As an experiment I started taking regular chain stretch measurements, from when I first put on a chain until it is worn out. A Park CC-2 was used and each measurement was done in triplicate, then averaged to improve precision. Measured chain stretch was then plotted against the total riding time at the time of the chain stretch measurements (Figure 1).

These measurements came under real world commuting conditions on a 2x10 speed drive train, with a constant pacing over the same mixed terrain (35% gravel, 65% asphalt). The chain was wiped clean, re-lubed and wiped every 50-100 km. A Shimano XT chain was used.

Question 1

In the right panel of Figure 1 (logarithmic), it is clear that there are two distinct wear rates/phases, a slower rate until a ride duration of roughly 40 hours, where a clear change in the pattern occurs.

Does anyone know the mechanisms driving the two different wear rates?

My suspicions is that the change relates to wearing through one material type into another, but it would be nice to get confirmation from people who know more.

Question 2

The wear pattern observed in real world testing (Figure 1; left panel) appears to be quite different from the patterns I have seen in posted laboratory testing (Figure 2), which used a machine to run, dirty and lube the chain.

Note: Both axes in Figure 2 are plotted on a linear scale so it should be compared to Figure 1, left panel, which uses the same figure scaling.

In real world testing, once into the second wear phase, the wear rate appears to be quite linear when plotted on the linear time scale (Figure 1; left panel). This differs from laboratory testing that showed a distinct curvature with duration for the entirety of the observation period (Figure 2).

Does anyone know the reason for the different observed patterns in wear for real world testing versus laboratory testing?

Figure 1. Chain stretch plotted against duration over the course of the life of a single chain (Shimano XT). Left panel has been plotted on the linear scale, while x-axis on the right panel displayed in a log10 scale. Shading indicated 95% confidence band. Horizontal line indicates suggested replacement point for 10 speed chains. chain stretch plot

Note: Patterns remain essentially unchanged if plotted against distance, instead of duration, due to constant pacing in the real world test.

Figure 2: Chain wear patterns from laboratory testing (accessed Feb 26, 2016 but now offline). Whipperman Test

Note: These results were found online. I did not participate in this study and I do not have the original data.

  • @andy256 the XT chain (CN-HG95) package doesn't say, but I assume so, the package claims SIL-TEC, what ever that is. Some of the chains in Fig 2 should be plated (e.g., KMC), but don't show the same wear pattern as Fig 1.
    – Rider_X
    Commented Jun 18, 2016 at 0:07
  • 1
    Great work! Is the chain plated, eg with nickel? (Sorry, the comments are out of order now :-)
    – andy256
    Commented Jun 18, 2016 at 0:09
  • 1
    Without anything to back it up, my guess is Factory applied lube has a lot to do with it, its much more effective than applied lube, after 40 hours its all gone. Its well known its best not to clean new chains so you do not strip the factory lube.
    – mattnz
    Commented Jun 18, 2016 at 3:25
  • @mattnz - shimano factory lube on high end chains like the HG95 is very light. This chain was making no lube noises within the first 20 km. So I started my wipe/lube/wipe protocol after the first hour. KMC chains seem to use a proper baked paraffin wax that lasts 500 km. Really depends on the manufacturer and chain model I suspect.
    – Rider_X
    Commented Jun 18, 2016 at 3:39
  • 1
    Shimano's product information claims that "Durability is assured through a chromising treatment on link pins, and the heat treatment of rollers, pins, and plates". One thing to suspect is that the chain starts wearing faster after chrome has worn out. I'll leave it to a chemist to figure out a test that could check if the surface of worn pin is chrome or steel.
    – ojs
    Commented Jun 18, 2016 at 11:16

2 Answers 2


My suspicions is that the change relates to wearing through one material type into another, but it would be nice to get confirmation from people who know more.

Another hypothesis is that as the pin-to-link mating surfaces wear, they create an increasing gap for the entry of dirt. Once dirt gets between the parts, wear accelerates due to a "wet sanding" action. At first smaller particles are able to enter, then larger ones as the clearance increases.

Eventual displacement of the fairly heavy-bodied factory lubricant might be a factor.

Exaggerated diagrams of the idea follow. Here is the state of a new chain: the pin is firmly pressed into the link with no clearance:

enter image description here

The arrows indicate the direction on of the stress under load. Then as the link and pin wear, there must appear a gap opposite to the where the load presses the parts together:

enter image description here

The "game changes" at this point. Of course the dirt works its way around the pin as the linkage rotates, and passes through sections of the drive cycle where it is not under load.

  • It might not even be particles getting in - the bits that wear away are probably still there and they will help wear the remaining chain.
    – Nuі
    Commented Jul 5, 2016 at 22:23

I've answered https://bicycles.stackexchange.com/a/40942/19705 and it got me thinking about this two-phase wear.

Perhaps (and this is only a supposition) at new the chain meshes nicely with the teeth of the cassette and chainring. So your 48 tooth chainring has ~24 teeth in contact with the chain, and the load is shared between multiple teeth.

Same at the back where you're in 11 tooth cog, so its contacting 4~5 teeth solidly. (fewer because the jockey wheel of the RD mech prevents a 180 degree wrap, only 3/8 of the rear cassette cog is in mesh with chain.)

This is the first stage of chain wear, the linear change. The chain moves around the chainring at the same rotational speed.

At the knee-curve in the graphs, a subtle change happens. The chain no longer contacts as many teeth cleanly. Instead of 24 teeth in the front, the microscopic wear adds up so that slightly fewer teeth/links are sharing the same load. This increase in load wears those teeth faster and stresses that chain roller more, accellerating the wear.

The extreme case is where only one tooth and one roller is carrying the full load for an instant. Upshot of this is that the chainring is now moving faster than the chain, by half a chainlink length per revolution. This is the second phase of chain wear, and the chain is moving slower than the chainring.

This difference in speed adds another source of friction, which is an added wear factor, and increases the loss of material from the chain.

The extreme case of this is when the chain slips off the teeth of the chainring, which clatters along under the chain until rider decreases pressure at the end of the power stroke. Your chainring did 1/4 of a rev and the chain didn't move much at all.

Why is the lab result so much different to the real-world result?

Its not. Looking again at the lab graph there's a distinct turn at 40-55 hours for ~11 of the 19 tested chains, and 3 of them were already off the chart by that time.

As a corollary , I suggest a follow up experiment - a distributed experiment by SE users.

Next time you change your chain, please start a routine measurement of chain wear using whatever tool you have. Otherwise ride normally, don't do anything different in your riding style.

Personally, I intent to use a vernier caliper on 10 specific links forward and backward from the master link (not including the master joiner) and will do this on a weekly basis. This length will be correlated with hours of riding that week, and distance travelled that week.

Results should be shared with @Rider_X periodically (6 monthly?)

How's that sound?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.