Accuracy, Precision and Virtual Power...

I'm a big fan of using indoor trainers for building cycling power and strength.  I recommend the use of trainers for interval work all year round.  When the weather is nice, it is hard to beat riding outdoors for longer endurance rides.  But, utilizing a trainer allows you to target and execute on very specific intensity levels without the concern or interruption from traffic, intersections or weather.  How beneficial can quality trainer work be?  Check out this record setting 24-hour cycling performance, where the vast majority of this athlete's training was done on an indoor bike trainer.

One of the best ways to target and quantify workout intensity while cycling is to use power as a metric. Unlike heart rate, power is not impacted by fatigue, core temperature or hydration level.  Power also reacts very quickly to changes in intensity (which HR does not), allowing for quantification of short duration VO2 max sets, such as 30 second billats.  Although prices for power meters have dropped substantially in the past couple of years, the price of power meters is still a significant obstacle for many cyclists.  Likewise, electronic "smart" trainers than can actively control resistance (e.g. you dial in 200w and it provides 200w of resistance, like a Wahoo Kickr or CompuTrainer) tend to cost as much or even more than a power meter, putting them out of reach of many athletes.

During the past couple of years, a few software programs / apps have created a simulated power metric, called "virtual power" (e.g. TrainerRoad).  Using an Ant+ or Bluetooth speed sensor mounted to your rear wheel, the app reads your speed and uses a mathematical equation to estimate power based on a known or estimated resistance curve.  The user can then see their "virtual power" and change their intensity level to meet their specific workout targets.  Utilizing virtual power for indoor training purposes allows the user to get many of the benefits of training with power, without the need of the up front cost of an actual power meter.  

Each manufacturer designs their trainers with their own proprietary resistance curve, which is typically a function of the trainer design.  As such, some indoor trainers are much better than others for use with virtual power.  To understand why, it is important to understand the concepts of precision and accuracy.  Below is a great graphic providing an overview of both concepts, and how they relate to each other:

Ideally, your trainer would have both high accuracy and high precision.  In my experience (both personal and with athletes that I coach), I've found the Kurt Kinetic Road Machine to be a great trainer to provide both high accuracy and high precision.  They have been designed specifically for this purpose, and unlike other manufacturer's, Kurt Kinetic actually posts their resistance curve on their website.   

I've had a couple athletes that I coach use both virtual power and then actual power (from a power meter) using two different fluid-based indoor trainers.  The first used a Kurt Kinetic Road Machine, and the second used a CycleOps Fluid 2 trainer.  Both were using TrainerRoad to calculate virtual power.  I looked at a series of segments from multiple rides to see how the actual power compared to the virtual power calculation.  Here's how the Kurt Kinetic trainer performed:

 Kurt kinetic, virtual power from website data.

Kurt kinetic, virtual power from website data.

This first graph (above) is a plot of the virtual power from the equation listed on their website.  The graph below is how TrainerRoad estimated virtual power versus the published data:

 virtual power from trainerroad, plotted on kurt kinetic equation... a good fit of data!

virtual power from trainerroad, plotted on kurt kinetic equation... a good fit of data!

As you can see, the virtual power (blue diamonds) from TrainerRoad lined up very close to calculated data from the Kurt Kinetic website.  This shouldn't be a surprise, as they likely use the published resistance curve. Below is the actual power data (red squares) plotted on top of the previous graph:

 Actual power plotted against virtual power for the kurt kinetic road machine.

Actual power plotted against virtual power for the kurt kinetic road machine.

As you can see, the actual power was very close to the virtual power calculation... well within the power meter's level of accuracy (within plus or minus 2%).  This is an example of a trainer that is both precise and accurate.  

Next we'll look at a CycleOps Fluid 2 trainer.  I find these to be very popular locally, with a lot of athletes using them.  Unfortunately CycleOps does not publish a resistance curve equation, just a picture of their resistance curve on their website.  

 Cycleops fluid2 trainer plot and polynomial trend, based on trainerroad virtual powerA

Cycleops fluid2 trainer plot and polynomial trend, based on trainerroad virtual powerA

As you can see, the virtual power calculations from TrainerRoad follow a very consistent and predictable trend for the CycleOps Fluid 2 trainer.  Now, let's see how the actual power compares to the calculated virtual power:

 cycleops fluid 2 virtual power plot versus actual power.  red squares are the actual power with blue diamonds being the virtual power.

cycleops fluid 2 virtual power plot versus actual power.  red squares are the actual power with blue diamonds being the virtual power.

In this case, you can see the TrainerRoad virtual power model for the CycleOps Fluid 2 is neither accurate nor precise. The average error is around 15%, but it ranges from less than 1% error on a few data points at the high power end to over 50% on the lower power values.  One could argue that the problem is simply that TrainerRoad's virtual power model is creating the accuracy problem.  That may be a contributing factor, but part of the accuracy problem could also be a result of the lack of precision or repeatability of the data... it's hard to create an accurate mathematical model with wide swings in data.  Without publishing an actual equation for the speed / power model, it's difficult to tell what the manufacturer had targeted for a resistance curve.  Likewise, although the plot above is the aggregate data, there was distinct ride-to-ride variation (one day to the next).  Even worse, there was variation within the ride... the longer you ride, the more the resistance increases for the Fluid 2.  Looking at one example, the average for three segments approximately three minute in length within a single longer interval set resulted in the following speed / power relationships: 14.6/126, 14.3/129, and 13.9/133.  As the segment continued, the speed dropped by 4.8%, while the power rose by 5.6%.  In this particular example, the virtual power would have shown power dropping from 154 to 143 as speed dropped, while power actually increased from 126 to 133.

So, what does this mean?  If you want to use virtual power, your best bet is to get an indoor trainer that is both accurate and precise. 

  • With a trainer that lacks precision and accuracy, your FTP (functional threshold power) would not be comparable to anyone else's, and you may not have confidence that you were actually executing on your targeted power zone.  I would be inclined to watch my heart rate closely under these circumstances, to ensure I was on target.  Or... if you are thinking about upgrading your trainer, rather than spending $300+ on a new trainer, consider the possibility of simply investing in one of the newer lower costs power meters instead (e.g. 4iiii Precision or Stages). Having actual power trumps the accuracy and precision issues of a trainer, and can be used outdoors for training too.  
  • If your trainer is precise (repeatable day to day and within the workout), but not accurate, then comparing your FTP with someone else's FTP is meaningless.  But, with a precise trainer and virtual power, you can establish a virtual FTP value and feel confident that when you are targeting zone 2 or zone 5 work, you are actually training within those zones.  
  • With a trainer that is both accurate and precise, you have much more confidence that you can compare your power output with others, and that if you are targeting Zone 2 or Zone 5 work... you are actually training within those zones.   

Train smarter... not harder.

Notes:

  1. I have no vested interest in the Kurt Kinetic trainers, and get no income if anyone buys their product.  I have simply found that every time I have verified trainer performance with a power meter, the Kurt Kinetic Road Machines have followed the predicted model very closely.  Here is a great video giving more details about their fluid resistance unit.