We grade ourselves in public. Quarter-by-quarter.

How we score a hit

Open methodology. The formula and the data are both public.

Catalyst calls — 50-row sample

Archive — not the current window. These 50 rows are frozen from 3 March – 2 April 2026 and are not regenerated; they show 45 hits and 5 misses, which is not representative of the live rate above. Treat this as a historical worked example of the grading format, not as evidence. The live, continuously regraded numbers are the ones in the stat row at the top of this page and in /data/hit_rate_audit.json.

DCF audit — 5 names, 6-month outcome

A-grade calls should outperform SPY by ≥ 5% over 6 months. F-grade should underperform by ≥ 5%. Each card shows the bet we made and the realized outcome.

What we haven't graded yet

Honest list. We don't claim a hit rate on anything below — sample size or data plumbing isn't there yet.

Coinbase live P/L (real money, public ledger)

The bot is trading real money on Coinbase Advanced. Every fill is shown below. Equity, realized + unrealized P/L, and fees paid are pulled directly from Coinbase's fills API — not the bot's self-report. If the number is negative, we're down. If positive, we're up. No spin.

Source: /data/coinbase_pnl.json. Updated every cron tick (~5 min). Includes Coinbase taker fees in the math. Strategy: 24/7 spot trader, 10-coin universe, signal stack = crypto-treasury equity flags + BTC ETF flow + DeFi calm + ETH gas relief + Coinbase premium + intraday momentum, +5% TP / -4% SL / 168h time-stop, conviction-weighted sizing.

Referral link — both you and we earn fee credits when you sign up and trade. Disclosures.

Agent paper-trading audit (live)

In addition to the 2026-Q1 SEC audit above, an autonomous agent now places paper trades against Alpaca off the same scanner signals. Live stats refresh from /data/agent_outcomes.json every 60 seconds. As fills accumulate, this section becomes the second independent audit of the same scanner — different timeframe, same methodology.

Source: /data/agent_outcomes.json · live from Alpaca paper account · also feeds tune_scoring_config.py at a 30% blend weight (SEC × 0.7 + agent × 0.3) once any signal bucket crosses 10 trades.

Crypto-treasury equity bot (live ledger)

A separate bot trades the crypto-treasury equity basket — MSTR, MARA, RIOT, COIN, CIFR, CLSK, IREN, BITF, HUT, HIVE, TSLA, SQ, HOOD, BITO — whenever the SEC catalyst scanner flags one of those names. Same signal, two ledgers in parallel: Alpaca paper (always-on proof) and Tradier live (gated on funding). This is "trade crypto" using our actual edge, not pretending the SEC scanner predicts spot BTC.

Source: /data/crypto_treasury_decisions.json (today's decisions) + /data/crypto_treasury_ledger.json (audit trail). To enable real Tradier orders, set TRADIER_LIVE_ORDERS=1 in the env file. Per-position cap: $25.

Coinbase Advanced 24/7 spot ledger

The 24/7 leg of the trading harness. Equity bots only fire 9:30-4 ET — this runs every cron tick, always. Two stacked signals: (a) crypto-treasury equity flags from our SEC scanner score ×3 weight, (b) BTC ETF heavy-flow + DeFi calm + ETH gas relief score ×1 each. Threshold ≥ 2 to enter. Dry-run by default until COINBASE_LIVE_ORDERS=1 + COINBASE_API_KEY + COINBASE_API_SECRET are set in env. (Hyperliquid was tried but blocks US jurisdictions; Coinbase Advanced is US-legal and 24/7.)

Source: /data/coinbase_decisions.json + /data/coinbase_ledger.json. Setup: Coinbase account → Settings → API → New API key (default permissions: trade) → paste key + secret into .sec_email_env. ~10 min if account exists.