How to Test Tire Performance

In the 15 years of Bicycle Quarterly, one of our discoveries has been that testing bicycle performance isn’t easy, and that taking shortcuts often has led to erroneous conclusions.

Carefully designed tests that replicate what happens when real cyclists ride on real roads have allowed Bicycle Quarterly to debunk several myths. Certainly, the biggest change in our understanding of bicycles has been about tires.

Tires, more than anything else, change the performance, feel and comfort of your bike. We now know that fast tires can increase your on-the-road speed by 10% or more. But how do we know which tires are fast?


Lab testing is the most common way to test tire performance, usually on a steel drum (above). In the past, these steel drums were smooth. Now the testers have added some texture to simulate the roughness of the road surface. Unfortunately, that doesn’t address the fundamental flaws of drum testing:

1. The curved drum pushes deep into the tire

Since the drum is convex, it pushes deep into the tire, unlike a real road, which is flat. The more supple the tire, the deeper the drum pushes. This makes the tire flex more, which absorbs more energy. That is why a stiff tire performs well on the drum, and a supple tire does not. We know that the opposite is the case on real roads.


Increasing the tire pressure also makes the tire harder, and so the drum won’t push as far into the tire. That is one reason why drum tests show higher pressures rolling much faster (above). According to this data, increasing your tire pressure from 60 to 120 psi (4.1 to 8.3 bar) reduces the resistance by 30%!

Bicycle Quarterly‘s real-road testing (above) has shown that the opposite is true, especially for supple tires: They roll slower at 100-120 psi than at lower pressures. (Higher power = slower.)

This problem with drum tests has been recognized for a long time. There is a way around this problem, but it’s very expensive: Make the drum so large that it’s barely convex. One of the best drum testing rigs is in Japan, and from what I’ve heard, it measures about 7 feet in diameter. That means it’ll push into the tire much less, and thus it won’t make stiff tires seem faster than they really are.

On the other end of the spectrum, you have efforts to measure tire performance on small-diameter rollers,  like those used for training. That will always be futile: Anybody who has ridden on rollers knows how high the resistance is, because the rollers push so deep into the tires. And if you want to increase the resistance further, you just let some air out of the tires…

Not surprisingly, tests on small-diameter rollers show the ultra-supple Compass ‘Extralight’ casing rolling not much faster than the ‘Standard’ version. On real roads, the performance difference between the two is quite noticeable.

TOUR magazine in Germany has designed a test rig that eliminates the problems associated with the convex drums: Two wheels carry weights that are off-center, so they rock the wheels like a pendulum. This test rig rolls back and forth on a flat surface. You could even use it to test on real roads. The longer the test rig rocks from side to side, the lower the tires’ rolling resistance.

This test showed the Compass Bon Jon Pass as the fourth-fastest tire they’ve ever tested. (‘Rollwiderstand’ means ‘rolling resistance;’ the dark bars are for ‘rough asphalt;’ the light ones for ‘smooth asphalt;’ to convert the pressure from bar to psi, multiply by 14.5)

It’s interesting to compare the same tires – Compass Bon Jon Pass 700C x 35 mm tires (standard casing) vs. Continental 4000 S II in two tests:

  • tested on a steel drum:
    • Compass (6 bar / 90 psi): 15.8 W
    • Continental (7 bar / 100 psi): 12.9 W
    • Conti has 18% lower rolling resistance.
  • TOUR magazine used their rocking test rig:
    • Compass: 17 W
    • Continental: 17.5 W
    • Conti has 3% higher rolling resistance.
    • Compass is fourth-fastest of all tires TOUR tested (graphic above).

It’s clear that the drum test disadvantages a supple tire – the stiffer Conti performs much better. Adding to the confusion, gets lower resistance values than TOUR – it should be the other way around with the drum pushing into the tire.

There is another odd thing: TOUR shows the wider Compass tire in 4th place on the smooth road surface, but in 5th place on the rough surface, where it gets beaten by the narrower Conti rolling at higher pressure. That isn’t how it works in the real world, where the advantages of wider tires and lower pressures are greatest on rougher roads. That brings us to the second problem of these lab tests:

2. No rider on the bike


Without a rider, you have no significant suspension losses. Suspension losses are the energy that is absorbed when vibrations cause friction between the tissues of the rider’s body. Without a rider, there is nowhere to absorb the energy – steel weights don’t behave like human tissue.

Without suspension losses:

  • vibrations wouldn’t slow you down.
  • wider tires would be slower than narrow ones.
  • higher tire pressure would make your bike faster.

On the road, with a rider on board, all these statements are false – because suspension losses absorb energy, and reducing suspension losses is key to making a bike go faster. Understanding suspension losses has revolutionized our understanding of tire performance. It’s the underpinning of the ‘wide tire revolution.’

The lab tests described above are like a return to the last century, when we all ‘knew’ that narrow tires rolled faster because they could run at higher pressures. So we ran 19 mm tires (above) and inflated them rock-hard for optimum performance. That was long ago – when did you last see a short-reach racing brake with so much tire clearance?

Today, even professional racers run 25 mm tire at 80 psi. They have found that this is faster, no matter what the steel drum tests say. Racers have concluded: When tests don’t replicate the real world, they aren’t of much use.

At least TOUR‘s test rig gives us some indication about the energy absorption in the casing. It neglects one half of the equation – the suspension losses – but it’s useful if we understand its limitations. On the other hand, tests on small-diameter drums are just misleading – because if you design a tire to perform well in these drum tests, it’ll have a stiff casing and ultra-high pressures. And that means it won’t perform well on real roads.

