ROADMAP
Verdict's standard is simple: every structural claim on a verified card should be reproducible math, and every miss should become a permanent test. The engine gets harder to break with each chart it meets — that grind is the product.
One thing you will never find on this page: a promised win rate. Verdict grades structure for study — it is not a signal service, and calibration claims will only ever come from published, measured outcomes.
Pixel extraction — candles measured from any screenshot, sub-tenth-of-a-percent price accuracy, verified against real TradingView exports.
Verified facts — fair value gaps, equal highs/lows, sweeps, and BOS/CHoCH computed deterministically from extracted data. In verified mode the AI cannot claim a structure the math didn't find.
Honest fallback — screenshots that can't be extracted with confidence are labeled as vision estimates, never passed off as measured.
Regression suite — every chart style that ever failed is a permanent test case. The engine can't silently break on a chart it has already learned.
Re-verdict continuity — a follow-up analysis may only flip bias by naming the structural event that invalidated the prior read.
Watch-for triggers — every verdict ends with objectively checkable price conditions, not vague advice.
Chart-style coverage — 1M through daily, light and dark themes, drawings, phone captures. Each refused upload becomes the next detector patch and test case.
Evaluation lab — hundreds of historical verdicts simulated against candles the model never saw, with hard promotion gates. A prompt change ships only if it beats the incumbent on held-out data.
Deployment — verdictcharts.com, with the extraction service running alongside the app.
Accounts and auto-journal — every verdict saved with one-tap outcome logging.
Published calibration — do A-grades actually outperform C-grades? Does confidence 80 beat confidence 55? Measured from real outcomes and shown, not promised.
Pro tier — unlimited verdicts, journal analytics, priority processing.
Multi-timeframe context — a higher-timeframe read informing the entry-timeframe verdict.
Session analytics — killzone and session-level performance breakdowns from your journal.
Prop-account awareness — drawdown-aware grading for evaluation and funded accounts.
API access for communities and educators.