I have released a new version of GPS Speedreader that fixes a few bugs (thanks for all reports!) and introduces a couple of new features. This version will also show polar plots for data that do not have doppler headings (e.g. watch data), using positional data instead.
Two new features are "Sorted speed" graphs and detailed turn (jibe and tack) analysis. The turn analysis is still a bit buggy, but may be useful for anyone working on improving jibes or tacks. It will hopefully get more accurate in future versions. For more details about the new features, check boardsurfr.blogspot.com/2023/09/sorted-speeds-and-jibe-analysis.html

That's a very informative and well-written blog post. Thanks!

Wow! These are very interesting additions/enhancements!
Many thanks Peter.
I trimmed and analysed my own two best 1 hrs and, as you say in your blog, it is VERY interesting to analyse with the new features.
Here is my recent 2nd best 1Hr. (@ 10Hz). It really highlights how much one missed Gybe can hurt you!



This was a 1Hr working downwind slightly in Lake George in about 20-25 knots of wind on my 80L slalom and 5.7m sail. It is interesting because my best (much older) 1hr was done square to the wind all the way, back and forth in the same area of LG in the same way Kato's PB was done, and in much more wind (pretty solid and steady 30 knots) on the IS50 speed board and 5m sail back in 2012. (see further below)

That led me to having another play around with the Polar plot and I think another reason why the plot of speed angles can be deceptive is gybing. My one hour polar plot shows some quite high speeds off the wind at very broad angles, yet during the Hr run I didnt actually sail much more than about 10-15 degrees off the wind, except of course, in the gybes. Could you do an explanation of what each of the various Polar Plot view options mean please - or if I have missed it, point me to it. I am sure many others will be very interested too.

Data from 1hr 2022:



Data from my best 1 hr in 2012 (@1 Hz)


Other polar plots from 2012:












Sorted speeds 2012:



Turn analysis 2012:

Easy to see why this one was faster. Less turns (longer runs) and no dropped Gybe like in the other one.

One thing puzzles me about this one: It has 11 Starboard Gybes and only 6 Port Gybes? And yet they should be even each way? (In the other 2022 1 hr the turns we even number each way)

This is a very welcome feature! Thank you.

BTW I tried the turn analysis. Maybe this helps for bug fixing, but the best alphas that I have do not show up in the turn analysis feature. It says my top turn didn't go higher than 8.05 knots, but my top 5 alphas are all over 9kts.

If I take the track of only my best alpha it doesn't identify a turn at all. But the alpha code finds the alpha just fine.

Select to expand quote

sailquik said..
Turn analysis 2012:

When looking at the minimum speeds and percent scores, keep in mind that you are comparing filtered 1 Hz data with unfiltered 10 Hz data. The sampling rate alone will lower your minimum speed by about 0.5 to 1 knot just from random noise near the minimum speed (and also possibly from body movements). I suspect that the Kalman filter in the GT31 increases this difference even more. I don't have any decent data from a GT31 and a 5 or 10 Hz unit used in the same session to check this, but here is a comparison of 10 Hz Motion data to 1 Hz Coros data from Boz:


The different in minimum speed in this jibe is more than 2 knots (16.92 vs 14.79). The second-biggest difference is about 1 knot less, and the 14.79 knot speed is just one point that's 1.5 knots lower that the points on both sides, so that particular point was likely reading a knot or more too low due to random noise.

Basically, the 1 Hz data overstate the minimum speed somewhat, and the 10 Hz data understate it - in this case, by (very roughly) similar amounts.

Select to expand quote

aeroegnr said..
If I take the track of only my best alpha it doesn't identify a turn at all. But the alpha code finds the alpha just fine.

Alpha is easier because you can look at at least 200 meters, and require that the ends are close together, which makes it easy to find all turns. Looking at turns is harder, since we're looking at just 50 meters from each side of the turn center, and do not have constraints on how far the ends can be apart. Also, we can require a minimum average speed for alphas since 2 knot alphas are irrelevant, but not for turn analysis where we want to include crashed jibes and tacks. Wing tracks make it even harder, since they often have jibes with a diameter of 20 meters, and tacks with close to zero diameter, and turns can be very close to each other.

But alphas have been part of the software since the start, and the turn analysis is new, and still needs to be improved. If you can send me your file, that would help. Send me a PM if you don't have my email.

Select to expand quote

boardsurfr said..
Basically, the 1 Hz data overstate the minimum speed somewhat, and the 10 Hz data understate it ....

Yes, of course. That is a bit deceptive and I have noticed that before. But in this particular case I know I planed out of each turn (only just for the slowest one )

I may have some tracks with 1Hz and 10Hz together (I know I have 1Hz and 5Hz) I'll see if I can find some examples.

Of more interest is; why in this 1 hrs report are the Port and Starboard turns uneven?
Turns:
1 Sb
2 Pt
3 Sb
4 Pt
5 Sb
6 Sb
7 Pt
8 Sb
9 Pt
10 Sb
11 Pt
12 Sb
13 Sb
14 Pt
15 Sb
16 Sb
17 Sb

My first thought is that there were more turns in between the consecutive ones on the same tack that the Software does not pick up?

Checked the speed graph and that is indeed the case. 22 gybes in all and if I count one for each of the double ups, I get 21. Not sure where the other one went? But that explanation makes more sense. :-)

I'll send you this data if it might help you improve your algorithm.


The 10Hz Hr I plotted didn't miss any turns. (25) Perhaps this is more of an issue with 1Hz tracks?

Select to expand quote

sailquik said..
Of more interest is; why in this 1 hrs report are the Port and Starboard turns uneven?
..
The 10Hz Hr I plotted didn't miss any turns. (25) Perhaps this is more of an issue with 1Hz tracks?

As I said, the initial version of the turn analysis is still a bit buggy. Most of the initial tests were done with 5 Hz and 10 Hz data, so 1 Hz data have the most problems. I've fixed a lot of these and hope to release a new version with the fixes soon. Thanks to everyone who has sent tracks where they noticed problems.

I just uploaded a new version of GPS Speedreader (2.2.2) to ecwindfest.org/GPS/GPSSpeedreader.html

This version has better accuracy in turn analysis, better handling of 1 Hz data, and fixes a bug when reading GPX files that contain time stamps with fractional seconds. Thanks to Mike, Greg, and Andrew for reporting problems and sending files.

Thank you! These past two updates are incredible, between finally having polars and the turn analysis.

This is the updated run when things finally clicked with my foil jibes, able to finally fly through them without touching on some of them on my freeride (foilx and ptm 926 w/ 7.0) gear.

I definitely have a preferred jibe side (port), but my second best jibe was on my "bad" side. Can't wait to improve the jibe speeds and numbers a bit now that I can fly through at least some.

Went out today on iqfoil gear, very gusty conditions and some decent chop. Had to go from +1 to +0.5 after the +1 got super hectic when the wind was stronger.

Anyway, it looks like my jibes are about 10% above that 9kt threshold on either that gear or the freeride gear. That's a helpful metric to measure against. It's worlds ahead of where I was a year ago but also I have tons of room for improvement.

My best min speed on the IQFoil was also similar speed to the ptm 926. Have much higher entrance speeds (as you would expect), but exit speeds are similar. I think there's plenty of room for my jibes to improve. Didn't get one fully clean like I did on the ptm. Looking forward to the next sesh with the right wind in flat water where the wind has a lot more fetch and less obstructions causing gusts.

Hello Boardsurfr,
Nice job on the new update, thanks for all of your effort.

Great work Peter!

I ran the old 1hr PB again and it now finds all 22 turns.

Thanks again.