A better lab test?

Is there a way to design a realistic lab test for tire performance? After all, Bicycle Quarterly‘s test procedures – testing only on totally calm days; when temperatures are constant; with a rider who has trained to keep the same position for lap after lap – are fine if you are doing scientific research. But they are not feasible for commercial applications, where you need to be able to just mount a tire on a wheel, take it to the lab, and get an immediate reading of its performance – without having to wait until the weather is right, the wind has died down, and the temperature is constant.

At Bicycle Quarterly, we’ve been thinking about this. Current drum tests load the tire with metal weights that don’t absorb much energy as they vibrate. Is there another material that behaves similar to a human body? ‘Ballistic gelatin’ is used to simulate gunshot wounds in human tissue. It closely simulates the density and viscosity of human tissue. Using a material like that to weigh down the wheel might simulate the suspension losses.

Suspension losses vary with speed (higher/lower vibration frequency), so TOUR‘s rocking rig probably would not work – it simply moves too slowly to replicate suspension losses at normal cycling speeds.

That brings us back to the steel drums. You’d have to make the drum huge to reduce the problems with the convex surface. The drum surface itself would have to be a true replica of actual pavement, not just a diamond tread. You’d probably want to map a bunch of road surfaces with a laser and then use EDM (electrical discharge machining) to engrave an ‘average’ road surface into the steel drum surface. You could make interchangeable plates covering the drum with several road surfaces that feature different roughnesses. And why not a gravel road, too?


To validate your test rig, you’d take a fast, a middling and a slow tire, and test them on the road, just like Bicycle Quarterly has done. If your drum test results match those on the real road, then you can be confident that they replicate real-world conditions.

Back to Real-Road Testing


As you can see, making a useful test rig is a huge undertaking, which is probably why nobody has done it yet. For now, tire companies continue to develop their tires with the help of simple steel drum tests. That may be the reason why they don’t offer their supple high-performance models in truly wide versions: The steel drum tests indicate that you lose performance quickly as you run tires at lower pressures. And since supple, wide tires cannot support high pressures, steel drum tests suggest that wider tires should strong and not supple.

At Bicycle Quarterly, we’ll continue to test tires on real roads. To get good results, we can’t just put a power meter on a bike and go for a ride, then change the tires and repeat. We must keep the conditions the same for all tests. First, this means testing in a controlled setting, like a track. Second, we must control the variables tightly: test only on days with no wind and constant temperature, test each tire multiple times, and do a rigorous statistical analysis of the results.

The statistics are important, because there always will be some ‘noise’ – even in a lab test, because the tire warms up the longer you run it on the machines. The statistical analysis shows where you are recording real differences between tires and where you just see ‘noise.’

After more than a decade of testing tires under real-world conditions, we can say with certainty:

  • Supple casings, more than anything else, determine the performance of your tires.
  • Wider tires roll as fast as narrow ones on smooth surfaces, and faster on rough ones.
  • Higher tire pressures don’t make the bike faster.

There is little doubt about these findings any longer – they’ve become widely accepted, even though the lab tests still haven’t caught up to the new science. But for us as riders, what matters is how well our tires perform on real roads, not on a steel drum.

Further reading:

About Jan Heine, Editor, Bicycle Quarterly

Spirited rides that zig-zag across mountain ranges. Bicycle Quarterly magazine and its sister company, Compass Cycles, that turns our research into high-performance components for real-world riders.
This entry was posted in Testing and Tech, Tires. Bookmark the permalink.

80 Responses to How to Test Tire Performance

  1. Stephen Poole says:

    Brilliant post! Lots of useful info pulled together in one place. Thanks very much Jan. 😉

  2. alderbanks says:

    A long decent on a washboard road is a punishing experience on low volume/high pressure tires. It may be hard to quantify how uncomfortable that can be, but most of us have experienced it.

  3. Brian Hanson says:

    Why not use a treadmill with a unicycle setup so a person (or ballistic gelatin) with sensors to measure the suspension forces/losses? You could even vary the treadmill surface to simulate smooth and rough roads. Seems like a cheaper solution.

    • I thought about your idea of testing on a treadmill. That would be an interesting idea, and it would eliminate the convex roller.

      BQ mechanic Cale here had a similar idea. You’d need a metal treadmill to simulate the hardness of the road – something like the track of a Caterpillar bulldozer. I think that is where you might run into trouble, because you need to measure the energy it takes to turn the track at a certain speed to calculate the resistance. And with so many joints, the resistance of the track itself might not be constant. You could calibrate the treadmill before each run, but in the end, it might be easier and cheaper to make a huge drum…

      • Staffan Widell says:

        I totally think that a treadmill with a real person on board the bike is the best option. To measure the resistance, simply put a chord from a fixed point in front of the thread mill with a meter that measure pull and connect the other end to the headset of the bike. As the rider is being “pulled” by the chord over the threadmill surface, you can measure things like posture, stiff or loose grip, sit or standing and get a reading from how much the bike pulls the chord.

      • Staffan Widell says:

        And, as addition to my previous post, you can disregard any trouble about measuring the power output via the threadmill. As long as it is moving at a fairly constant speed, it should pose no problem. Having a chord from the front would only allow to measure rolling resistance though but I bet that if you did the same setup but used a measuring chord from the back of the rider, and synchronised the data from a crank power meter, you should in theory get the correct data, even when pedaling. Maybe this set up could eventually “prove” the concept of planing? 🙂

      • One possibility may a treadmill similar to a Woodway, which uses slats instead of a track. If you could construct your own, you could make the slats replaceable to simulate different road surfaces. Another option may be a horizontal disc that could provide a flat surface for the tire, but a smaller diameter disc may also introduce a turning factor on the wheel.

