Golden Cheetah v2 released into the wild

amaferanga

Well I got to the bottom of it. The ride data and new cyclists I was creating were being stored in the Virtual Store directory. No idea why, but I deleted all the GC stuff in the Virtual Store directory and its all sorted.

bigpikle

For the first time yesterday it seemed my power file included a wild spike - I'd like to think I briefly created 16,000w during one of my bursts, but I suspect not....

I used the Fix Power Spike function in GC, in fact I used 5/6 times, but it still hasnt brought the spike down to the correct level and then I used the Editor to find the spikes and delete the numbers in there. I saved the data in the Editor and saved the ride again but when I restart GC it once again gives me Max Power of >16,000w for the ride. I wouldnt mind but it totally screws up the PM and CP graphs for ALL my data, as it skews the data ridiculously.

I dont want to delete the ride but is there a way to permanently fix the spike? I'm happy to simply manually enter a max power figure as it was a 90 mile endurance ride and the spike happened within 1 minute of getting the bike out of the garage, so it wont afect my real ride data if I use the highest number from the rest of the ride as the real maximum output?

Any help is appreciated here. I did manually recalibrate the PT with my 705 as soon as I got the bike out, so am a little confused where the spike came from in the first place as well?

liversedge

Bigpikle wrote:

Any help is appreciated here. I did manually recalibrate the PT with my 705 as soon as I got the bike out, so am a little confused where the spike came from in the first place as well?

You must remember to save the file after you change it. Do you get a warning on exit about saving files? (it may not appear if you previously clicked 'do not show me this again').

I fixed a few bugs related to metric recalcs in the beta v3 builds. Oh, you mentioned a bug in weekly summary a while ago, this has also been fixed (thanks for highlighting it). We're hoping to get v2.1 out soonish, but v3 may be a while off yet.

If you're on Windows then a beta v3 installer is gonna appear very soon on goldencheetah.stand2surf.net, but be aware it uses a new file format that will be ignored in earlier releases.

bigpikle

thanks for the reply.

sadly, I'm saving and saving and saving....and even rebooting, but the changed data is NEVER reloaded when I restart GC. It always reopens with the same 16,000w spikes even though I removed the entire rows in the Editor and did all the other stuff, while constantly saving

I'm using it on OS X by the way. Is there a way to manually edit a power file before importing it to GC or any way to edit the file once its imported, other than through the standard UI? I'm also wondering if there is a cached file somewhere I could delte to force it to recalculate using the newly edited file?

bigpikle

thanks for the reply.

sadly, I'm saving and saving and saving....and even rebooting, but the changed data is NEVER reloaded when I restart GC. It always reopens with the same 16,000w spikes even though I removed the entire rows in the Editor and did all the other stuff, while constantly saving

I'm using it on OS X by the way. Is there a way to manually edit a power file before importing it to GC or any way to edit the file once its imported, other than through the standard UI? I'm also wondering if there is a cached file somewhere I could delte to force it to recalculate using the newly edited file?

bigpikle

thanks for the reply.

sadly, I'm saving and saving and saving....and even rebooting, but the changed data is NEVER reloaded when I restart GC. It always reopens with the same 16,000w spikes even though I removed the entire rows in the Editor and did all the other stuff, while constantly saving

I'm using it on OS X by the way. Is there a way to manually edit a power file before importing it to GC or any way to edit the file once its imported, other than through the standard UI? I'm also wondering if there is a cached file somewhere I could delte to force it to recalculate using the newly edited file?

bigpikle

thanks for the reply.

sadly, I'm saving and saving and saving....and even rebooting, but the changed data is NEVER reloaded when I restart GC. It always reopens with the same 16,000w spikes even though I removed the entire rows in the Editor and did all the other stuff, while constantly saving

I'm using it on OS X by the way. Is there a way to manually edit a power file before importing it to GC or any way to edit the file once its imported, other than through the standard UI? I'm also wondering if there is a cached file somewhere I could delte to force it to recalculate using the newly edited file?

liversedge

Bigpikle wrote:

thanks for the reply.

sadly, I'm saving and saving and saving....and even rebooting, but the changed data is NEVER reloaded when I restart GC. It always reopens with the same 16,000w spikes even though I removed the entire rows in the Editor and did all the other stuff, while constantly saving

I'm using it on OS X by the way. Is there a way to manually edit a power file before importing it to GC or any way to edit the file once its imported, other than through the standard UI? I'm also wondering if there is a cached file somewhere I could delte to force it to recalculate using the newly edited file?

