Accuracy vs. precision, and why it matters in cycling

For most of my youth, I used "accuracy" and "precision" interchangeably. As I got deeper into cycling at the enthusiast level, I heard phrases like:

- "That heart rate monitor isn't accurate."
- "Brand x's power meter isn't very precise."
- "The gauge on my floor pump isn't precise enough to accurately set tire pressure."

Do you actually know the difference between accuracy and precision? I'm embarrassed to admit I didn't, and I'd worked in the bicycle industry most of my entire adult life. Until I had the opportunity to meet some machinists, who taught me the distinction between the two.

Accuracy vs. precision - what's the difference?

Accuracy is how close a measurement is to the "truth", or accepted standard. If I'm supposed to drill a one inch hole in a block of aluminum and the hole measures two inches in diameter, I am wildly inaccurate.

Precision has two components:

First, is whether that measurement is repeatable. If I measure again, do I get the same result? If the length/weight of the object hasn't changed, then neither should my measurement (at least not by much.)

The second component of precision, is the number of decimal places; i.e. the ability to measure to 1 vs. 1.1 vs. 1.11 and so on. A tool that can measure to more decimal places is said to be more precise.

Real-world examples of accuracy vs. precision

Bathroom scale example

Let's imagine a cheapo bathroom scale. I'm 150 lbs. I jump on the scale three times and it says "160 lbs" each time - clearly the scale is not accurate. However, if it says "160 lbs" every time as I step on and off, the scale is precise, even though it isn't accurate. The result is wrong, at least it's consistent.

Now imagine I'm still 150 lbs, and I jump on the scale three times, obtaining readings of 149.6, 150.1, and 150.5 lbs. That's an average of 150.066666 lbs.... the scale is plenty accurate, but it isn't precise, because it gives results that aren't consistent - there is variation.

Dartboard example

A dartboard is another commonly used example to illustrate accuracy vs. precision. Imagine we throw 10 darts at the dartboard:

  • If there is a cluster of 10 darts all in the bulls-eye, we have been both accurate and precise.
  • If there is a cluster of 10 darts at the edge of the board, we have been precise, but not accurate.
  • If the 10 darts are all around the bulls-eye, but some are 2" away, some are 1" away, and some are right dead center, we've been accurate, but not precise.
  • If the 10 darts are scattered all over the board (or the floor), we've been neither accurate, nor precise.

Accuracy and precision on bicycles

There are several components, accessories, and techniques on performance-oriented bicycles where one, or both of accuracy and precision can become important.

If you've been riding long enough to use old-school, wired cyclocomputers with wheel magnets, you've witness a first-hand example. You and your buddies all complete the same ride, computer #1 says it was 20 miles, computer #2 says 22 miles, computer #3 says it was 19.5 miles, and so on. With the now widespread adoption of GPS-based cyclecomputers, riders in groups now find a much higher degree of both accuracy and precision when measuring ride distance - multiple riders are more likely to have similar results, and the results are more likely to be repeatable when you ride the same route in the future.

Power meters

Let's say you can average 200 watts over your favorite 10 minute climb, using a power meter crankset, or perhaps a set of power meter pedals. You train all Summer, improving your fitness, and find that you can now average 400 watts instead over the same climb. You have doubled your power output!

Now assume your power meter has an error - it's not accurate - and the real power values are actually 180 and 360 watts. Does that matter? Maybe not, as long as the results are consistent and repeatable, because you compare yourself only with your own prior measurements.

Perhaps you've undertaken a training plan that requires you to ride at a prescribed percentage of your functional threshold power. Again, as long as you continue to use the same power meter, and it provides consistent (precise) results, it may not matter that it's inaccurate.

Incorrect power meter data could make you overtrain, or undertrain

As soon as you introduce other riders or measurement devices, however, accuracy becomes critical. Let's say you're working with a coach who wants your team to do a group training ride at 200 watts for an hour, but your power meter actually measures 50 watts low. Whoops! You'll have to work much harder than intended in order to complete the prescribed workout compared with the other riders in the group.

This is also important if you own multiple power meters and have them installed on different bikes. For example, perhaps you've got a crank-based power meter on a road bike, but use SPD-compatible power meter pedals on a cyclocross or mountain bike. Accuracy suddenly becomes very important, otherwise you could be producing the exact same power output, but get wildly different measurements when you switch bikes!

