arcroad · Hyderabad · driven 2026-11-01

We drove the detours the map predicted.

On 2026-11-01 arcroad drove modelled places on Hyderabad's west corridor with a GPS recording. Of 5 direction(s) that could be compared with the model, 5 matched it within the tolerance fixed before the drive. Every pass is listed, including the ones that did not count.

3places now driven by arcroad
5/5directions matched the model
1passes rejected

Every pass

PlaceDirectionStraight acrossModelDrivenTimeVerdictEvidence levelEvidence kinds
T2A → B——3.08 km10.6 minNo model route to compareModelledS · candidate
T2B → A——3.89 km13.5 minNo model route to compareModelledS · candidate
S16B → A19 m2.38 km2.30 km9.0 minMatched the modelDriven by arcroadS · candidate
S18A → B13 m1.57 km1.58 km3.2 minMatched the modelDriven by arcroadS · candidate
S18B → A13 m1.38 km1.55 km11.5 minMatched the modelDriven by arcroadS+P · corroborated
S23B → A20 m—11.34 km13.1 minNo model route to compareModelledS · candidate
S24A → B20 m1.55 km1.77 km9.4 minMatched the modelDriven by arcroadS · candidate
S24B → A20 m10.49 km11.60 km25.2 minMatched the modelDriven by arcroadS · candidate

Evidence kinds: S a timed drive, P a usable photo at the place during the pass, R a repeat pass. One kind is a candidate, two corroborated, three with a photo or an official record triangulated. Times include waits at signals and U-turn queues (Sunday afternoon and evening).

Rejected passes

Visited but not scored

Limits

How this was scored

Thresholds were written before the drive and are not changed after seeing results. A drive matches when it is within the larger of 150 m or 25% of the model's distance. The code is public: pipeline/fieldscore.py, pipeline/groundtruth.py, pipeline/evidence.py; docs/ENGINE.md triangulation rule. The raw recordings stay private; only these per-pass numbers are published. See the places on the 3D map · arcroad home