Predict & Monitor
Real-time distance-to-trigger for every live rule, plus a real backtest of which stop-loss/take-profit combination would have compounded fastest on actual historical prices.
Queued trades
Live rule monitor
COG's learned strategies
Runs automatically twice a day (plus every manual "Run predictions" below) and tracks an exponential moving average of each strategy's backtested performance across every evaluation — a branch that keeps validating well over many separate runs and real elapsed time is a stronger signal than one good backtest. This is the "constantly backtest and learn" loop; the one-off table further down is a single snapshot.
Not enough repeated evaluations yet (needs 2+ separate runs of the same branch, still positive and a "candidate" verdict). Check back after the next scheduled run or click "Run predictions" below.
Strategy research lab
Runs the full strategy catalogue (buy & hold, DCA, momentum, trend-following, mean-reversion, moving-average crossover, RSI, MACD, volatility breakout) against real historical prices for every held/targeted/COG-flagged ticker, computes Sharpe, Sortino, CAGR and max drawdown from the actual equity curve, and accumulates an EMA + confidence score across every run — same "constantly backtest and learn" loop as the branches above, generalized beyond stop-loss/take-profit. Ranked by risk-adjusted return (Sharpe), not raw return. Also runs automatically twice a day. Research only — never applied to a live rule automatically.
Not enough repeated evaluations yet (needs 2+ separate runs of the same strategy, still positive EMA return and Sharpe). Click "Run strategy lab" above, or wait for the next scheduled run.
COG-aligned live rule
Scoped to DCA specifically: it's the one strategy type that's both backtested above AND actually executable by the trading loop (signal-based strategies like RSI/MACD have no live evaluator yet). Creating a rule here only ever makes a dry-run draft on the Automate page — going live is still a separate, explicit action.
No COG-aligned DCA recommendation yet — needs at least 2 evaluations of a COG-flagged ticker with positive EMA Sharpe and return. Run the strategy lab above (COG signals are included automatically) to contribute more evidence.
COG prediction ledger
Every distinct belief COG's world-model expresses gets logged here with its price at the time. After a 180-day horizon, it's scored against the real price — bullish is "correct" if the price rose, bearish if it fell (moves under 2% count as "flat", neither). This is purely observational — it never creates, blocks, or influences a trade. It exists to answer one question with real numbers instead of a guess: when COG says something, is it actually right?
No predictions logged yet — the scheduler records one automatically whenever COG expresses a new or changed belief.
Portfolio performance
COG shadow portfolio
A hypothetical $1000bet on every bullish COG call, bought at the recording price and marked to market today (or at resolution). No real money moves. Bearish calls are excluded — nothing in this system's execution layer can short, so simulating one would model a capability that doesn't exist. The real-portfolio figure is context only, not a controlled comparison — different capital, timing, and assets.
No bullish COG calls logged yet — the shadow portfolio populates automatically as the ledger accumulates them.
Crypto shadow portfolio
A $1000 SIMULATED position per real, evidenced crypto/dYdX strategy (2+ real backtest evaluations, positive EMA Sharpe and return), entered at the real price when evidenced, resolved 30 days later against the real price then. No real capital moves — same research→backtest→shadow staging as everything else here, applied to dYdX and KuCoin holdings.
No evidenced crypto strategy yet (needs 2+ real backtest evaluations with positive Sharpe and return) — populates automatically as evidence accumulates.
Run a fresh backtest
Backtests real historical daily closes (Yahoo Finance) for every currently-held and targeted ETF/stock, across a grid of stop-loss / take-profit combinations, and ranks them by compounded return. Paper/backtest only — nothing here places a trade; applying a recommendation means adding the matching rule yourself on the Automate page.
Click "Run predictions" to backtest against real prices.