How strange! Any bloody annoying I'd imagine. Never heard of it before -- do you see duplicate entries in the ride list? If not then send me the file in question and I'll take a peek!

bigpikle

OK - think I've solved it

I went into the folder where GC stores all the ride data and found that there were 5 files related to yesterdays ride, dated yesterday and today. It seemed like there should only be 2 or 3 based on what was there for past rides, so I deleted the lot and started again.

I just re-imported the original file from my 705 again, edited the spikes and re-saved it, and so far it seems like its worked. I quit GC and re-started it and the edited data seems to be preserved. I guess there was an old file there for some reason that GC was constantly opening, and simply deleting the ride through the UI wasnt actually removng it.

The PM and CP charts now seem accurate as well - a Relative Intensity of 1.3 for a 6 hour 90 mile ride did seem a little unusual

liversedge

Bigpikle wrote:

OK - think I've solved it

I went into the folder where GC stores all the ride data and found that there were 5 files related to yesterdays ride, dated yesterday and today. It seemed like there should only be 2 or 3 based on what was there for past rides, so I deleted the lot and started again.

I just re-imported the original file from my 705 again, edited the spikes and re-saved it, and so far it seems like its worked. I quit GC and re-started it and the edited data seems to be preserved. I guess there was an old file there for some reason that GC was constantly opening, and simply deleting the ride through the UI wasnt actually removng it.

The PM and CP charts now seem accurate as well - a Relative Intensity of 1.3 for a 6 hour 90 mile ride did seem a little unusual

Excellent. Well, except for the fact that it got out of kilter in the first place. I have seen some reports of changing the date/time of the ride causing duplicate rides to be left behind. Be interesting to know if you may have done that, I need to check the code too!

bigpikle

no - the only thing I did was to run the power spike tool several times and re-save each time. I then tried deleting the ride using the UI and only later found the Editor function (I didnt have that showing in the standard view) and by that time the duplicate ride files were already created.

I have started to tick 'over-write existing data' when importing a file now, as I guess that should avoid duplicate files existing if I delete a ride and try to add it again.

Thanks for the assistance as it prompted me to find the solution.

amaferanga

Just spotted the latest development build has real time mode without the need for virtual machines and all that gubbings. Very exciting.

Plus a new interface, and some of the Training Peaks metrics (NP, TSS, etc.) though not all the buttons/functions seem to be working yet.

One thing I've noted before from looking at one or two rides is that NP is usually higher than xPower and I'm seeing it consistently higher on most of my rides. Sometimes only 5W higher, but up to 20W higher.

liversedge

amaferanga wrote:

Just spotted the latest development build has real time mode without the need for virtual machines and all that gubbings. Very exciting.

Yeah, we wanted folks to try it out. The SRM and PT code seems "ok" but interested in everyone's experience. It will generate an 'antlog.bin' file which is useful for debugging (its a trace of all the ant messages as 12-byte fixed length records).

It works with both the Garmin sticks (old and new). There is also a virtual power calculation in there for 2 or 3 different turbo trainers for folks who just have a wheel speed sensor (i.e. no power meter).

Plus a new interface, and some of the Training Peaks metrics (NP, TSS, etc.) though not all the buttons/functions seem to be working yet.

We also have upload/download to TP.com, the guys at TP have been really cool. Not all the buttons are implemented (diary view, measures, athletes) but we're on the way. The configurable layout of 'Home' will eventually be used for all the other options (diary, train etc). Its quite fun to drag and drop charts onto the home page and muck about with them... this build doesn't bring a default setup so its just blank...

One thing I've noted before from looking at one or two rides is that NP is usually higher than xPower and I'm seeing it consistently higher on most of my rides. Sometimes only 5W higher, but up to 20W higher.

