# Routing and Scoring Model

## Goal

Optimize for a loaded bikepacking trip, not a generic bike route.

## Route cost model

A candidate route is scored by segment and by whole-trip logistics.

```text
segment_cost =
  distance_cost
+ climb_cost
+ steepness_cost
+ poor_surface_cost
+ traffic_stress_cost
+ access_uncertainty_cost
+ hike_a_bike_risk_cost
+ remoteness_cost
+ weather_sensitivity_cost
- official_cycle_route_bonus
- confirmed_good_segment_bonus
- scenic_bonus
```

```text
stage_cost =
  ride_effort_cost
+ sleep_risk_cost
+ water_gap_cost
+ food_gap_cost
+ daylight_risk_cost
+ weather_risk_cost
+ bailout_gap_cost
+ repair_gap_cost
```

## Segment attributes

Each segment should carry:

- distance meters,
- ascent/descent,
- max grade,
- average grade,
- surface,
- smoothness,
- OSM highway class,
- bicycle/access tags,
- traffic stress proxy,
- cycle route membership,
- protected-area overlap,
- POI distance to next water/food/repair/sleep,
- community report score,
- confidence score.

## Bikepacking profiles

### Loaded gravel

- Prefer gravel/compact dirt and low-traffic roads.
- Penalize technical MTB trails.
- Penalize grades above 12% more heavily when loaded.
- Penalize unknown surfaces in remote areas.

### Hardtail bikepacking

- Allow more rough surfaces.
- Allow singletrack where bicycle access and difficulty are acceptable.
- Penalize sustained paved highways and high traffic.

### Road touring

- Prefer paved low-traffic roads and official cycle routes.
- Avoid rough surfaces, tracks, and paths with unknown rideability.

### E-bikepacking

- Add range and charging constraints.
- Penalize long remote stretches without accommodation/charging.
- Respect local e-bike access distinctions where available.

### Beginner-safe

- Strongly prefer known surfaces and legal certainty.
- Avoid high traffic, remote stretches, long dry gaps, and steep grades.

## Confidence model

The app should show both recommendation and confidence.

Example route summary:

```text
Distance: 382 km
Ascent: 5,420 m
Known surface: 88%
Unpaved: 42%
Longest water gap: 36 km
Longest food gap: 58 km
Access confidence: Medium-high
Sleep certainty: 4/5 nights confirmed options
Bailout gap: Max 71 km from train/ferry/bus
Warnings: 2 steep loaded-bike climbs, 1 uncertain access segment
```

## Hike-a-bike risk

Estimate using:

- steep grade,
- poor smoothness,
- `mtb:scale`,
- `sac_scale`,
- surface type,
- trail width if available,
- rider profile,
- actual ride speed reports,
- user reports.

A route can include hike-a-bike if the profile allows it, but it must be visible.

## Critical gap detection

Calculate gaps along the route:

- water gap,
- food gap,
- sleep gap,
- repair gap,
- bailout gap,
- charging gap for e-bike.

Warnings should be generated when a gap exceeds profile thresholds.

## Stage planning algorithm

1. Generate candidate endpoints every 5–15 km around the daily target window.
2. Score each endpoint by sleep options, water/food, daylight fit, safety, and bailout.
3. Prefer endpoints with at least one high-confidence sleep option.
4. Add Plan B endpoints before and after Plan A.
5. Rebalance subsequent stages to avoid one impossible day.
6. Surface warnings for stages where no good endpoint exists.

## Example thresholds

| Profile | Target km/day | Max grade warning | Max water gap | Max food gap |
|---|---:|---:|---:|---:|
| Beginner-safe | 40–70 | 10% | 25 km | 40 km |
| Loaded gravel | 60–100 | 12% | 40 km | 70 km |
| Hardtail remote | 50–90 | 15% | 55 km | 90 km |
| Road touring | 70–130 | 10% | 40 km | 70 km |
| E-bikepacking | 50–100 | 12% | 40 km | 60 km + charging |

Thresholds should be configurable per user.