Reader Story: How Much Traction Does a New Tire Buy You?
Pinkbike reader Matt Morrison has continued to run experiments with his data acquisition system, following up on his attempts to determine whether pedal kickback actually happens. This time around, the goal was to figure out how much of a difference there is between a fresh tire versus a worn out tire. - Mike Kazimer
That first ride with a fresh tire is almost as good as new bike day. Well, not quite, but you get the idea. I’ve been running the same rear tire on my Stumpy Evo for longer than I can remember. It has survived the abuse of countless rides in Santa Cruz, endless laps in Whistler, and even unfriendly cacti on the Arizona Trail. A few weeks ago, a friend took pity on me and gave me a brand new tire. This prompted me to ask myself, How much traction does this new tire actually buy me? And what does better traction even mean?
I took this opportunity to run a little experiment. With my data acquisition system and a bit of code, I devised a way to measure traction loss events while on the trail. I did a few final laps with my old tire before swapping it out for new rubber. I was careful to ensure that all aspects of the bike setup, like tire pressure and suspension setup, were held constant between runs except for the single variable of the rear tire.
I’ve long thought traction was simple: you have it until you don’t! But this experiment taught me that it is more complicated than that.
And one last thing: definitely don’t buy a new tire after watching this. If it holds air and keeps you rolling, then your current one is just fine. Besides, didn’t it seem like I was having a little more fun on the old tire anyway?
![photo]()
Cass Labs
That first ride with a fresh tire is almost as good as new bike day. Well, not quite, but you get the idea. I’ve been running the same rear tire on my Stumpy Evo for longer than I can remember. It has survived the abuse of countless rides in Santa Cruz, endless laps in Whistler, and even unfriendly cacti on the Arizona Trail. A few weeks ago, a friend took pity on me and gave me a brand new tire. This prompted me to ask myself, How much traction does this new tire actually buy me? And what does better traction even mean?
I took this opportunity to run a little experiment. With my data acquisition system and a bit of code, I devised a way to measure traction loss events while on the trail. I did a few final laps with my old tire before swapping it out for new rubber. I was careful to ensure that all aspects of the bike setup, like tire pressure and suspension setup, were held constant between runs except for the single variable of the rear tire.
I’ve long thought traction was simple: you have it until you don’t! But this experiment taught me that it is more complicated than that.
And one last thing: definitely don’t buy a new tire after watching this. If it holds air and keeps you rolling, then your current one is just fine. Besides, didn’t it seem like I was having a little more fun on the old tire anyway?
Farewell, old friend.
Cass Labs
Author Info:
Must Read This Week
ITS A GOOD TIRE, I WILL FIND A USE FOR ITTTTTT!!!!!!!
I only have three bikes, but I definitely have more than a year's supply with just the tires that are on the bikes.
But controlled slippage is super fun.
slow or low angle -> whatever
I’m curious if the same data tells us about corner slip vs straight line slip? Straight line traction means I can brake later and make the corner. Corner traction means I don’t break my collar bone.
Gonna stick with the idea that there are tons of good ways to save money, cheaping out on tires is not one of them.
In fact, it kind of invalidates the data slightly if it's not the same line every time. Just being on a slightly different line, especially on a camber where from the wheel's perspective you're leaning to turn up the camber, is going to change both the traction and the wheel paths. And since relative wheel speed is the only data, any change in wheel path already changes the difference in wheel speeds regardless of traction. That changes the inputs to the data, makes it useless.
The biggest issue that is the most noticeable and the reason I swap tires... is because on steep tech climbs (have a lot where I live), the lack of grip is very noticeable and becomes super frustrating. For me the centre goes first, makes tech climbs shitty. Especially noticeable as we get into fall riding with moist greasy rocks and roots.
Fast non tech descents (flow) are generally faster... dead tires just roll so fast!
Oooh, I see we're speaking moistly... that's it, one more time please, and make it reeeaaal slooow.
I also ride tyres to way more bald than the old tyre in this vid..
Is the underlying assumption that a used tyre requires a higher slip ratio to generate mu (I'm not sure that the slip at which peak mu occurs should change, but maybe it does), or is it that if there is not enough grip the rider panics and in the vain attempt to slow down they grab a fist full of brake and lock up without actually improving the deceleration?
If it's the latter then you're measuring the level of uncontrolled panic braking that occurs with a worn tyre, which is a cool metric.
Since you have the high resolution wheel speed measurement, you could measure the former: Straight line braking on a consistent gradient/surface, don't touch the front brake, so this axle can be used purely for bike speed measurement/deceleration. Then try braking with different amounts of rear brake pressure, keeping your body in a similar position. It should be possible to plot deceleration Vs slip for both new and old tyres. Then with some assumptions about CoG height and measurement of the gradient you could calculate weight transfer and vertical tyre load. Then your graph could be converted to tyre mu Vs slip ratio.
Still, with this rudimentary setup it should still be possible to calculate straight line traction, which nevertheless would be very cool result indeed to discover the % grip loss longitudinally.
"A common alternative formulation of this process goes as follows:
Compute from the observations the observed value tobs of the test statistic T.
Calculate the p-value. This is the probability, under the null hypothesis, of sampling a test statistic at least as extreme as that which was observed (the maximal probability of that event, if the hypothesis is composite).
