Personal research project — paper-testing & tuning, not raising capital.
Paper research · advisory · not a yield promise

Strategy Tournament

Strategies compete backtest → paper → live, ranked by net return and Sharpe. Only credible-data winners promote. This is how the desk VETS candidates — not a yield promise. The desk itself holds a steady RWA-anchored conservative book — ~3.3% realized track-to-date.

source: static / offline ·

The honest framing — read this first

The promotion gate is fail-closed. It refuses to promote a strategy to live when the data is degenerate (locked-vol / near-constant → Sharpe explodes) or flagged untrustworthy. So "0 promotions" is the CORRECT honest outcome — not a failure, not an empty page.

We show the trustworthiness flags openly. If the dataset is flagged untrustworthy, every strategy in the tournament is shown flagged — we never render a mock ranking as a live winner.

assessing data trust…

Reading the trustworthiness verdict from the tournament dataset…

The full tournament cockpit · /api/tournament

live · fail-closed offline

The promotion ladder — where each strategy sits

A strategy climbs backtest → paper → live only by earning it. The live counter under each rung is how many strategies sit there right now. The final rung is guarded by the fail-closed gate.

1 Backtest

Historical simulation over the deterministic price/APY series. Ranked by net return + Sharpe.

strategies here
2 Paper (Nd)

The top strategies shadow-trade live paper daily — no capital. Accruing a real forward track.

strategies here
3 Live (advisory)

A candidate PROMOTES only if it clears every criterion ON CREDIBLE DATA. Advisory — never auto-executes.

strategies here

All four criteria must clear — ON CREDIBLE DATA

Sharpe ≥ 1.5

risk-adjusted return

≥ 7 days paper

enough live samples

APY ≥ 3%

clears a floor

Drawdown ≥ −15%

survivable loss

Why most DON'T promote

Even if a strategy mathematically meets all four criteria, the gate refuses when the dataset itself is degenerate (locked-vol / near-constant → Sharpe explodes into the tens/hundreds) or flagged untrustworthy. A degenerate Sharpe can never reach live. The refusal is honesty, not a failure.

The promotion verdict — what the gate decided this run

static / offline

Loading the promotion verdict…

Personal research project in paper validation — not investment advice, not a regulated service, not raising capital; results are simulated. Full disclaimer & risk disclosure →