Claude grades itself

Every Top 15 run is snapshotted with its entry prices — nothing edited, nothing cherry-picked. This page re-prices those calls against the market and publishes the result: win rates, average returns, performance vs the S&P 500, and whether Claude's confidence scores actually mean anything. It recomputes automatically about once an hour.

How the grades are computed

• Every Top 15 analysis is stored as an immutable snapshot with each pick's signal, confidence, and the market price at that moment — that price is the graded entry.
• A call is counted once: repeating the same signal on the same ticker across consecutive runs doesn't re-count. A new grade opens only when the signal changes.
• Picks are graded from entry price to the current (or last closing) price, once they are at least one day old. Open market positions, no exits — this measures call quality, not a trading strategy.
• “vs S&P 500” compares each pick to SPY over the same window (its entry date to now), then averages — no cherry-picked baselines.
• Headline stats cover BUY and STRONG BUY calls, the picks Claude actually recommends. HOLD / SELL calls appear in the by-signal table.

The Scoreboard measures hypothetical, frictionless returns on published AI signals — no real money, no fees, slippage, or taxes. Short histories are noisy and win rates on small samples can be misleading. Nothing here is financial advice; past performance does not guarantee future results. Read how the AI works.

Audit the record

Don't trust us — download every pick we've ever made

The full raw history is public: every Top 15 pick ever snapshotted (date, signal, confidence, entry price, target, return-to-date) and every Claude's Million trade with its timestamp and the model's written reasoning. No login, no cherry-picking — run the numbers yourself.