Reject the null hypothesis, in favor of the alternative hypothesis, if and only if the p-value is less than (or equal to) the significance level (the selected probability) threshold (α), for example 0.05 or 0.01."
And so as the p-value is higher with a 1 sided t-test, it would pass a higher significance test.
So the lower the p value, the higher the confidence that your differences are not random nut come from two different distributions.
When it's not random, the difference has a meaning, in other words, it is significant. How low the probability has to be before you decide it's low enough, depends on the field you are in. 5% is enough for medicine, but doesn't even get close in particle physics.
Of course, that is just statistics. Systematic errors, poor design of the experiment and wrong assumptions about the data are independent from this.
www.ncbi.nlm.nih.gov/pmc/articles/PMC4635100
That’s a simple example.
link.springer.com/article/10.1007/s10654-016-0149-3
Thats a longer treatment. It’s not semantic. The t test 100% does not give you probabilities that your scientific hypothesis is or is not supported by the data. It tells you if the t stat ( a measure of difference in terms of variance) is large enough to maybe indicate difference, and the p value says if the difference is likely to be sampling error or due to a true, underlaying difference in populations sampled. P = 0.06 is not closer to worn tires have less traction than P = 0.01. There’s no such thing as more significant or close to significant.
To answer your comment: a t test certainly can give you probabilities that your hypothesis is *supported* by the data. If (and even in real scientific articles that is a big if) your measured quantity is directly related to your scientific hypothesis. Even then, a hypothesis that is supported by the data doesn't have to be true. It's just not wrong in a way that would produce different data. And fyi, p=0.06 is indeed not closer to supporting the hypothesis than p=0.01, you have that right, because lower p values indicate a *higher* statistical significance.
The papers that you link have to be seen in the context of a trend in certain fields of science, where the focus of researchers has shifted too much to only concentrating on finding differences in datasets that have low p values, without paying enough attention to what those numbers actually tell you about the thing you wanted to investigate in the first place.
E.g. in this case, if the hypothesis is: "rear wheel angular velocity that is much smaller than front wheel angular velocity is more likely with worn tires than with fresh ones" , it can be reasonably well tested with a statistical comparison between two data sets of relative velocities. Of course, if you take tighter turns, your relative angular velocity also goes down so that doesn't prove much about traction.
Let's put things into perspective here. This guy is doing a t-test with a sample size of 3. Good luck getting significance on anything that isn't bloody obvious beforehand with that sample size. This is a nice piece of quasi-scientific entertainment. That's fine. It's fun to watch.
But if we want to do a solid investigation of tire traction, we should not even be discussing p values here. This is engineering, not medicine. You're not limited by the number of healthy volunteers that will take your untested medicine. Go find a long straight fireroad and do a couple hundred braking exercises per tire with rear wheel brake only. Quantify slip with high enough time resolution to actually see the slip/wheel lock up events. Quantify deceleration. Show me histograms, not p values.
en.wikipedia.org/wiki/Threshold_braking
"Braking beyond the slipping point causes the tire to slide and the frictional adhesion between the tire and driving surface is reduced. The aim of threshold braking is to keep the amount of tire slip at the optimal amount, the value that produces the maximum frictional, and thus braking force. When wheels are slipping significantly (kinetic friction), the amount of friction available for braking is typically substantially less than when the wheels are not slipping (static friction), thereby reducing the braking force and eliminating steering ability. Peak friction occurs between the static and dynamic endpoints, and this is the point that threshold braking tries to maintain.[2]"
1. Facial expressions of people as you ride past with a bike with headset vs non headset routed cables.
2. Speed difference with position of the SaddleSpur seat: perfect, too far back, too far forward and backwards
3. Blind test of sram vs shimano vs... brakes looking at power and modulation.
4. Speed test of 180 droppers vs 240mm droppers since all the PB staff seem to insist these " short" droppers make all our bikes un-rideable.
Anyone notice more flats with a more worn tire?
just ride your bike, have fun, if it’s wrecked and full of missing nobs change it if you can be arsed. If not just crack on.
Personally I have a few kicking around but not loads that change depending on conditions but other than that then they stay as the lay until a rip a hole to big to plug and then a new one goes on. I’m not a pro, don’t get them for free and don’t go fast enough to worried too much about a little wear.
I switched to an asagai, and found that I could break and corner harder than a DHR2, but noticed - switching back to the DHR2 - I could not only have more fun, but match my performance by trying harder/focusing.
All wheel drifts with same tire front and rear for life!
your mileage may vary; I'm not a terribly skilled rider, and I'm not focusing or racing for tens of minutes at a time.
Magic Mary rear is worth trying it in winter conditions... and strangely the rolling resistance is much better than the DHRII.
The MM rolls faster, grips and corners better, lasts longer and looks awesome... what else?!!
Ignore the statistical numbers that I got from applying an unproven formula for a singular situation... Sure, ok.
I also think the findings basically only apply for the surface conditions tested. For example increasing/reducing tread depth theoretically affects grip on hard surfaces vs soft soil in opposite ways.
That said, who reports p-values as %? And describing something like p=0.17 as close to statistical significance is playing very fast and loose with statistics.