      • I’ve been thinking about the treadmill more, and if you put a power meter on the bike (since you have a rider, who pedals!), you don’t need to worry about the friction in the treadmill…

  4. zigak says:

    Instead of using steel weights or gelatine on a big drum, just put a few pork legs in there instead.
    This brings an additional question – if muscles of a rider are contracted i.e. stiff, do they absorb less energy?

  5. Stephen Poole says:

    Another option might be to make something like a belt sander, with a rigid flat plate and flexible belt(s) with different surfaces. The plate wouldn’t need to be very big, just a little larger than the contact patches to be tested. There are smaller earth movers out there, so a (possibly used) track might not be completely silly either.

    A really big drum would need a really big lathe to make it, or else a “Compass style” laser or water jet cutter. 😉

  6. Frank Toman says:

    Here we are talking about tyres and no one is mentioning the Antelope Hill 700/55? C’mon already! Roll it out ; )

  7. Paul says:

    I’m not a scientist, but I can say with certainty that a bike which goes in a straight line is faster than one swerving back and forth (see modern carbon) if both riders in ssme gear at same cadence.
    Why do bicycle designers insist on bikes which, even with a pro atop them, make bikes which cannot hold a line?

    • That is one of the great mysteries. I’ve been thinking that perhaps the swerving serves as an additional energy input, like a skateboarder who weaves and keeps the skateboard moving without pushing off the ground, but that seems unlikely…

      Most of all, I really enjoy a bike that rides in a straight line, as if it was pulled toward the horizon by an invisible rubber band, yet it turns with precision when the road turns twisty. But that is the topic of another blog post…

      • Atle K says:

        Maybe the bike industries makes bicycles “responsive enough” (= exciting) to sell?
        A new customer tries it, finds it fun, agile and easy to turn in the parking spot, and – maybe it’s only me – gets annoyed after six hours on seat. So I got a Salsa Casserole who is more relaxed than normally. The next one will be custom, I want a longer and more relaxed frame.

        I think personal preference matters, somebody prefers a “exciting” bicycle who wants a lot of inputs, so it still feels fun after six hours. (and usable in a peloton), while others want it to be less agile.

        I had a trackbicke with short rake/trail, and the steering got almost unresponsive over 25km/t, I loved it. Used it as fixed gear and 3 gear IGH. Thight places made me more careful/slower than normally. (safety first 🙂

      • Paul says:

        I’m reminded of the builder of 3Rensho bikes who, when asked why his chainstays were always so long replied, “when my riders accelerate, I want them to go forward!”

    • Stephen Poole says:

      This is totally a “YMMV” thing – I have no problems going in a straight line with most bikes, carbon included, but had terrible problems with the Soma GR, which others have liked – to my surprise (and disbelief). *Nothing* fixes things for everyone, IMHO; individual testing is necessary. Saying pro cyclists can’t or don’t ride straight is not believable based on what I’ve observed – 60+km/h finish line sprints do involve some line changes or swerves, but I doubt control is lacking due to the bike. It’s similar watching runners or XC skiers at the Olympics – a lot of those guys basically collapse after the line due to exhaustion.

      • I think what Paul was referring to is the weaving of the front wheel you observe with many riders on modern racing bikes, including pros. However, as you point out, those riders and bikes perform very well in an all-out sprint. I am not strong enough a sprinter to tell you what happens when you push the pedals with 1000+ Watt. Perhaps the bikes designed for that are compromised everywhere else, but it isn’t a big deal, since the race is won or lost in those few seconds at the end.

  8. Monty says:

    HI Jan, thank you an excellent article. Yesterday I had an interesting tire experience. Myself and a friend did a long asphalt ascent and then hit a rough gravel descent. I was using 40mm 700c tires at 40 lbs and he was on 32mm 700c tires at about 70lbs. On that downhill and then pedaling on flat and up hill I left him far behind and yet he is a much stronger rider (we know from riding our roadies together). The only difference was the wider tires at lower pressure.
    In the real world I am totally convinced of your ‘truths’.

  9. Road tests to do it all, including aero.

  10. larryatcycleitalia says:

    Thanks for this. I’ve been singing this song for decades based mostly on personal experience with bicycles and motorcycles, but it’s nice to see some validation from others. There’s still work to do though as I was at a cycling event just last week where most of the others were still pumping their tires to ridiculous pressures while I was letting the air out of (borrowed, way-too-stiff carbon bike) mine!

  11. Balor says:

    “Since the drum is convex, it pushes deep into the tire, unlike a real road, which is flat. The more supple the tire, the deeper the drum pushes. This makes the tire flex more, which absorbs more energy. That is why a stiff tire performs well on the drum, and a supple tire does not. We know that the opposite is the case on real roads.”

    I get the argument from ‘rollers’, but I find it very unlikely that flexible tires are more vulnerable to this effect and not vice versa.

    Inflexible tires may not get deformed as much, but we are not testing tires with a certain tire drop, but at a certain pressure. Еxtra legere Compass tires do get more tire drop at same pressure than stiffer tires – even on smooth surface! They get deformed more… but actually don’t suffer from than as much, because they are, well, more supple!
    I’m not at all convinced unless you provide some concrete data with convincing test protocol to back up your claim.

    Argument about using a block of ballistic gel is solid, I’ve suggested it myself to Jarno. You can also use a, say, polyurethane spring instead of a weight with hysteresis tuned to that to a human body.

