F1805 · an Elite AI Empire product

"Great backtest."
Is it real, or overfit?

Empire Validator runs your strategy's results through a 26-method validation battery — deflated Sharpe, walk-forward, regime-split, look-ahead leakage, survivorship, multiple-testing correction — and returns one honest verdict: tradeable, or a story you told yourself.

See pricing Join the waitlist

# Bring your own returns. We don't need your alpha — just the trades. from validator import Validator v = Validator(trades=my_fills, prices=my_universe) report = v.run_all() # 26 checks, no config required DEFLATED SHARPE 1.81 → 0.42 after 240 trials ✗ likely overfit WALK-FORWARD IS 2.1 / OOS 0.3 ✗ degrades out-of-sample LOOK-AHEAD clean ✓ REGIME SPLIT profit concentrated in 1 regime ⚠ VERDICT: DO NOT DEPLOY — re-test with the 3 flagged fixes.

Deflate

The overfit detector

Most "amazing" backtests are the survivor of hundreds of silent trials. Validator computes the deflated Sharpe ratio — adjusting your headline number for how many strategies you tried — so you see the real edge, not the lucky one.

Forward

Walk-forward & out-of-sample

Rolls the train/test window forward across your history and reports in-sample vs out-of-sample decay. A strategy that only works on the data it was built on gets caught here.

Leak

Look-ahead & survivorship checks

Scans for the classic ways a backtest cheats: future data in features, delisted names dropped from the universe, point-in-time violations, and fill assumptions that the market would never have given you.

Stress

Regime split & multiple-testing

Splits performance by market regime so you see whether the edge is broad or a single lucky window, and applies multiple-testing correction so 200 backtests don't hand you a false positive by chance.

Built by a team that trades its own book. Empire Validator is the validation methodology stack we use to decide whether a candidate strategy is allowed near capital — packaged as a harness you point at your trades and your universe. It ships the test battery and the scoring math only. None of our strategies, signal definitions, thresholds, or data come with it — and yours never leave your environment.

Who it's for

Quant funds, prop desks, systematic-trading teams, and serious retail algo traders who need a rigorous, defensible "is this real?" gate before allocating — and a number they can show an investor or a risk committee.

Why not just VectorBT/backtrader?

Those generate backtests. Validator interrogates them. It's the layer that tells you whether the pretty equity curve your engine produced will survive contact with live markets. See the comparison →

The pitch in one line

"A backtest is a hypothesis. Validator is the peer review — before you bet the fund on it."

Early-access waitlist

Public launch 2026. Early-access = founding pricing locked (20% off forever) + a free validation of one strategy + priority onboarding.

"Great backtest."Is it real, or overfit?