NuroPicks Roadmap
Last updated: 2026-04-19. Status: living doc. Source: internal 2026-04-16 feature-priority research.
Status legend:
- shipped - in production on
main, usable today. - partial - a working version exists but does not yet match the target bar.
- planned - not built yet, approach decided.
- speculative - on the list, but we may drop it.
Tier 1 - MVP (must ship at launch)
| # | Tool | Status | Notes |
|---|---|---|---|
| 1 | AI match predictions (XGBoost ensemble) | partial | /predict returns model output today; full XGBoost-ensemble train on 5-10 seasons still open. Target 53-58% accuracy. |
| 2 | AI chat with "WHY" explanations (SHAP) | shipped | Generalized 2026-04-22. Canonical picks.shap_top schema ({ base_prob, model_prob, features[] }) defined once and consumed by every surface: src/bot/services/pick-shap-explainer.js (Discord embeds via picks-publisher.buildEmbed) and web/src/lib/pick-shap-explainer.ts (web /record Recent Edges panel via explainPickStructured). Graceful fallback to legacy why_edge strings when shap_top is null. Pipeline stub emits the canonical shape. 26 unit tests across both ports. Next-gen work: train the xgb scorer that produces real drivers (Tier 1 item 1). |
| 3 | Multi-book odds comparison | shipped | /odds via The Odds API covers 40+ books. See docs/DATA_STACK.md. |
| 4 | Bet tracking + auto-settlement | shipped | /bet, /mybets, /settle live. Settlement currently layer 1 (The Odds API /scores). Layer 3 SportsDataIO graduation planned for real-money-tracked parlays. |
| 5 | CLV personal tracker | partial | Picks-side capture shipped 2026-04-22: migration 0020-closing-line-capture.sql adds event_id/commence_time/sport_key to picks; src/bot/services/closing-line-capture.js fills closing_odds from the odds-feed cache after kickoff. /record and web/src/lib/record.ts already consume it. User-side CLV (bets.clv_delta) still pending — requires /bet to persist to Postgres first. |
| 6 | Daily streak mode | shipped | /daily live with streak multiplier + 100 XP streak-bonus cap (exploit patched 2026-04-19). |
| 7 | Multi-metric leaderboard | shipped | /leaderboard live. Min-picks-to-qualify + ROI x Volume x CLV scoring still partial; current implementation favors volume. |
Tier 2 - Phase 2 (next 8 to 16 weeks)
| # | Tool | Status | Notes |
|---|---|---|---|
| 8 | Copy-betting | planned | Dabble shows 50% adoption, eToro proved the model. Target 3% rev on volume. |
| 9 | Account health monitor | planned | First-mover. Nobody tracks sportsbook limiting patterns. Feed layer 3 required. |
| 10 | Automated Kelly bankroll management | partial | /bankroll surfaces basic unit sizing and quarter-Kelly. End-to-end automation across linked books still open. |
| 11 | Line movement + AI context | planned | "Line dropped 2.5 pts: QB injury reported 10 min ago" style annotations. |
| 12 | OCR betslip scanner | planned | Google Vision 98% accuracy. Import existing bets from screenshots. |
| 13 | Reputation tier system | partial | Level-based XP progression live with LEVEL_BADGES (Rookie Sharp 5, Rising Sharp 10, Veteran 25, Elite Sharp 50, Legend 100). Channel/feature gating by tier still open. |
Tier 3 - Phase 3 (post-launch iteration)
| # | Tool | Status | Notes |
|---|---|---|---|
| 14 | Capper marketplace | planned | Research recommended 85/15 split; we are going 50/50 (see blog/2026-why-we-split-affiliate-revenue-with-cappers.md). |
| 15 | Parlay correlation analysis | planned | Which legs actually correlate vs independent. Basic version in /parlay today; full correlation model planned. |
| 16 | Referee/umpire bias analytics | speculative | Unique data angle, but build cost is high and the addressable audience small. Hold. |
| 17 | Weather + travel fatigue impact | planned | Auto-flag rest advantage. Low build cost, integrates cleanly. |
| 18 | Cross-exchange arbitrage scanner | speculative | Novig/ProphetX exchange growth-dependent. Revisit in 2 quarters. |
Table stakes (everyone has, we must match)
- Multi-book odds comparison - shipped
- AI betslip generation (Gambly/Unabated commoditized since their merger) - partial
- Basic bet tracking - shipped
- Push notifications - partial (Discord notifications yes; mobile push planned)
- Dark mode UI - shipped (Neural Dark by default; no light mode)
Differentiators (nobody has, we should own)
- AI that explains WHY with SHAP in plain English - partial
- Account health monitoring across sportsbooks - planned
- Automated Kelly-criterion bankroll - partial
- Tamper-proof verified capper records at scale - planned (oracle layer flagged in
docs/DATA_STACK.md) - US-market copy-betting (Dabble is UK/AU only) - planned
- Public CLV performance (transparency moat) - partial (track record exists, public dashboard open)
- Esports AI picks with cross-book arbitrage - speculative
- Daily streak gamification inside Discord - shipped
What we deliberately do NOT build
- Direct sportsbook scraping (DK, FanDuel, etc.) - ToS violation. We use The Odds API.
- ESPN unofficial API dependencies. No SLA, no stability guarantee. Blocked in
docs/DATA_STACK.md. - Real-money wagering on-platform. We are an analytics and capper platform. Wagering stays at licensed sportsbooks.
- Anonymous user profiles with no RG posture. Every user surfaces
/self-exclude+/report. Gated tiers require age + state verification. - Paid flat-fee capper tiers. We pay cappers from the 50/50 rev-share, not from a flat retainer.
Minimum viable launch set
7 Tier 1 tools + Discord bot + web dashboard + newsletter integration = launch-ready. 4 of 7 Tier 1 items are shipped, 3 are partial. Tier 1 gaps are the primary /loop focus.
Cadence
Roadmap reviewed on each full-ship cycle. Items move from planned to partial to shipped as /loop cycles land. Status changes get logged in IMPROVEMENT_LOG.md under the cycle that moved them.