    However, you didn’t note *tire compounds* anywhere.
    As tests on SAME drum, same size and same pressure (and even rough same weigh) tires but different compounds tires show, compounds can account for nearly DOUBLE the rolling resistance, other effects notwithstanding.

    Drums don’t show absolute values reliably – they may overestimate rolling resistance compared to flat surface, and underestimate (or, more likely, completely ignore) suspension losses.

    But they can reliably show how similar-looking tires with different compounds may be drastically different in rolling resistance.
    You *know* that compound hysteresis is what actually causes rolling resistance (given no suspension losses). Not mentioning them is disingenuous.

    • On steel drums, completely different factors are important for performance than on real roads. Just witness that Jarno wrote on “the casing thickness doesn’t make a huge difference in the rolling resistance tests” That is totally the opposite of what we (and everybody else) observes on real roads.

      Conversely, in all of BQ‘s tires testing, we haven’t found tread compound to be a factor that stood out.

      On the drums, tread compound is very important, and harder rubber performs much better – for the same reason higher pressures work better. On real roads, it’s the opposite. I recall how at the start of our tire tests, we looked at Tour’s test data (back then on a drum), and they showed the hard Michelin Pro2 Race as fast as the soft Vittoria Open CX Corsa. In our real-road testing, the Vittoria was way faster.

      So if you optimize your tread compound on a steel drum, you’ll end up with a slower tire on real roads. This means that testing the best rubber compound is difficult. We hope to do more testing on this in the future, and we’ll report on whether tread rubber compounds matter, and if yes, which are fastest.

      • morlamweb says:

        Conversely: keep the existing drum setups, and design tires to have high resistance on that test (supple sidewalls, low pressures, etc). By building a tire to “fail” the steel rollers or drum tests, then maybe you’d end up with a good real-world tire.

      • 😉 Unfortunately, that isn’t how science works… You could probably design a tires that performs poorly in both tests.

      • Balor says:

        Unfortunately, when you look at total accumulated drum test data, it does NOT prove your theory. At least it does not prove that it is significant.
        If you only look at Jarno’s test of “Bon Jon” vs “Marathon Almotion”, it might look credible, right.

        HOWEVER, when you look at tests of road tires – say, Schwalbe Durano Or Continental Gatorskin (notoriously stiff tires) – they are in the back of the pack.

        Tires like Corsa Speed or Turbo Cotton, that are very supple (yet narrow) tires are MUCH faster and are at the top. If drum tests favour stiff tires that much, why don’t we see Gatorskins beating Corsa Speed?

      • For narrow tires at high pressures, this is the case. But with wider tires, drum tests no longer work. Otherwise, the ‘wide tire revolution’ would have happened decades ago. But as long as tire companies developed tires using drum tests, it seemed that wide tires rolled faster if they were stiffer and running at higher pressures.

    • Apart from anything else, I don’t understand why you (I mean Balor) “are not testing tires with a certain tire drop, but at a certain pressure”. I’d have thought it obvious that different width tyres require different pressures to run at their most efficient. So comparing them at the same pressures is doing some of them a disservice. Or do you compare them all across the same range of pressures (eg, in increments of 10 across 40–120 psi)? Or am I misunderstanding something?

      • I hadn’t thought about that, but you are right: If tire pressure affects the results – as it does on a drum test – then you should compare tires at the same tire drop, as that is how you’d run them on the road. In the end, it doesn’t really matter, as the test is flawed. All the real-road data shows that tire pressure doesn’t make a difference in performance.

      • Tom Anhalt says:

        “All the real-road data”, Jan? I think you mean just YOUR data. There’s plenty of other data which shows that for a given configuration, increasing tire pressure DOES reduce rolling resistance, up to the point the tires stiffen up and begin transmitting energy into the rest of the system.

      • I think we can reconcile those differences. It’s simply that supple and stiff tires behave differently.

        When we did our initial study, we did see a ‘break point’ for some of the stiffer tires. For example, the Rivendell Rolly-Poly got faster as we increased the pressure from 35 to 55 and then to 85 psi. Then the pressure leveled off: At 105 psi, it wasn’t any faster than at 85. This suggests that this tire’s break point is somewhere between 55 and 85 psi. (At least for our test road and the weight of our test rider.)

        And if you look at the Rubino in the graph above, there is a break point at 95 psi. If we had stopped testing at 110 psi, we’d get a graph that matches the ones you showed in your earlier comment. However, we tested beyond 110 psi, and we saw the resistance decrease again, indicating that the real world is more complicated than the neat graphs that you showed.

        With supple, wide tires, the break point was so low that it was meaningless. Even the (no-so-supple, but wide) Mitsuboshi 650B x 38 mm rolled as fast at 35 psi as it did at 55. It rolled slower at 25 psi, but that pressure is marginal for that tire –when you take a corner fast, the sidewalls collapse.

        What this suggests is that stiffer tires do benefit from an increase in pressure, but supple ones don’t. So if you want to optimize the performance of a slow tire, you might want to run them just at the ‘break point.’ But that is somewhat meaningless, because if you want a fast tire, you’ll always run a supple one, which doesn’t benefit from higher pressures.

      • Tom Anhalt says:

        Wait, so a Vittoria Corsa KS with a 320tpi (TRUE tpi) casing isn’t “supple”? Even the 25C Continental GP4000S II used in the Silca testing isn’t exactly a “stiff” tire.

      • It appears that the results of Silca’s testing don’t match the results we got from the similar Vittoria Open CX Corsa. Without seeing Silca’s original data and details of the testing procedures, I can’t comment on the their tests.

