Yeah that seemed a bit odd to me too, but from a physics perspective it does make sense: https://github.com/abrensch/brouter/issues/309#issuecomment-855833481

It doesn’t make sense in practice, because one doesn’t ‘waste’ energy in a descent. One can ride it down with 0 Watt. Nobody ever avoided a route because there was a downhill on the way :-)

The problem is going up. And the steeper the uphill is, the worst the problem is.

In the example I had taken, if I take the 12% road, I need to take a rest on top of the section (in the best case, otherwise I may even need another break in the middle). And I am pretty dead after the common section, and I carry that in following sections again (recovery is not my strong point :-)). If I take the 8% road, I can ride it in one go, then I may or may not struggle in the following sections because I am a bit burnt. If there was an hypothetical twice-as-long road at 4%, I would start the following sections as fresh as a daisy.

Note that going up can be decorrelated from going down. I absolutely don’t care about which kind of gradient is exhibited by the descent that might come after.

In their logic, they also forget that even with a nice, flattish descent that they wouldn’t penalise, it is always both faster and easier on the legs to ride around the hill than climbing up and down through a pass, even though the detour is a bit longer. Because the climb is what kills it all, it rarely works as an investment as they call it; it rather hurts your legs and diminishes your abilities for the rest of the route.

Basically it aims for energy efficiency by avoiding excessive wind drag.

Yeah, and I don’t think that’s the right metrics in most cases :-)

Personally, if we set aside true winds, the only situation in which I sometimes care about wind drag is over long downhill false-flat sections (and only when I am in a shape, for that’s the only combination where the speed can have an impact on me). This situation is specifically the one case that they don’t penalise…

Anyway, as you said, that piece of software is highly configurable, so one can configure it as one likes when one disagrees with their logic.

Edit : would you happen to know which Brouter parameter should be set/tweaked to avoid/limit the occurrence of U-turns? By “U-turn”, I mean that when I set an intermediate waypoint, my intention is generally to go through it and then keep going on the same road or general direction, not turn back and come back the way I came.

See Example . You see how I engaged my intermediate waypoint on the D32 and D17, but the router goes to the waypoint, and then comes all the way backwards to take the D618.

Of course, in practice, I can move the waypoint or add new ones, that’s how I always did so far with other, non configurable, routers. It’s not a big deal.

Yes I think that mapping the inclines would be the way to go here. Typically maps have some elevation data from other sources but this should be more precise.

OK, I’ve added the “incline” tag on the small street (I had to split several paths to do so, because they had been drawn as parts of other roads/streets).

Brouter is also super flexible!

I see. That’s very interesting.

You can set the downhill and uphill costs in the profile tab. Increasing them to 90 made it prefer the bigger road in both directions in your example.

Wait, by default they have a penalty on downhill, but not on uphill??? What the… I see 60 for “downhillcost”, and 0 for “uphillcost” for most predefined bike profiles… Do you see the same?

While we’re at it, I have another case in the area that always annoys me: Example on brouter.

I guess the case is rather common: the router routes through smaller roads/street while the bigger road is more convenient. The smaller street is super steep (>12%), while the bigger road was built much later with a rather smooth 7-8% gradient. There is very little traffic on the bigger road. The smaller street is not only a problem when going up: when you ride downwards, the brakes suffer a lot, yet you are supposed to make a full stop at the bottom crossroad, an you really need to (as there is zero visibility there), unless you want to be run over by vehicles from the main road 4 feet farther…

I thought that the smaller road/street being partially marked as “highway=residential” would have helped picking the bigger road, but nope.

Do you think setting “incline” would help routers to pick the main road? Or any other tag?

There’s no one tag to do this but a combination of mapping the negative sides of the tunnel road and the positive sides of the bigger road should be enough.

That’s what I feared: the routers can use anything with any weight, and we cannot know about it. It could be a good idea to reach people doing the routers and tell them that stuff like unlit tunnels should be weighted negatively.

Some ideas:

cycleway=no for the tunnel road and perhaps something like cycleway=shoulder for the bigger road

I can add “cycleway=no” for the tunnel road (cycleway is not set there), but the bigger road has “cycleway:both=no” already set, and well, that’s the reality.

Maybe the bigger road has better asphalt? Tagging smoothness values of both can help.

That depends on the year :-) But something is constant: the potholes inside the tunnels; so yes, I can at least set “smoothness=bad” under the tunnels.

Are there official cycle routes going through? Many routers take these very seriously. Note that it may be that there’s an outdated route mapped through the tunnel road.

I cant’ see such route on RideWithGPS’ “Cycle OSM” view (I guess I would if there was one), and I don’t see any of the tags present on the page you linked in OpenStreetMap objects data either.

I’ll perform the two small modifications I picked from your list, and see after a few weeks or months if something seems to have changed, or not.

Thanks for the pieces of information you gave me.

And hey, since all of these tools rely on OpenStreetMap data,

By the way, I have in my area a road that is very dangerous for bicycles because it goes through several unlit tunnels. It’s a former single railway line turned into a one-lane, one-way road; most cars enter the tunnels way over the allowed max speed without seeing at all what is inside.

All routing tools based on OpenStreetMap insist on routing through this road, despite the fact that there is another (larger) road on the other bank of the river. They just go in parallel 25 yards away for each other.

A couple of years ago, I edited OpenStreetMap data to mark the tunnel with “lit=no”.

But that was in vain, routing tools don’t take this parameter into account and keep routing bicycles through the (smaller) dangerous road. They prefer to route through the smaller road, which is generally the right choice, but not at all in the present case.

Does someone know if/how I can mark this road as “non-suitable for bicycles although they are allowed there”?

Yeah.

Yesterday, on Paris-Tours, the case showed a few similarities but was a bit different, in that the favourites did not appear to have understood that it was a race for tough guys, which means said tough guys must at some point take the matter in their own hands (as in a spring Belgian race) and not wait for their drained and unique surviving domestique to do the job in their place.

This profile is from stage 7 of the Vuelta… 2015!

Today is flat, flat, flat…

Lenny Martinez get the Red Jersey. With a 2:30 lead over the big favourites, I could see him keeping it quite late in the race.

It is possible, but there are other contenders closer than the favourites, and they may try something earlier. However the closest 2 are super-domestiques and they may not be given the liberty from their team to try anything, and the 3^rd one is already pretty far.

It certainly sets him up really well for a good overall result.

It is possible too, but so far he has shown some growing difficulties after some point when efforts accumulate day after day.

Evenepoel held it together well after being dropped

It is often like this with him: it feels like he’s badly cracking, but after a while he stops the loss (strangely often around 30 seconds in my admittedly faulty memory) and sometimes comes back a little or completely.

A bit like Almeida, even though Almeida doesn’t give so much the feeling of cracking but more of letting go; and then finishes more frantically, catching up with the best.