Best Bike Split Questions

I am using best bike split for the first time for a 210 mile gravel race. I have a few questions I was hoping the group might be able to answer. I think I am entering information incorrectly because I am getting a speed and time I think is unrealistic, or I have an incredible amount of inefficiencies in my drive train somewhere. My FTP is 275 and I am using a IF of .6. Aero helmet, gravel bike with gravel tires. I am pasting some screenshots below. Is my CdA way off? I find it hard to believe I will be rolling at over 17 mph at 152 watts. I am hoping some of you more savvy riders with more experience using Best Bike Split may be able to help. Or I could contact them directly but, what fun is that.


My understanding is that BBS does not work very well for non-road surfaces, as the coefficient of rolling resistance can be highly variable, and difficult to know.

In your modeling above, my guess is the Crr is set too low - although I don’t know what it “should” be for the kind of gravel you are riding.

One way to estimate the Crr is to ride a short section of the actual course, and reverse engineer the Crr so that the BBS modeling matches what you did in real life (time, power, etc). But this will just be an estimate for that section of the course. If the quality of the gravel is variable, you’ll still face the same problem of not knowing Crr for the full course.

I think Jonathan encountered a similar limitation with BBS when he tried to use it to predict Leadville times.


Yup, non-pavement rolling resistance is the core issue. From memory, Jonathan took known power profiles and/or times and plugged them into BBS.

He back figured from the known equipment to come up with a modified CDA. The CDA was artificially high in order to account for the rolling resistance.

It’s a hole in the BBS model that should be filled. But I also see it will be difficult as the range of rolling resistance for non-paved surfaces will have a wide range. But the can likely start with a modifier that could be better than the CDA fugde.

1 Like

Best bike Split now has an option for gravel surfaces.



And for a course like Leadville, an average Crr will give very wrong segment times. The paved climb from turquoise lake to Carter summit is very different than the Jeep trail sugarloaf climb. So the predicted time for the overall course may be right, but segments could still be way off.

Over time, BBS may be able to do some specific Crr modeling for popular courses like Leadville and DK, but until they do, I’m not sure there’s a good way to apply BBS for variable surfaces.

1 Like

But the quality of gravel can vary a lot between courses, and within a course, so it will still be difficult to know what Crr to put in the model.

1 Like

Just pointing out that they are trying to ‘fill the hole’.

At the moment non of the modelling takes into account variability of road surface but I do understand that the variability (and the effect of that) may be quite a bit more significant on rougher surfaces.



I used Best Bike split for a recent century. The prediction gave me a finish time of 6:30 for a NP of 130. I did the century with a riding time of 7:48 with a NP of 129. I think they have a ways to go in the modelling algorithm before anything useful can be produced.


Good to know. I don’t use it, so I haven’t kept up with it other than what I see from others.

Thanks for the update. :smiley:

Well, many other people have had very close and successful results. One example does not a pattern make.

It’s a math model and subject to the data it is given. Garbage in, garbage out.

Thanks for all of the information. Sounds like it is a rough guess when it comes to gravel. I will take it into consideration and use this as an excuse as to why I failed my goal time…

If you want to go by anecdote, BBS predicted 27:50 at 300 NP for me in a recent time trial and I ended up with 27:45 at 300 NP, so…

Sure. This and the fact that the results are apparently good enough that pro teams and athletes use BBS for many of their race plans is sufficient validation, IMHO.

Nothing is perfect and I expect that there will be outlier results. But as I said above, it’s using input data in the form of GPS data, rider info, and equipment info. Get any one of those entered with inaccurate data, and your results will be negatively impacted.

First off, you attached the weather graph, which is probably not very useful. The power plan graph might be more useful.

The CdA looks reasonable. CdA isn’t going to change between road and gravel, really, unless you ride much more upright or wear baggier clothing. If you ride gravel relatively upright, that CdA could be low.

The Crr is what is going to really make gravel slow. That’s going to be hard for any model to do well because it’s so dependent on the surface, which they know nothing about. That Crr is much higher than a road Crr, though.

The other big difference is pacing. If you haven’t followed a BBS-like pacing strategy before – it’s like the difference between riding on the road and riding on the trainer. They model that you will be putting down decent power all the time. In my experience, that means taking hills easier than you naturally would, but it’s pretty merciless on flats and mild downhills. Bear in mind that for gravel, this means that the model is also not capturing any slowdown (going slower, coasting, or braking) for safety’s sake.

That said, for a tame gravel surface, 18 mph at 150 W doesn’t sound too unreasonable to me. On the other hand, 12 hours with the aero position and unrelenting pace that BBS models and I’d need a wheelchair.


Thank you. This is really helpful.

Is the course local? Can you ride part of it, or do you know how the gravel is across the whole of the course?
If you can ride some of the route, or on a comparable surface you could then try the BBS Race Analysis. That may give you more accurate figures as it uses weather forecast data and your actual ride data to suggest figures for your ride profile.

I’ve used it several times to get ballpark estimates of times for gravel events. It does take some effort to set things up so the results are realistic. Basically, I use a past event, load in the route to BBS, restrict the optimization so the minimum IF is the same as the IF from the actual ride, and then adjust the bike profile’s rolling resistance and mechanical loss until the predicted time is roughly consistent with the actual time. You’ll know you’re close to finding the right “recipe” for you, your bike, and your course conditions when the overall time, and times on various segments (e.g., long climbs, flats), are roughly right. It’s not a “plug and play” type of deal.

For my purposes, I have ended up using the following:

  • Bike profile that includes a bike weight of 22 lbs, Rolling Resistance of 0.0115 and Mechanical Loss of 0.08
  • Race profile specifying “poor” pavement and “heavy forest” terrain.

I have not tried the newer selection options for “off-road” road conditions or “mountainous” terrain.

1 Like

I have using BBS for a few years and had several email discussions with their support staff early on. The good news is that they are extremely responsive, particularly Cody Stephenson. However, I found their analysis of hilly road events/races (100 ft/mi) is “a bit aggressive” so I am not surprised by @T_Field experience.

What I learned from Cody is that their real expertise, and development of their algorithm, came from long, flat TTs. Many of the world’s best and elite Ironman competitors use BBS with very accurate results. However, their algorithm needed (and still seems to need) some improvement for hilly events. As such, despite now having a gravel option, I wouldn’t be at all surprised if it has a long way to go before it is reliable for that use.

FWIW: I still use it for hilly events, but apply my own downward scale factor on its predictions. I also use it relatively effectively for hill TTs.

1 Like

In order to get myself close for my last gravel race, I had to increase CRR and drag to max and I also went into the advanced settings and changed my max downhill speeds (to 25 mph for gravel) and my climbing speed (as I understand it, that wants you to consider when you will come out of aero position, which I set fairly high for climbing (I set to 12 mph)

Then for power, I set it to the power (not normalized) I thought I could hold for 4-5 hours (mid 180s for me) and it got me within a few minutes of my actual finish.

Of course this all comes with a huge YMMV… :slight_smile:

This is a great idea. I just adjusted some settings to get it close to a past event. Thanks for the information.

1 Like