⌂ Research Index · tradebotsecrets research record · evidence-first, negatives included
Paper Trading Live
evolive testing ground · 4h deep cohort · forward from 2026-08-21 08:00 UTC

Paper Trading Live

All 8 champions of the 4h deep cohort trading forward on real Apex Omni candles, refreshed every 4 hours by replay-from-anchor. A note on the word prediction: this system does not predict price paths — that is a validated null here. Its prediction each candle is its position; what this page tracks is whether that standing call captured the next move, and what it cost or earned when it didn't.

The candles & the cohort's standing call

Muted candles are pre-anchor history (context only). Colored candles are the paper period. Background tint is the cohort majority call — green long, violet short, none flat. The amber line marks the anchor; the countdown chip above tracks the pre-registered bottom call (2026-09-26).

longshortflat / abstain— one row per seed champion (11…88), one column per paper candle

Call accuracy & running PnL forward in time

Left axis of the first pane: cumulative hit rate of the cohort's majority call against the next candle's direction (flat = abstain, unscored). Second pane: mean wealth across the 8 paper accounts with the min–max envelope, buy-and-hold dashed, banked ledger as the amber floor. The honest chance bar for direction is whatever the period's up/down split is, not 50%.

Trade history

Every fill and sweep across all 8 paper accounts, newest first. Sweeps move realized gains to the banked (funding) ledger — the ratchet that survives anything.

Live-trading readiness. The execution path is wired (live/live_trading_runner_v1.py → evolive2's SDK executor, read-only reuse) and held shut by five gates, all defaulting safe: LIVE_TRADING_ENABLED, evolive2's APEX_OMNI_EXECUTION_ENABLED kill switch, testnet-first APEX_ENV, the mainnet double-confirmation, and credential presence. The go-live checklist lives in .env.example — to be walked only after this page shows a profitable paper record.