20w is a big difference!! the algorithm is pretty spot on, its been thru a few "eyes" ... be interested in seeing the file in question!

amaferanga

liversedge wrote:

20w is a big difference!! the algorithm is pretty spot on, its been thru a few "eyes" ... be interested in seeing the file in question!

If you PM me your email address I'll get the file to you. It was from a race that lasted 59 minutes. AP 273W, xPower 298W, NP 322W.

jibberjim

amaferanga wrote:

liversedge wrote:

20w is a big difference!! the algorithm is pretty spot on, its been thru a few "eyes" ... be interested in seeing the file in question!

If you PM me your email address I'll get the file to you. It was from a race that lasted 59 minutes. AP 273W, xPower 298W, NP 322W.

This isn't an unsurprising difference... xPower is a different calculation and the above differences are somewhat believable.

amaferanga

jibberjim wrote:

amaferanga wrote:

liversedge wrote:

20w is a big difference!! the algorithm is pretty spot on, its been thru a few "eyes" ... be interested in seeing the file in question!

If you PM me your email address I'll get the file to you. It was from a race that lasted 59 minutes. AP 273W, xPower 298W, NP 322W.

This isn't an unsurprising difference... xPower is a different calculation and the above differences are somewhat believable.

Well yes, but the paper by Skiba on the Bikescore algorithm suggests that xPower and NP should be quite similar. If it was a 5W or even 10W difference I wouldn't care, but 24W is significant. I have another race file with a difference of 28W over a period of 2h17min which means that Bikescore differs from TSS by 35 points. Again I'd say that is significant for metrics that are as I understand it touted as being pretty much the same.

jibberjim

amaferanga wrote:

Well yes, but the paper by Skiba on the Bikescore algorithm suggests that xPower and NP should be quite similar.

Whatever the paper says, the maths says something very different.

amaferanga wrote:

Again I'd say that is significant for metrics that are as I understand it touted as being pretty much the same.

They're the "same" in that they are used to measure the same thing, that doesn't mean the numbers are interchangeable. A performance chart created using xPower and a performance chart created using NP will show the same thing, but that doesn't mean the numbers themselves will be identical.

amaferanga

jibberjim wrote:

amaferanga wrote:

Well yes, but the paper by Skiba on the Bikescore algorithm suggests that xPower and NP should be quite similar.

Whatever the paper says, the maths says something very different.

amaferanga wrote:

Again I'd say that is significant for metrics that are as I understand it touted as being pretty much the same.

They're the "same" in that they are used to measure the same thing, that doesn't mean the numbers are interchangeable. A performance chart created using xPower and a performance chart created using NP will show the same thing, but that doesn't mean the numbers themselves will be identical.

I'm not an idiot and I do understand how the algorithms differ, but unless Skiba bodged his data then the maths tell me the algorithms can and do produce similar results sometimes. And I never said I expected identical results. I do expect similar results most of the time, but that's not really what I'm seeing particularly for highly variable rides such as races.

liversedge

I've looked at this before, and generally NP is higher than xPower, typically by about 10-15w for tempo and threshold sessions.

Here is a chart of a complete season showing NP vs xPower.



Uploaded with ImageShack.us

jibberjim

amaferanga wrote:

I'm not an idiot and I do understand how the algorithms differ, but unless Skiba bodged his data then the maths tell me the algorithms can and do produce similar results sometimes. And I never said I expected identical results. I do expect similar results most of the time, but that's not really what I'm seeing particularly for highly variable rides such as races.

The difference between the two is how the averaging is done 25second EWMA and 30second moving averages produce quite different results in very highly variable conditions (ie races) with the big sprint efforts impacting differently in the two methods.

Nowhere that I can see has Skiba ever looked at the situation, his examples in the bikescore paper are not heavily variable in the same way - there's only a single sprint effort in all the examples and that's not going to be enough to make the difference in an overall calculation.

The algorithms simply don't allow for similar results in that situation - I personally find the xPower one a lot better in the race situation - I produce NP busters in just about every race I do - the only times I don't is when they're nowhere near maximal.

