Golden Cheetah - How much data required for CP estimate?

springtide9

Hi...

Finally have my Powertap back after the wheel went out of true (wasn't build up right)... so today I have finally entered my 4th lot of data into GC and hit the "Calculate CP value"

I'm assuming that the CP value is the same as FTP?

Just to add a little bit more confusion (on my part)... there is also a "Daniels EqP" power as well as an "xPower" (which I can't find any documentation on):

e.g. http://bugs.goldencheetah.org/projects/ ... eetah/wiki

The main question is my CP value. It seems that everytime I enter new data the CP value jumps around randomly.
Now I'm guessing that after GC has enough data it will stop doing this (as it builds the graph), but wonder roughly how much data it needs to give a reasonable accurate value?

I am also assuming that I should only add "tough" rides or rides with intervals?

Any advice on how to read the data would be most useful.

Apart from not really understanding the terminologies and what I'm really looking at in terms of data, it looks like a great bit of software. Are there any other software recommendations that I should give a trial?

liversedge

Ideally, you need to make sure that any data you use has some maximal efforts in the 1,6 and 10 minute range.

The maximal value towards 60 mins will also be used, but if you do a 60 minute maximal effort there is no point in trying to derive you maximal 60 min power!

springtide9

liversedge wrote:

Ideally, you need to make sure that any data you use has some maximal efforts in the 1,6 and 10 minute range.
The maximal value towards 60 mins will also be used, but if you do a 60 minute maximal effort there is no point in trying to derive you maximal 60 min power!

Thanks.

I was reading the documentation which explained how the CP is calculated from the data, and that seemed to imply that it basically finds intervals within the training files to use to complete points on the power/duration graph (so that it can estimate your CF)

I am also guessing that by looking at the graph/data I should also start being able to see my strengths and weaknesses and use this information to focus any training I do to improve
i.e. I think I have a reasonable 1 - 5 min power as well as 60+ min power compared to my estimated power for 10-30 mins (although this obviously could be due to the limited intervals I've done with the PT are mostly less than 10 mins in length)

I am not sure whether owning a Powermeter is making me more informed or just causing more confusion. I guess it's not rocket science and I just need a little bit of time understand the data I'm seeing.

liversedge

springtide9 wrote:

I was reading the documentation which explained how the CP is calculated from the data, and that seemed to imply that it basically finds intervals within the training files to use to complete points on the power/duration graph (so that it can estimate your CF)

Yes exactly. So have a maximal effort in those ranges. You don't need to be ultra precise, just go and kill it for about a minute, six etc.

springtide9

liversedge wrote:

springtide9 wrote:

I was reading the documentation which explained how the CP is calculated from the data, and that seemed to imply that it basically finds intervals within the training files to use to complete points on the power/duration graph (so that it can estimate your CF)

Yes exactly. So have a maximal effort in those ranges. You don't need to be ultra precise, just go and kill it for about a minute, six etc.

OK, so I have four lots of data....

After getting GC to look at the data I got:

Data (1) CF of 280 watts (~ 4 hr reasonably hard ride)
Data (2) CF of 250 watts (~ 90 min reasonably easy ride with Strava segment intervals)
Data (3) CF of 270 watts (~ 90 min moderate with Strava segment intervals)
[delay for 4 weeks as PT got repaired]
Data (4) CF of 303 watts (~ 90 min moderate with Strava segment intervals)

Now I would love the last CF to be true, but something tells me that after I upload a bit more data this will drop. It looks to have increased after a good 6 min & 10 min interval as the curve now has a lot of unfilled space under the curve (with just the touch points of the 6 & 10 min power)

I will upload the graph and hopefully it will help you (or someone else) help me understand better:



Not sure whether it's the done thing to share information like this or not, so apologies if it's not.

amaferanga

Unless you're doing proper maximal efforts then I wouldn't worry too much about the details of the curve yet. And by maximal I mean what you can do in a well rested state if you go absolutely balls out for the given duration. Picking bits from a normal ride where you sort of went hard up a hill will only mislead you. Also, CP will be overestimated if you've not done proper maximal efforts of short durations, but you have done a (say) 20 minute test.

springtide9

amaferanga wrote:

Unless you're doing proper maximal efforts then I wouldn't worry too much about the details of the curve yet. And by maximal I mean what you can do in a well rested state if you go absolutely balls out for the given duration. Picking bits from a normal ride where you sort of went hard up a hill will only mislead you. Also, CP will be overestimated if you've not done proper maximal efforts of short durations, but you have done a (say) 20 minute test.

Thanks for the reply.

I guess I should now look at the curve and try and put in some hard efforts to fill in the blanks(gaps)?

Obviously some of the longer efforts are not easy as they are done on the road; and traffic / lights etc seem to tend to get in the way! I assume the more data GC has the more accurate the curve.

Thanks again...

Alex_Simmons/RST

springtide9 wrote:

I'm assuming that the CP value is the same as FTP?

That depends on what GC reports as CP. Often CP models overestimate FTP by a bit.

springtide9 wrote:

The main question is my CP value. It seems that everytime I enter new data the CP value jumps around randomly.

As with any model, GIGO applies.

springtide9 wrote:

Now I'm guessing that after GC has enough data it will stop doing this (as it builds the graph), but wonder roughly how much data it needs to give a reasonable accurate value?

The critical power model only needs two good data points that have been recorded within a reasonably close time period, say a week or so. Three can be useful as a check.

I'd recommend one maximal test of ~ 3-6 minutes, and another of ~ 20-25 minutes from the same week. That's all you need.

Everything else is just noise. Cherry picking data across anything more than a couple of weeks is a sin of sins.