        However, I have a lot of confidence in our results:

        We tested on three surfaces (ranging from super-smooth to rough, but not bumpy, pavement), with two methodologies (roll-down and constant speed power meter), with multiple repeats, and with a robust statistical analysis.

        After our initial testing showed such revolutionary results, we had independent observers for the next round of tests, because we were concerned that somebody might accuse us of fabricating the data. Ten years ago, the idea that wider tires and lower pressures could be faster was still highly controversial!

        Before publishing the results, we had the articles peer-reviewed, not by hobbyists, but by scientists with degrees and careers in testing equipment: Jim Papadopoulos and Andreas Oehler. They had some suggestions about improving our methodologies, which we followed. The whole thing was a huge undertaking that took many months – waiting for perfect testing conditions, spending multiple days testing, analyzing results, testing more, writing the article, discussing with the reviewers, re-writing, etc.

        During my earlier career in geology, I published numerous articles in scientific journals, so I am very familiar with the requirements for a peer-reviewed study. We made sure ours exceeded academic standards. We thought about publishing our results in a ‘real’ academic journal – like Papadopoulos did with their study of the zero trail, zero gyro bike that was still stable – but we decided to invest the time into further research instead. We might still submit an article at some point, now that there is so much interest in the topic.

      • Vittoria Corsa KS with a 320tpi (TRUE tpi) casing

        The Vittoria certainly is a supple tire, but the ‘Italian’ way of counting TPI counts the threads of overlapping casing layers multiple times. With three casing layers overlapping on most tires, an ‘Italian’ 320 tpi casing would be 106.7 tpi by most other standards. In fact, if you made a true 320 tpi casing, it probably would be too fragile: When the tire flexes, one thread bears the biggest load. If that thread is too thin, it’ll break, and then the next and the next… You can see that when tires are run at too-low pressures for too long, and a cross-hatched pattern appears on the sidewalls.

  12. Preston R Grant says:

    Long ago, in my days as a marathon runner, I much preferred running on asphalt compared to concrete due noticeably less shock, even when wearing shoes designed to absorb shock. We have many runners in our area, and they mostly get off the concrete sidewalks, and run on the asphalt bike lanes. So there is a perceptible difference between the hardness of the two materials. Getting back to bikes, it seems that on an asphalt road, a skinny tire, pumped up rock hard, would depress (sink into) the asphalt surface more than a wider tire, thus slightly incresing rolling resistance. Of course this increase is obvious when going from any good paved surface to a dirt road, or across a grassy area, but I am not sure that it has been considered as a factor between concrete and asphalt.

    • I think you are onto something: I remember riding on a very hot day after moving to Texas, and seeing my 19 mm tires making visible tracks in the asphalt.

    • Gert says:

      In my experience with running (limited) the advantage of asphalt is grip. Meaning you loose less energy by sliding on the surface.
      According to this article the softness of asphalt compared to concrete has almost no influence

      It is however very noteworthy how much the Young modulus of asphalt varies with temperature. So it is possible to cut into it in Texas on a hot day with a narrow tire
      (A totally aside on this is how much asphalt temperature affects puncture likelihood)

      Back in the mid nineties Tour had an article on high profile aluminium rims, and how they were felt to be hard to ride. They stated that 95% of all suspension on a bicycle was in the tires, so the influence of the rim stiffness was unimportant.
      However due to different frequencies that rims and frames dampen compared to the frequencies the tires dampen, how the last 5% are dampened may very well be felt.
      So going back to running on asphalt. The softness of asphalt may very well only dampen very little compared to the shoe, but I may very well be felt beyond its numerical value

      • Regarding TOUR’s article on rim stiffness: I also used to believe that rim stiffness didn’t matter. Then we rode two bikes that had identical front ends – my Alex Singer and a TOEI test bike. Both had flexible fork blades, both had 700C wheels and the same tires. Unexpectedly, the TOEI felt a little harsher on a really rough stretch of road. We couldn’t figure it out, until we noticed that the TOEI had a taller ‘aero’ rim. We swapped front wheels, and then the Singer rode harsher. I was surprised!

  13. Dan Michael says:

    “When you stop pedaling, the rollers stop turning almost immediately. Compare that to how easily a bike coasts on a real road.” Is this a valid comparison? In the former case, the kinetic energy of only the rotating wheels and rollers is absorbed by the frictional forces. In the latter, the much larger kinetic energy of the rider / bicycle and wheel rotation is being resisted.

    • You make a good point, but the energy lost on the small rollers still is significant: Take another comparison: How long does a roller spin by itself after you stop turning it by hand, and how long does a bicycle wheel spin in the stand? Both turn for a significant time. Then you put both together, and they stop almost immediately.

      • Mx says:

        The main point is that of a physical inaccuracy in explaining the difference as you did: “When you stop pedaling, the rollers stop turning almost immediately. Compare that to how easily a bike coasts on a real road.” That’s just incorrect from a physics standpoint and tends to cast doubt in other pronouncements in the post.

        But I am reading with curiousity. The ballistic gel experiment should be revealing.

      • The issue of small rollers pushing into the tire has been recognized for a very long time. It’s not something new that we came up with. I was just trying to illustrate a well-known fact.

      • Tony says:

        Mx makes my point about casting doubts. Interesting food for thought, though. Keep it up!

      • If one has ever ridden rollers with a flywheel or a device to create additional inertia (such as TruTrainer rollers), they coast beautifully and that additional inertia makes them feel much more stable.

        That being said, I thought it was fairly common knowledge that the easiest way to increase resistance when riding rollers was to let air out of the tires. The increase in power required to hold a given speed can be significant.