Alex_Simmons/RST

xPower can also sometimes produce a number not insignificantly under average power for regular ride durations (not the short duration 1-2W variance you might see for NP), for which I have yet to understand why you would ever want a normalising algorithm to do that.

amaferanga

I had NP busters pretty much every race towards the end of last season, where as xPower does seem more believable.

cookiemonster

re: the "power spike" issue

Ive just been through the same - imported a ride from quarq/705 and found that my 1s peak power was 18,000w in the CP calculation. I ran the remove power spike tool but the crazy readings remained. I couldnt find the a high power reading in the editor tab though - so it looked like the peak power had been removed, but CP calculation didnt update. I then deleted the ride, but the power spike still remained in CP even though the ride was no longer listed.

I had a read and came across the timing issue (where the garmin can reset the time - for example by acquiring a gps signal 5 minutes into the ride), and dug through the xml file looking for discontinuities in time. None existed.

I'm a new user of GC, so not much data in there, so I tipped everything out and started from scratch. I then just edited out the high power spike in the tcx file directly (16,000w btw ), and reimported it. All working fine now.

So I guess the question is, what do you have to do to update the CP calculation after you have either a) run the power spike tool, or b) deleted a ride?

Havent coded for 10 years, but I will attempt to resurrect my C++ knowledge and have a look myself

jon

liversedge

cookiemonster wrote:

re: the "power spike" issue
So I guess the question is, what do you have to do to update the CP calculation after you have either a) run the power spike tool, or b) deleted a ride?

Thanks for the detailed diagnosis. Indeed, the CP plot only refreshes once and then ignores updates to ride files. This is a bit of a pig to fix, the CPplot code is muddled up with the calculation code and needs to be refactored.

I'll raise a bug for this and proly take a look sometime in the next few weeks.

Cheers,
Mark

wthrelfall

I have an issue, probably my bad but my trimp values all seem to have reset to zero since 5 rides ago.. i haven't altered anything that i'm aware of...help?!

wthrelfall

oops, i meant everything before 5 rides ago. latest 5 rides are all fine.

liversedge

wthrelfall wrote:

oops, i meant everything before 5 rides ago. latest 5 rides are all fine.

Check your HR zones in the config pane and look at the first start date, perhaps your HR zones only start 5 rides ago?

wthrelfall

liversedge wrote:

wthrelfall wrote:

oops, i meant everything before 5 rides ago. latest 5 rides are all fine.

Check your HR zones in the config pane and look at the first start date, perhaps your HR zones only start 5 rides ago?

That's sorted it, thanks.

gandhi

I just got an ANT+ dongle so I could play with the real time display stuff. Is there a way to set up a training plan (e.g. countdown intervals etc.)? Is there a way to customise the data fields displayed?

I've been on the wiki and I couldn't find documentation on how to do those things - is there some stashed away elsewhere? (I'd be happy to contribute as I learn how to do stuff.)

liversedge

gandhi wrote:

I just got an ANT+ dongle so I could play with the real time display stuff.

Super cool, I assume you downloaded the v3 beta from goldencheetah.stand2surf.net for the native ANT+ stuff (i.e. no quarqd) ? Be interested in your experience, the code is very new

Is there a way to set up a training plan (e.g. countdown intervals etc.)?

Yes. We support computrainer format workout definitions and there is a workout wizard in the tools menu to create one segment by segment or from a ridefile itself (assuming it has altitude).

Is there a way to customise the data fields displayed?

If you are using v3 you will see that the 'home' screen lets you drag and drop charts and move / resize them. This will be the UI for the whole of GC.

As it happens I'm working up the realtime stuff this weekend to allow you to customise / drag and drop charts/dials etc and move them around. There is also some really cool stuff in the works with streetmaps and videos, but that will be along in the next few months, not days.

I've been on the wiki and I couldn't find documentation on how to do those things - is there some stashed away elsewhere? (I'd be happy to contribute as I learn how to do stuff.)

That is one of many achilles heels for us. If you want to contribute that would be gratefully received -- in fact, documentation is probably more important than code to us right now. The wiki is here: http://bugs.goldencheetah.org/projects/ ... eetah/wiki

Let us know how you get on!