Edgeproof Request an audit

Independent Validation · Algorithmic Trading

We try to kill your strategy before the market does.

An independent testing house for trading algorithms. We run your strategy through the battery most backtests never survive — and give you a straight verdict before it touches real capital.

Code reviewed under NDA · No signals · No hype

Certificate of Validation No. EP-26-0047 · Edgeproof Protocol v1.0 · Illustrative Sample
STRATEGY: Momentum breakout · 15m index futures
SUBMITTED: Python · 4,100 lines · 6yr backtest
CLAIMED EDGE: In-sample Sharpe 2.4
31/100
Edgeproof Score · Reject band (below 40)
Bottom quartile · Edgeproof reference set
Data integrityno look-ahead foundPASS
Out-of-sample holdoutSharpe 0.31FAIL
Walk-forward ×123 of 12 windows profitableFAIL
Cost + slippage stressedge −38% at 1 tickFAIL
Monte Carlo ×5,000p = 0.21 vs. random entryFAIL
No. EP-26-0047 · Verify at edgeproof.example/verify

The Problem

Why backtests lie

An equity curve going up and to the right feels like proof. Usually it's one of these three things instead.

Overfitting

Test enough variants and one will look brilliant by pure luck. AI makes it worse — a thousand strategies a day, each fit to the past, blind to the future.

Frictionless fills

Backtests fill instantly at the price you asked for. Live orders don't. Fees, slippage and latency quietly eat edges that looked enormous on paper.

Regime luck

A strategy tuned on one regime often dies in the next. The market that made your curve is not the market you'll trade in.

None of this means your strategy is bad. It means an equity curve is a claim — and claims get tested.

The Audit

The test battery

Every strategy runs the full battery — no skipped tests, no cherry-picked windows. Either way, you'll know exactly why.

Standard Battery · Six ExaminationsIncluded in every audit
IData integrityLook-ahead bias, survivorship, repainting, bad ticks. Most fatal flaws live here.
IIOut-of-sample holdoutUntouched data your optimizer never saw. The first honest number.
IIIWalk-forward analysisRolling re-optimization across regimes. Does the edge persist, or did it happen once?
IVMonte Carlo resamplingThousands of resampled paths, one question: is this distinguishable from luck?
VCost & slippage stressRealistic fills, fees and latency. We find the friction level where your edge disappears.
VIParameter sensitivitySmall input changes shouldn't collapse the system. Fragile parameters are overfitting in disguise.

The Deliverable

One of three verdicts

You get a full written report and a 45-minute walkthrough call either way. The verdict is just the headline.

Validated

The edge survived the battery. You get the evidence, its limits, and the conditions under which it holds.

Marginal

Something real is there, but it's fragile. You get exactly what breaks it — and what to fix before it ever sees money.

Rejected

The edge doesn't survive contact with reality. It stings for a day, and it's still the cheapest loss you'll ever take.

We have no incentive to flatter your strategy. The audit is a flat fee — we get paid for the truth, not the outcome.

Services

Two ways to work together

For strategies that exist

Validation Audit

$2,000$750 founding rate · 10 business days

First 5 audits, in exchange for an honest, anonymized testimonial — 5 of 5 remaining.

  • One strategy through the full six-test battery
  • Institutional-grade data, realistic execution modeling
  • Written verdict report you can act on — or show a partner
  • 45-minute walkthrough call, straight answers
Request an audit

For prop traders about to deploy, buyers doing diligence on a purchased system, and builders who want an independent check.

For strategies in your head

Custom Build

From $10,000 · Scoped after a free call

  • Strategy design and implementation in Python or C++
  • Built against your broker, platform or API
  • Validation baked in from day one — not bolted on after
  • Delivered with the full audit report as standard
Book a scoping call

For traders with a defined edge and no time — or no desire — to fight with code, data feeds and execution plumbing.

How It Works

Four steps to a verdict

I

Submit

One paragraph on what you've built. Code is shared only after the NDA is signed.

II

Test

The full six-test battery on quality data with realistic execution. No shortcuts.

III

Review

The written report, then a 45-minute call through every result.

IV

Verdict

Fund it, fix it, or kill it — your call, made with evidence instead of hope.

Who Runs This

Built by a builder

[Your Name] — Founder, Edgeproof

[2–3 sentences in your own voice. Example shape: "I've spent the last X years building and stress-testing algorithmic strategies in Python and C++. Most of what I built, I ended up rejecting — and that discipline is the product. Edgeproof exists because the testing that kills bad strategies is the same testing that lets you trust a good one."]

[Optional: one line on tooling — data sources, frameworks, or platforms you work with.]

Get Started

Submit a strategy for review

Tell us what you've built in plain English. An email address is the only thing required — everything else is optional. You'll hear back within one business day.

Nothing you send here is shared with anyone, and source code is never requested before an NDA is in place.