Blog Homepage

Identifying threshold power - Cherry picking data

It's not always convenient to schedule and complete a performance test during a training program.  Fortunately there are several other ways in which you can keep tabs on changes in performance and identify if your training zones need updating. Previous blog posts have looked at using histograms and the various testing option, however there is another method that can be used to determine threshold power, it’s called Cherry Picking and it can be a useful method for riders who use a power meter.


This method literally involves picking out your best power output for set durations from training or race files over a select period, or quite simply ‘cherry picking’ the best numbers that show your current ability.  These numbers then need to be analysed to determine your threshold power.  However, before you can do that you need to ensure that the data meets certain requirements so that you can be sure the results are valid. 


Effort Duration

When selecting data using the cherry picking method you should select durations that closely reflect those found in the power profile test, ideally 4 minute and 10 minute power.   However, as you will see below, for this method to work these durations are your starting points.


‘Currency’ of data

For this method to provide accurate results, the date from which the ‘cherry picked’ data is from is of the utmost importance.   The data should ideally come from within the current training phase and preferably no more than three weeks ago.  Using data from too far in the past, or using some data from a few weeks ago and some from the current week may not yield accurate results for your current threshold power.


Intensity of effort

This is one of the trickiest aspects of selecting data for this method.  For the cherry picking method to provide accurate results the values used must all be maximal values. The problem here is that seldom do we push ourselves to our maximal limits in training or in races (as doing so would likely cause us to ‘blow up’ or get pretty close to it) – the exception generally being towards the end of a race.  Therefore, when selecting individual power data to use in this method it’s important to consider the individual rides/races that the data comes from.  For example.

If you have found a 4 minute best power of 384 watts but it came from a race where the effort (at that time) was actually of a longer duration, for example 5min22sec at 368 watts, then you can be pretty sure that your actual best 4 minute power is going to be in excess of 384 watts.

Conversely you may have a maximal effort in training or a race of 9min 45sec that would be better to use than a 10 minute power value that didn’t reflect a maximal effort.

The key here is to use your critical power curve to indentify the date and ride file that contains the data and then examine that ride file more closely.  An example of this can be seen below.



This image is a copy of my Critical Power curve (from Golden Cheetah) for the current season.   I have selected the 4 min power value and you can see this is 367 watts from 27 October (which was a scratch race).  The critical power curve for the season is the red line, whereas the individual ride file’s power curve is the black line.  You can clearly see that there is a section (from around 2 minutes out to nearly 5 minutes) where the season best power comes from this one race.   In itself this is a good indication that my fitness is improving but it presents a problem when using the 4 minute power value as a ‘maximal’ value.

The first thing to do here is to use the ‘find power peaks’ function in Golden Cheetah to determine where the best 2 minute and 5 minute power occurred during the ride.  In this case both occurred at the end of the race, with the 2 minute peak happening at the end of the 5 minute peak.   This means the shorter duration is unlikely to be of much use but the longer duration value may be more suitable.  The next step is then determining exactly how long that final effort was and at what point was it the current best seasonal power.  This can be done using the critical power curve.  Just use the mouse to select a time/interval duration (or type it in) and gradually increase the value (in this case from 4 minutes) until you find the point that the peak race value is below the season best value.   For this ride file this occurred at 5min46sec (306 watts).  Using the same technique for the 10 minute duration I come up with a value of 281 watts for 10 minutes and 14 seconds (from a file only 4 days earlier).



Now that we have our two power/duration values the last check we need to make is the terrain.   Invariably, most people are able to sustain a higher power while climbing, in part due to more time spent riding out of the saddle.  In some ways this artificially inflates the threshold values – perhaps making them more climbing specific.  However of more importance is matching up the terrain that both of your cherry picked values came from.  

In the example above both of my values came from races on flat to undulating terrain.  If, however, one came from a hill climb and the other was on the flat then it would be unwise to use that data pair.   The outcomes for this would be:

  • If the shorter effort was on a hill then it would likely over inflate the anaerobic abilities and the threshold value would be calculated lower than it actually is.
  • If the longer effort was on a hill then it would over inflate the aerobic abilities and the threshold value would be calculated as higher than it actually is



Calculating your threshold from the cherry picked data

Assuming you are satisfied with the quality of the data points you have obtained the final step is actually calculating your threshold power.  This is done using the Monod Critical Power model, which calculates the slope and intercept of these power/duration pairs to determine your Critical Power (threshold) and Anaerobic Work Capacity.

There are several Monod Calculators on the web (and my Power Profile test calculator uses this model to calculate the test results) but it’s quite simple to create your own spreadsheet to do these calculations (or just Google Monod Critical Power Spreadsheet to find one).


To finish up with the example used in this post, my ‘cherry picked’ critical power is:

5 min 46 sec @ 306 watts

10 min 14 sec @ 281 watts

Critical Power = 249 watts


The final step here is then a sanity check.  Is this value higher or lower than the current threshold power? and does it look realistic?    Given my previous analysis indicated that my current threshold power was somewhere between 260-270 watts it’s highly likely that the cherry picked data was not close enough to my max to provide a valid threshold value – and therein lies the issue with using this method, if the data is not of sufficiently high quality (effort intensity, etc) then the values can be out quite a bit. 

However, with good data it can be very accurate and like the histogram methodology, it can provide a good backup and secondary check of other testing data.

Jason Mahoney Tuesday 05 November 2013 at 09:22 am | | Tips

No comments

(optional field)
(optional field)

Comment moderation is enabled on this site. This means that your comment will not be visible until it has been approved by an editor.

Remember personal info?
Small print: All html tags except <b> and <i> will be removed from your comment. You can make links by just typing the url or mail-address.