Analytics & research
These pages help you research and trust the model before trading: replay the forecaster against reality, see where the realized edge is, and browse everything the bot scanned.
Hindcast /backtest
A leakage-free trust check: roll the clock back to an anchor, forecast from there, then overlay the bars that actually followed. Pick a ticker and an anchor (most recent window / ~2 days / ~1 week / ~3 weeks back) and Run.
How to read it: a DIRECTION HIT/MISS tag and a within / outside p10–p90 pill tell you whether the forecast was right and whether reality stayed inside the predicted cone. The chart shows grey history up to the anchor, orange for the bars that followed, a blue dashed predicted median, and the shaded p10–p90 cone.
Analytics /analytics
"Where the edge is" across completed trades (rolls/bookkeeping and never-filled excluded). Eight overview cards: Completed trades, Win rate, Total P&L, Expectancy / trade, Avg win, Avg loss, Profit factor, Avg hold. A best/worst extremes panel. A Performance vs S&P 500 chart (cumulative % return of the bot's realized P&L vs SPY buy-and-hold, both indexed to 0). And breakdown tables — win-rate/P&L by strategy type, by exit reason, by confidence/POP bucket, by Roundtable consensus, by days-to-expiry at entry, and by hold duration.
This is the page to check after paper-trading a strategy: positive expectancy and profit factor > 1 are what you're looking for.
Possible Positions /candidates
Everything the AI scanned, ranked by confidence (paginated). Columns: Symbol, Direction (LONG/SHORT), Confidence, Exp. move, Note, Made (EST), and an action (day-trade → Execute; spread → Execute; otherwise "analyze →" opens the Prediction Engine). Scan now forces a fresh scan. Click any row for the candidate detail, which shows the full Roundtable verdict and each judge's reasoning.