A convergence score built from insider, congressional, 13F and on-chain whale signals is traded by two paper bots and has to pass its own kill-test before it can call itself proven.
26Sources
1/7/30 dOutcome checks
2Paper bots
16 daysBuild time
1Kill-test status: zero live samples yet, score marked unproven
2Cohort backtest table: each signal type against SPY and a control group
Reliability tab: kill-test status and the historical backtest table.
1Kill-test status: zero live samples yet, score marked unproven
2Cohort backtest table: each signal type against SPY and a control group
What runs automatically
01 / INGEST
Signal ingestion
26 sources, SEC insider filings, congressional trades, 13F fund portfolios, activist stakes and on-chain whale wallets, are pulled in and normalised into one feed per ticker or coin.
02 / SCORE
Convergence score
A single source never moves the score alone: it climbs only when independent signals agree. Underneath, a trainable model is re-weighted from the journal's own outcomes and evaluated walk-forward against a simple rule scorecard.
03 / DECIDE
Two paper bots, journaled
Two autonomous bots, one for stocks and one for crypto, journal every entry, exit and skip with a reason attached. A risk layer caps position size and exposure, with a kill-switch on realized and unrealized loss.
Built with Claude Code in 16 days
Before
Signal trackers showed a trade, never whether it was worth following.
After
Two paper bots trade one score, and its own kill-test still calls it unproven.
[→]/More screens
Stocks tab, minutes after start: the 14-day signals table is honestly empty, no ticker has cleared the threshold yet.
Crypto tab, same fresh instance: public market data from Binance and CoinGecko populates almost immediately.
Under the hood
01Context
Trackers show a trade, not whether it means anything
Tools that surface insider trades, congressional filings or 13F fund moves already exist, but most of them stop at the data: a trade appears, and it is left to the reader to decide whether it is worth following. This panel goes a step further: it combines several independent smart-money signals into one score, trades that score for real with paper money and measures, in public, whether the score actually beats a benchmark.
02What we built
One convergence score, two paper bots
A dashboard pulls SEC insider filings, congressional trades, 13F fund portfolios, activist stakes and on-chain whale wallets into one convergence score per ticker or coin: a single source never moves it alone, it climbs only when independent signals agree. Stocks and crypto sit in their own tabs with their own feed. Two autonomous paper bots trade that score, one for stocks and one for crypto, and every decision, entering, exiting or skipping a candidate, is journaled with a reason, exposed through a web panel and a Telegram bot.
03Constraints
Paper only, and an edge not yet proven
Every signal is journaled with price and benchmark included, then rechecked at 1, 7 and 30 days; weaker signals are logged too, as a control group. The Reliability tab runs a kill-test on that journal: a strong signal has to beat both the control group and the benchmark, net of costs, before the score is allowed to call itself proven. As of this write-up the kill-test reports zero completed samples in both cohorts, so by its own rule the score counts as unproven. Both bots trade paper only, and this is a research and monitoring panel, not investment advice.
ANM SOLUTIONS / CONTACT US
Want something like this?
First 30-minute call is free. We scope it, you decide.