        Great article, as always. It is important to understand the limitations of and assumptions embedded in any test.

      • I stand corrected – you make a convincing arugment: It’s probably the lack of inertia that makes the rollers stop so quickly. I’ll update the article.

  14. Tom Anhalt says:

    You say all of that…and yet there is data like this showing that small roller testing is completely valid for determining on-road Crr values (below the “breakpoint pressure”, of course:

    And this:

    • I respect all the effort you put into testing bicycle performance, so please don’t take any of this personally.

      You already pointed to the biggest issue – your graphs match only up to the ‘breakpoint.’ What happens at 115 psi that suddenly makes your test not work any longer. Your roller test shows higher pressures reduce rolling resistance into infinity, but on real roads, that isn’t the case.

      The fact that the graphs are so smooth and clear in itself tells you something: Real-world data is never so smooth and linear. Just look at the real-road tire pressure graph in the post above. But it also shows that the ‘breakpoint’ doesn’t lie anywhere near 115 psi, at least for supple tires. It’s much, much lower – somewhere between 60 and 85 psi for a 25 mm Vittoria CX. And below that break-point, the tires were so soft that they were almost impossible to ride. That makes the ‘break-point’ meaningless for real-world riding, because you’ll always be above the breakpoint.

      To validate your method, I’d want to see three tires tested on the rollers and on real roads (under carefully controlled conditions). Do we get the same ranking in both tests, at different pressures (i.e., fastest on the rollers is fastest on real roads)? Are the differences roughly the same (fastest tire x percent faster than slowest one, etc.)

      In fact, you could test the three tires we’ve tested both in the roll-down and on the track and see whether you can replicate those carefully controlled real-road tests in your setup. That would save you the hassle of designing a real-road test.

      Beyond that, the problems with small-diameter drums have been recognized for a long time. Jobst Brandt and Frank Berto talked about this way back. They were concerned about drums that are ‘only’ 60 cm or so in diameter, yet yours are almost an order of magnitude smaller yet.

    • Sam Atkinson says:

      The current situation is that both of you currently have data that appears to directly contradict the other’s argument. I don’t think either of you are going to convince each other of much unless you find major rank order discrepancy outliers between your methods, and then get together and compare to try and figure out what’s happening.

      • I’m reminded of the early days of our tire testing, when flame wars raged for several weeks on, because it just couldn’t be true that higher pressures made no difference in tire performance. For a while, there were two alternative universes – on the one hand the ‘established’ mainstream, who ‘knew’ that narrow tires and high pressures rolled faster; and the ‘alternative’ world who tested wide tires on real roads and found that they rolled just as fast. The two worlds sort of ignored each other.

        Now that the ‘alternative’ world of wide tires has entered the mainstream, the two worlds clash once more, because these new developments make no sense when tested with the old methods.

  15. Dirk van Rossum says:

    Tensing the muscles dampens the vibration in the body, but still absorbs energy. Should have no effect overall, it is is still energy and power lost. One could argue that it reduces the bounce of the tires (i.e. lift off) and keeps then contact patch more in contact. But how much does it affect rolling resistance or even aero drag (when you sit up slightly as a result) Alternatively consider when riding on a really bumpy road and getting out the saddle. Arguably then one expends even more energy because the legs are then supporting bouncing body mass against gravity as compared to a fairly hard link through your butt, frame and wheels. This would affect aerodynamics far more, subject to speed of course. I also thought that a good comparative test rig could be made up of a modified treadmill unicycle rig and balistic gel. Would be a good step forward. Agree also that the impact on aero drag with wider tires needs to be taken into account and clarified.

    • the impact on aero drag with wider tires needs to be taken into account and clarified.

      Bicycle Quarterly‘s real-road tests include the aero drag and all other resistances. We tested both 25 and 31 mm tires (actual width) with very similar tread patterns in the wind tunnel, and we found that the differences were too small to be statistically significant. When you look at the frontal area, which determines most of a cyclist’s wind resistance, even a 54 mm tire isn’t huge. Otherwise, carbon bikes with massive down tubes would be much less aero than skinny-tubed steel bikes.

      • Tom Anhalt says:

        “Otherwise, carbon bikes with massive down tubes would be much less aero than skinny-tubed steel bikes.” Ummm…they are. That’s easily measured.

        Also, aero drag isn’t just about frontal area…there’s form drag to account for as well, especially with non-zero apparent wind angles. That can start having significant effects if a tire is wider than an aero section rim next to it.

      • If there is an aero disadvantage, it’s small – otherwise, the guys I raced against who were on Cannondales never would have won a sprint!

        More seriously, our wind tunnel tests showed that, with a pedaling rider, a bike has such a high coefficient of drag (cx) and the air is so turbulent that the contribution of form drag is small. About 90% of aerodynamic drag is due to frontal area, which is why lowering your stem 20 mm gives you more advantage than a set of aero wheels. It’s also why most fairings don’t work – they just increase the frontal area.

  16. Richard Freeman says:

    What effect does mechanical suspension have on rolling resistance? If suspension loss is mitigated with supple tires, and a soft (supple?) human on board further reduces resistance, it logically follows that other suspension methods would lower rolling resistance too.

    Could something as simple a sprung saddle make a supple tire roll even faster?

    You once found planing in an aluminum softtail road bike. Could part of that perceived effect be lowered rolling resistance?

    • We tested a RockShox Ruby fork when looked at suspension losses. What surprised us that even on the smooth pavement that served as a reference, the suspension fork was faster than a stiff fork from a hybrid bike. However, a flexible steel fork was just as efficient, both on the smooth and on the rough. So yes, all the available data indicates that stiff forks are slower.

