Design principles

Part of the «Waze для рыбалки» wiki · Status: v1 — principles agreed, details open
Related: Synthesis · Open decisions

P1. Rival goods stay gated (the founding principle)

Public layer = data that costs nothing to share (ice, conditions, rules, aggregate activity). Spot-precision data = gated: waterbody/sector by default, delayed publication, private circles (viewer/contributor/admin roles — HuntStand model), give-to-get access (Glassdoor model). Public exact pins: never. Origin: secret-spots research; Fishbrain's "spot burner app" brand never washed off.

P2. Verify exact, publish coarse (geolocation — Piotr, 2026-07-02)

• Verify privately, on-device: precise GPS, in-app camera capture (not gallery), EXIF time/geo, mock-location flag check, impossible-travel detection.
• Reward with weight and reputation, not disclosure: "verified on-site" badge = the app captured the evidence; report gets higher map weight; author's reliability score grows.
• Display always coarse: waterbody/sector only. Precision buys trust, not spot-burning.
• Opt-in only: requiring GPS ⇒ refusal or fake coordinates from the «не палить точки» crowd.

P3. Corroboration for freshness, reputation for honesty

• Confirmations ("still biting"/"nothing here") extend a report's life; absence lets it decay. Confirmation ≠ truth (Waze study: reliability score uncorrelated with accuracy) — surface age + count, let users calibrate.
• Separate the fun leaderboard (volume/diversity/streaks — cheap, low-stakes) from the trusted reliability score that governs map weight (eBird decoupling).
• Thank-you economy as primary reputation input: "this tip worked" is hard to fake and aligns status with honesty.
• Private logbook first: people log obsessively for themselves; publishing is a separate incentivized step.

P4. Anti-abuse ships at launch, silently

gdebenz was gamed within days. Defenses: per-region plausibility filters (eBird model — marlin in Ķīšezers auto-flags), confirmation-ring detection, velocity/teleport anomalies, reverse-image search on photos, silent down-weighting instead of public bans (Stack Overflow model). Expect strategic fakes (decoy reports to divert crowds) — unique to fishing.

P5. Auto-suppression on sensitive waters

Auto-obscure/block precise reports on salmon/trout rivers Oct–Dec, spawning bans, protected waters (eBird sensitive-species model). Doubles as the reason regulators endorse us. VVD has prosecuted from social-media photos — never auto-attach exact geotags to catch photos; support pseudonyms and home-water privacy zones.

P6. Heatmaps need k-anonymity + delay

Minimum N reporters per waterbody/time-window before anything renders; publication delayed so the bite window expires before the crowd arrives. Naive aggregation deanonymizes on small lakes (Strava lesson).

P7. Zero friction to read, near-zero to write

Browser-first PWA, no registration to read, one-tap 4-state report (клюёт / вяло / глухо / людно) + species + method, ~15 seconds with cold hands. Accounts gate reputation features, not reporting. (gdebenz: every visitor is one tap from contributing.)

P8. Honest forecast

Conditions score, not bite promise. Probability bands, per-species, show the why. Solunar demoted and labeled; pressure as trend only. Validate publicly against user-reported catches. (forecasting)

P9. Bilingual LV+RU everywhere

Parallel communities that barely overlap; etiquette vocabulary and moderation nudges in both. Default-Latvian posture with full RU localization (political climate consideration).

P10. Community data free forever

The most-hated competitor move is paywalling community reports. Monetize premium layers only.