Similarly, if you're using an online training tool like Zwift which relies on measuring the rider's power, you could have a big advantage (or disadvantage!) if your power meter is inaccurate.

Confirming power meter accuracy

How would you validate this data to determine accuracy? By comparing your own power meter with other measurement devices, for example, using both power meter pedals and a power meter crank on the same bicycle at the same time, then reviewing the results. We'd expect two precise devices to show the same peaks and valleys at the same points during repeated workouts, and the gap, if any, between the measured power of the two devices gives us insight as to their relative accuracy.

Reviewer DC Rainmaker illustrates this approach in his discussion of accuracy challenges with Shimano's power meter crankset - put a bike with a power crank and power pedals onto an indoor trainer which itself measures power, then compare the recording each of the three devices made of the same ride.

Hopefully they're all very similar, but if they aren't (as in the case of the Shimano crank), you'll need to think about what, exactly, the data means. Depending on your intended use of the product, it might be a dealbreaker, or just an interesting footnote.

    Torque wrenches

    If you're working on high-end frames and components (read: carbon) you're likely using a torque wrench. If not, you should be! A torque wrench is a great investment for riders who work on their own bikes.

    An accurate torque wrench must be close to the standard. 15 Nm of force on your torque wrench should be almost the same as 15 Nm of force measured with a much more sophisticated device in a lab. And of course, a quality torque wrench should be precise, producing repeatable results. If you set a torque wrench to 12 Nm, but it actually applies 20 Nm of force, you could damage fragile carbon components by over tightening, or the opposite could also be true - riding around with loose parts that you think are tight.

    You shouldn't drop your torque wrench onto the workbench or floor, because doing so can affect its accuracy. Instead, set it down gently, and store it in a protective case if possible. Better tools can also be re-calibrated by returning them to the manufacturer, something you may want to take advantage of periodically, to make sure your torque wrench is still reading correctly.

    Tire pressure

    I have previously stated that the gauge on most floor pumps isn't good enough to set tire pressure for cyclocross and gravel racing. Why?

    Take a look at the gauge on most floor pumps, and zoom in on the "hash marks" - some cheap models are only marked with a label every 5 PSI, leaving the rider to guess or estimate the PSI when the needle gauge falls between the hash marks.

    Accuracy and precision may be less important in high tire pressure applications

    If you are riding a road tire at 100 PSI, it is possible that neither accuracy, nor precision are all that important. You've got a basic pump and you end up riding 103 PSI this weekend, 100 PSI next weekend, 98 PSI the weekend after, then 96 PSI. Can you tell the difference? Most riders cannot at those pressures. The basic pump will do the job, because even a 10% error can be hard to feel at higher pressures. 

    Accuracy and precision are both more important in low pressure applications, like cyclocross or gravel

    At lower pressures, a 10% error is huge! If we aim to put 20 PSI in a front and rear set of cyclocross tires, we might actually end up with 18 PSI rear and 22 PSI front - that's a huge difference that even casual riders can feel. Katie Compton previously stated she could feel 1/2 PSI differences, for example. I can't, but I can sure feel a 4 PSI difference.

    In disciplines like cyclocross and gravel racing, small changes in tire pressure can make a huge impact on the way your bike handles and feels. 10% is way too much error in this application.

    To measure tire pressure, we want our inflation to be both accurate, and precise - so it's repeatable from race to race. That's why I suggest using a handheld gauge with a higher degree of precision (more decimals) like Meiser. They've got one gauge, for example, that maxes out at 30 PSI, with a hashmarks between each PSI increment, i.e. the needle points at 9, 9.5, 10, 10.5, 11 PSI... more than precise enough for the world's best riders. A gauge of this style is good for cyclocross, gravel, and mountain bikes.

    Speaking of not dropping your tools, as with a torque wrench, don't drop your Meiser gauge. They're fairly priced and a good value, but I've broken two by dropping them :(

    Riders who are at home with access to an air compressor may want to connect it to an inflator head with a digital gauge, like the Prestacycle Prestaflator Pro Digital. That one measures down to 1/10th of one PSI. The Topeak digital pressure gauge is another good option for riders who want a portable tool to take to races or drop in a pocket.

    Back to blog