      • Richard Freeman says:

        What about the rear? Even a superlight steel frame is vertically rigid in the rear triangle. At the same time, the rear tire is more heavily loaded, so it has more resistance and therefore more to gain from bump absorption. Look at all the ways it’s been done – spring saddles, moving seatposts, Softride beams, softtail frames, hinged rear triangles, Trek IsoSpeed, the list goes on. How about inserting one of those into the next round of tire tests? Just one simple A/B test of the same bike and tire with different seatposts could be interesting.

  17. Andrew Cohen says:

    Just want to add to the chorus of thanks for all the work you guys do. As always, putting conclusions out there like you do invites criticism, but I feel like you guys are doing a great service with your real-world testing of components like tires, as well ass your honest and realistic bike tests.

  18. Perhaps I’m simplyfying things here, but surely a stiffer tyre would be typically run at proportionally lower pressure in order to obtain the same level of suspension benefit as that of a suppler one? Which leaves us only with the compound and its hysterisis afecting rolling resistance?

  19. Jason Miles says:

    I think it might be useful to separate rolling resistance and rider vibration absorption into two different lab tests. They really are different forms of energy loss so the most efficient vehicle might need to use different solutions to increase efficiency. For example riders may be more efficient with suspension and smaller tires on certain surfaces vs larger tires and no suspension. Developing one rolling test that simulates both losses for all surfaces, bikes, and riders will be impossible.

    Assuming the rider is not skipping or hopping from bump to bump, the Tour test should provide accurate information for rolling resistance at any speed. This is because ideally the rolling resistance should not depend on the vehicle speed. I am sure this ideal assumption will not be accurate for rough surfaces because skipping and bouncing are very common and will have a big influence on the true rolling resistance.

    It would be cool to measure the frequencies and amplitudes of the vibrations a rider experiences on different surfaces and different bikes and tires. Then you could develop a shaker table to measure how much energy is absorbed by the rider for different conditions. You could then use this info to optimize tire size and suspension.

  20. RickH says:

    There is a trade off between different types of tyres for a given purpose. Choose one that is most important to you. Then find a gentle slope that levels off to flat-ish. Roll down this slope using different pressures and choose which one feels right to you and check how far each pressure gets you.

    • That is what we did in our initial tests. Make sure to use a riding position that you can easily maintain and replicate – like your hands on the hoods with your arms locked. Make sure there is zero wind, and that the temperature doesn’t change.

  21. Fred Mulder says:

    How about a real-world test that eschews consistent conditions but aims instead for broad aggregate results? I’m thinking of a power metered bike ridden over a coarse rather than a velodrome. Something like a section of the Paris-Roubaix route that includes smooth as well as cobbled surfaces. 100 riders do the course on Tire A; 100 riders do the course on Tire B. Every rider is given a broad directive at the start, such as “keep the wattage at such-and-such” or “ride the coarse as fast as you can”. Similarly, a broad set of controls could be instituted, such as limiting all testing to dry summer conditions, or making sure the riders are reasonably skilled. The idea would be to see if any consistent trends emerge from the murk.

    • That is always an interesting idea, but you’d need a huge amount of data. Sort of like a study of medical treatments, where you need thousands of participants. It would be a good way to assess flat resistance, but it’s too involved to do.

  22. Mathieu van Rijswick says:

    The assertion that a test on a drum favors rigid tires significantly compared to supple tires is highly improbable. Think about a harsh tire, e.g. Marathon Plus, with an inner tube inflated to a very low pressure, say 0.1 bar. If you put a weight of 1 kg (10N) on top, you will get an indentation of several millimeters. So the stiffness of the tire carcass is in the order of 5 N/mm.
    In the tests on the Rolling Resistance site, the tires are inflated to 6-8 bars and tested on a drum of 60 cm diameter and a load of about 400N. The stiffness of tires at those pressures is in the order of 150-250 N/mm for a flat anvil and 100-120 N/mm for a round anvil of 8 cm radius (experimental data for 28 mm tires from Joesh Poertner, Silca).
    A 35 mm tire inflated to 3 bar has a stiffness (spring constant) of about 120 N/mm due to the inflation pressure.
    The stiffness of a tire is mainly due to the compressed air. At practical pressures, adding 1 bar increases the stiffness by about 30 N/mm. The stiffness of the carcass is very minor. The assertion that a 60 cm drum pushes deeper in a supple tire compared to a stiff tire is without evidence.

    • The stiffness of a tire is mainly due to the compressed air.

      The ‘idealized’ assumption above isn’t true. Frank Berto gave me all his research notes from his tests of tire drop, and they show that the tire drop (how much the inflated tire compresses under a given weight) varies widely based on the stiffness of the casing. You can also experience that on real roads: The 57 mm-wide Schwalbe tire (forget the model name) on our Jones was great at 18 psi, but Compass tire of similar width (54 mm) is virtually unrideable at that pressure.

      And since casing stiffness greatly contributes to the spring constant of the tire, a stiff and a supple tire will be impacted very differently by the convex drum surface.

    • Francisco says:

      It is easy to show, based on purely geometrical considerations, that a very supple tire rolling on a drum having the same diameter as the tire will have almost double the tire drop compared to running on a flat surface. The contact patch will be slightly shorter and slightly wider on the drum, and the casing deflections will be slightly more than double those generated by a flat surface.
      To use your terminology, employing a drum with a diameter of 60-70 cm halves the “stiffness” of the inflated tire compared to a flat surface.
      The deflections measured by Josh are not productive in this discussion as they relate to sharp edges (he was interested in the Paris-Roubaix cobbles). Therefore, whatever the contribution of the tire’s construction in its total “stiffness”, a 60-70 cm drum multiplies its relative contribution by a factor of two or more. Therefore, a drum does push deeper into a supple tire compared to a stiff tire.

