Why does "Efficiency Factor" use normalized power and not average power?

Efficiency Factor = Normalized Power / average HR

Anyone know why Normalized Power is used instead of average power? This skews EF up for higher intensity efforts (higher EF) even more than what would happen anyway due to power vs HR having negative intercept.

If you are looking at EF over time as a fitness measure you probably don’t want that.

1 Like

NP is used because not all APs are created equal. Normalized power refers to how hard your body has worked for a given average wattage.

I agree. I’ve never liked NP. Let’s take each rolling 30sec power and raise each to 4th power. Then avg those. Then take 4th root of that. How can someone correlate that to the actual cost on the body? Then folks using NP for PR’s :rofl:. I know the crit folks won’t be happy with this response.

Do you understand where EF came from? The entire point is to establish a benchmark longer aerobic endurance workout, keep everything the same as much as possible (temperature, hydration, mood, rest, etc), and then track EF for only those benchmark workouts.

In other words, don’t use a hammer when the job calls for a drill.

He’s looking at it from a developer’s perspective. You may only use it for endurance efforts, but he has to calculate it for all efforts. Just like other data sites do. So imo a very viable question to ask.

1 Like

The simple approach is to hand everyone a tool chest and allow them fail at drilling a hole because they insist on using a hammer. Ideally the developer will not only give you a set of tools, but also warn you when the wrong tool is being used.

Yeah, that would be nice. And I believe that’s why this question is so important. At what point does it not make sense to use this? Is it Z3, Z4, Z5?

And you should probably mention that to TP and everyone else because they calculate it for any effort and never give any warnings.

sure but TP is targeted at coaches, and educates coaches with articles like this and this, along with a series of WKO webinars that discuss the use of tools like EF.

Here are two quotes from the two articles:

“We have known for decades that if heart rate during an all-aerobic (below lactate/anaerobic threshold) workout rises while the intensity (power or pace) stays the same then the athlete is not operating efficiently and his or her aerobic endurance is questionable.”

emphasis on “all-aerobic (below lactate/anaerobic threshold)”

“As the name implies, this metric is a way to measure efficiency during certain types of workouts as a way of looking at output versus input.”

emphasis on “certain types of workouts”

“To make the best use of this number you would want to do similar steady state workouts over time and track the EF to monitor the progress of your training and its impact on your efficiency.”

The EF metric was designed for use in traditional base training, and when you plateau its time to move on to more intensity (tempo and sweet spot).


It’s a pretty close approximation of the area under a curve of work vs time (a more precise formula would an integral function that most applications do no support. note: excel, for example, does not support calculus functions). Using 5th, 6th, and 7th, etc. powers will provide greater accuracy, but you are into decimal point precision that brings no value since TSS is an approximation anyway. As for the time period of 30 seconds, perhaps there could be some improvements there, but when the formula was developed (about 15 years ago), computers were a lot slower and calculation of TSS for a multi-hour ride using less than 30secs probably wasn’t practical.


Because using AP will give 0w measurements a disproportional weighting vs. using NP, which will create increased variability in the data.
Given that EF is supposed to be used for longer intervals of “steady” power to track differences over time, using NP creates a smoothing effect and allows better comparison data with less variability between similar sets.

1 Like

Hey Bob, hope you are doing well. My understanding is that Coggan analyzed blood lactate response to power above first lactate threshold (the aerobic one, LT1):

and determined the following:

Perhaps not surprisingly, an exponential function provided the best fit, but a power function of the following form proved to be nearly as good:

blood lactate (% of lactate at LT) = power (% of power at LT)^3.90; R2=0.806, n=76

Based on these data, a 4th-order function was used in the algorithm for determining the IF (the exponent was rounded from 3.90 to 4.00 for simplicity’s sake).

So instead of using an exponential function to relate power and blood lactate, Coggan approximated using a 4th order function.

Then the 30-second average part:

the important facts are 1) the half-lives (50% response time) of many physiological responses are directly or indirectly related to metabolic events in exercising muscle, and 2) such half-lives are typically on the order of 30s. Thus, to account for this fact the power data were smoothed using a 30 second (~1 half-life) rolling average before applying the 4th order weighting as described above.

That is all straight from Coggan’s paper Training and racing using a power meter: an introduction and I have the 25 March 2003 revision which can be found on the Internet.

@cwiggum so to answer your question, the formulas are based on an estimate of how your muscles respond (as measured by blood lactate) to increasing power levels.


Good to know there is science behind it. Thanks for the lesson guys!! I think I’ll still find myself mostly ignoring it though.

1 Like

I know that but even for those workouts EF done using average power would produce more consistent results by not accentuating whatever little bumps are in your workout. But anyway if you are consistent it won’t make much difference.

I tried plotting EF over time with various filters (low variability, lower intensity) to try include only the endurance rides and didn’t have much luck getting any “fitness” info out of it.

One approach would be to just calculate EF using average power. This would result in a lower number than using normalized power … and as a result may encourage people to only look at EF for steady paced / low variability rides where NP and AP are pretty much equal.

Assumption here is people are looking to increase EF, hence will have an incentive to pay attention to EF only for rides where they have a chance of getting a higher EF than last time I.e. steady paced rides.

Or maybe I’ve been listening too many to freakonomics podcasts…

its called efficiency factor, not fitness factor :rofl:

1 Like

So could you utilise EF tracking during high medium/long sweetspot (I.e. sub threshold) intervals as well as for entire low intensity rides?

The real problem with EF isn’t the numerator, it’s the denominator.

1 Like

For me, my NP : HR on a MTB course, as long as it doesn’t have really long descents is really close to AP :HR for steady efforts. That to me validates NP… The steadier the effort, the closer NP is to AP, but with a VI of 1.2 or so or more for MTB or other hard efforts I do on rolling terrain to prepare for it NP:HR is a pretty consistent with AP;HR from steady training efforts.