Surprise Me: how the app picks a ride when you can't

There is a particular Sunday morning where you are kitted up, the bike is warm, and you genuinely cannot think of anywhere to go. Not because there is nowhere — because deciding is work, and you wanted to ride, not plan.

That is the entire problem Surprise Me exists for. Here is what it is doing when you press it.

The naive version, and why it fails

The obvious implementation is: pick a random point 40 kilometres away, route there, route back.

Try it and you immediately see the problems. The return leg uses the same road as the outbound. The random point lands in the middle of an industrial estate, or a national park with no through road, or the far side of a river with no bridge. And “random” means it is just as likely to point you at the motorway as at the mountains.

A ride is a loop through several places, and it needs to be constructed like one.

How it actually works

1. Convert time into distance, correctly. You ask for two hours. Two hours of what? On curvy back roads, the achievable distance is far lower than the same time on highway — a road with 300 degrees of turning per kilometre is ridden a lot slower than the speed limit suggests. So the target distance is derived from the kind of riding you asked for, not from a flat average. This is the single most common reason naive route generators produce rides that take twice as long as promised.

2. Choose anchors, not points. Rather than one destination, the generator picks several anchor waypoints spread around a rough loop — different bearings from the start, at distances that add up to the target. Spreading by bearing is what produces a genuine circuit instead of an out-and-back.

3. Bias the anchors toward good ground. Anchors are nudged toward areas with denser, curvier road networks and away from urban grids and dead ends. This is why the loop tends to head for the hills rather than the suburbs.

4. Bias against what you have ridden. Anchors and the routing between them both lean away from roads on your ridden-roads map. For a rider with a season of recorded rides, this pushes the whole loop outward into ground they have not covered — which is usually where the good, unknown roads are.

5. Route between them with curvy weighting. The actual roads come from the same curviness model the manual planner uses.

6. Sanity-check the result. Total distance within tolerance of the target, no absurd doubling back, no unsealed sections unless your vehicle profile says that is fine.

Steering it without taking over

The first version of any feature like this gives you one button: re-roll. That is frustrating, because most generated routes are 80% good and 20% wrong — a leg through a town you dislike, or a fifty-kilometre highway drag to reach a good road.

So the editor lets you work in place:

• Lock an anchor you want to keep. Re-rolls will build around it.
• Drop an anchor that is causing the bad leg.
• Swap or reorder anchors to change the shape of the loop.
• Add your own — a lookout, a bakery, a road you have been meaning to try.
• Re-roll the rest, leaving the locked parts alone.

The mental model is a ride you are negotiating with rather than a slot machine.

Getting good results

Give it time, not distance. “Three hours” produces a better ride than “150 km”, because the generator can then trade distance for road quality.

Include your fuel stop. Add a fuel anchor if the loop is near your range. Curvy riding eats more fuel per kilometre than the same distance of highway, and a loop that assumes otherwise gets tense at the end. Our fuel planning guide goes into the maths.

Set your vehicle first. A big adventure bike, a sports tourer and a scooter should not be given the same roads. Vehicle profile drives surface tolerance, so a road bike is not sent down 12 kilometres of gravel.

Re-roll twice, then commit. The temptation is to keep re-rolling for the perfect route. The whole point was to stop deciding.

When not to use it

Surprise Me is not the right tool when the destination matters — meeting people, catching a ferry, or riding to a specific lookout at a specific time. Then place your waypoints deliberately and let the router handle the roads between them; the manual planning guide covers it.

It is the right tool when the answer to “where are we going?” is genuinely “somewhere good, back by three”. Which, if you are honest, is most Sundays.