      • Mathieu van Rijswick says:

        I am well aware of Josh Poertner’s history in Paris-Roubaix cobbles. However, his measurements in have a much wider relevance. The report contains tire stiffness data not only for a simulated cobble edge but also for a 16 cm diameter cylinder and for a flat surface. Obviously, a 60 cm drum is much more in-between the latter two. Moreover, his measurements were done at 3 different tire pressures.

        There is no disagreement about whether a drum pushes deeper into a tire compared to a flat surface. It does. The tire industry has for a long time established semi-empirical correction formulas for the drum’s curvature, e.g. ISO 28580 for car tires specified how to correct the RR measurement on a drum. The issue is whether the stiffness of tire casing is another important factor.

        What I want to point out is that you can easily measure the stiffness of the tire casing. That is, easy in comparison to a rolling resistance measurement. You can also easily measure the effect of the inflation pressure on tire stiffness. From Jan Heine’s comment I understand that typical stiffness data are already available in Frank Berto’s papers. The RR measurements done by Jarno Bierman on a rotating drum are done at a range of pressures. Instead of reading the RR value from his graph at the real pressure, you could read the RR value at a virtual pressure that corrects for the effect of the stiffness of the casing.
        E.g. if the casing stiffness is 5 N/mm and at common pressures the effect of an increase of 1 bar is 20 N/mm (typical data Poertner), you could read the RR graph at 0.25 bar lower than the real value. This method enables a fair and quantitative comparison of supple and stiff tires.

      • Thank you for the interesting comments. One issue when modeling tire stiffness is that you have both casing and tread to consider. How thick the tread is, how gradual its transition at the edge and how far it wraps around the tire all will affect the stiffness of the tire. It’s likely that these effects are not linear with pressure. So your calibration curve would be very complex.

        Often overlooked, but of crucial importance, is that all correlations must be proved by real-road measurements. Otherwise, you get things like the ultra-hard ‘energy saver’ tires for cars that in fact transmit more vibrations and thus apparently don’t save energy on real roads. (In cars, the suspension losses mostly heat up the shock absorbers, to the point where off-road racing trucks have watercooled shocks.)

  23. Even measuring the energy losses due to suspension loss, is this still neglecting the fatiguing effect of those vibrations? If I produce 100W at the pedals, and a few W are absorbed by the casing, that’s easy math. If I produce 100W and, say 10W is going back into suspension losses, I’m not only losing 10W of potential forward momentum, I’m experiencing something like the equivalent of 110W in fatigue. That’s 90W forward for ~110W fatigue, which is going to effect the total forward energy I can produce in a day’s ride.

    • Yes, the fatigue is another factor that we can’t easily measure. What was revolutionary in our tests of suspension losses was finding that even if you could ‘tough it out’ on tires rolling at high pressures, you’d go slower. But as you note, ‘toughing it out’ makes you tired, and then you’ll go doubly slower, because your resistance increases and your power goes down.

  24. maxutility says:

    I’m imagining a test rig where you tow a bike behind a vehicle. The tow rope could be hooked to an accurate scale to give you a resistance measurement. The bike could be ridden (coasting) by the same test subject multiple times without worrying about fatigue factors, but you would get the suspension loss data by having a human body on the bike. Hidden behind the pull vehicle you should avoid most wind issues, and it would be easy to do multiple runs for better data collection. It would be easy to vary speed, road surface conditions, tire pressure, etc. while keeping other factors stable.

    Of course this would not capture the effects of vibration on the rider’s fatigue levels which I suspect is a big factor making wider, softer tires faster in real world conditions.

  25. GAJett says:

    Why not do the testing using e-bikes? You would have a rider on board but, given a fixed position you eliminate uncertainties arising from pedalling styles and energy input variations, especially as the rider tires. Should be more straightforward to measure the wattage required to maintain various speeds than with a rider providing all the power. Would also allow data collection over longer distances, esp. at higher speeds.
    Just my 2¢ worth.

  26. Jan Koegel says:

    The reason why your tires are still not tested as the fastes so far and others with a similar flexible castings are better let me assume that the Panaracer compound is not competitive when it comes to rolling resistant. That must not be right, but from my own experience I know the Continental BC Compound as the fastest rubber so far. They claim their BC as 30% faster than their normal silica rubber and that demonstrate also the huge potential. Bottom line I value a supple ride much more than saving the last few watts. That’s way more meaningful – otherwise I would ride a time trail bike what I don’t:)

    • The big tire companies make much of the tread rubber compound, because that seems to matter in the drum tests that they all still use to develop their tires. On real roads, rubber compound matters much less. The fastest tires in our original tire tests were made by small companies, who don’t have any special tread rubber. The ‘high-tech’ tires from Conti, Michelin, and others all were much slower, because their casings were stiffer.

      Basically, you cannot design a wide tire that rolls fast both on real roads with a rider and rider-less on steel drums. You have to choose one over the other. We chose speed on real roads for our Compass tires…

  27. Stuart Fogg says:

    Using an actual bicycle with rider on a treadmill (or merry-go-round) you could measure the horizontal force required to keep the bicycle stationary rather than the power required to keep the surface moving. That would eliminate energy losses unrelated to the bicycle from the measurements.

Comments are closed.