invalid or missing points (as long as there is another point recorded at some point in time afterwards). They just don't allow them to start at an invalid point. Having different rules for the start and end makes no sense.

Select to expand quote

decrepit said..

firiebob said..
why anyone would want to join sets of minutes (15+45).

So Bob your place or ours for the coffee and cake???

It's a date, I'll give you a yell after weekend