Nobody sells a strategy that failed its backtest.
Every strategy for sale has a beautiful curve. What it does not have is a check run by someone with no reason to make it look good. Now it can.
Most backtests fail for boring reasons
Not fraud. Plumbing. A backtest run on today's index members never sees the companies that were removed. One run on restated fundamentals knows things nobody knew at the time. One that fills at the same bar's close trades on information that didn't exist yet. Each leak is worth points of CAGR, and none of them show up in the equity curve.
Survivorship
Testing on the current S&P 500 or Nasdaq-100 silently drops every delisted, acquired, and failed name, usually the worst performers. Our membership is reconstructed as of each historical date.
Lookahead
Fundamentals enter our tests when they were actually public, shares outstanding are the historical figures, and execution is at the next open, never the same bar that produced the signal.
One lucky window
A strategy tuned on a single start and end date may be an artifact of both. We roll the window, split out the fresh era, remove the biggest winners one at a time, and delay the trades to see what survives.
How it works
You send us your strategy in one form and pay $499 to start. We rebuild it and test it, and email you a private link to the finished report. You never need an account.
Send the thesis
Tell us the rules, the universe, the window, and the results you claim. You pay $499 to start, not the full price up front.
We check it can run
We read your strategy and confirm the test can be run fairly on it. If it is too incomplete to test, we refund you in full and tell you what is missing.
We rebuild it twice
A second, independent implementation rebuilds your strategy from the written thesis alone, on data as it was known at the time. No figure counts until the two implementations agree, and the unmodified strategy reproduces first.
You get the report
Status updates by email as it moves. The finished report arrives through a private link and unlocks in full once the final $500 is paid. Forward it to anyone. Make it public only if you choose to.
A replication is not a backtest.
A backtest grades its own homework
You run your code on data you assembled and read the number it gives you. Its failure modes are invisible from inside: the data usually holds only companies that still exist, fundamentals as they were later restated, index membership as it stands today projected backward, and your own code deciding what counts as a fill.
A replication is a rebuild by someone else
- A second implementation is written from your written thesis alone, blind to your code and your numbers. Your stated result has to reproduce first, and the two implementations must agree before any figure counts.
- The tests are built to kill, not to confirm: walk-forward, a fresh-era split, leaving one stock out at a time, placebos, parameter plateau, delay and cost stress.
- The verdict comes from a party with no stake in the answer. It can be "does not replicate." It has been, on our own strategies.
The data is the market as it was known on the day
S&P 500 and Nasdaq-100 membership as of each historical date, built from index announcements and filings.
Companies that later failed or were delisted are still in, with their final returns.
So a company's size in the test is its size then, not its size now.
Stamped with when they were actually filed, so a number exists in the test only after it existed in the world.
That is why the same strategy can show a spectacular result in a backtest and fall apart in a replication. The rules did not change. The data stopped helping it.
What we can model
A strategy is more than a signal. How money moves in and out, what the portfolio holds, and what taxes and costs take along the way all change the result. These are modeling choices you set in your thesis, and they are included with every report.
A lump sum, or scheduled contributions: weekly, monthly, quarterly, or annual dollar-cost averaging. Withdrawals as a fixed amount or a percentage of the portfolio, plus one-time deposits or withdrawals in the middle of the test. When money moves, we report the money-weighted return alongside the time-weighted return, because the two can tell different stories.
Individual stocks and ETFs. Bonds and bond funds. A cash or T-bill sleeve. Commodities, gold, and REITs. Several asset classes in one portfolio, tested together. Leveraged and inverse ETFs, with their daily reset handled honestly.
After-tax results, with short-term and long-term gains shown separately, and taxable versus tax-advantaged account framing. Dividend reinvestment on or off. Commissions, slippage, and execution delay. Ongoing fund fees shown as their own drag line, and inflation-adjusted (real) returns next to nominal ones.
Rebalancing on a calendar schedule, or only when weights drift past a tolerance band. Glide paths, where the weights change over time. Volatility targeting, where exposure scales with measured risk.
Where it matters, reports also show benchmark-relative statistics. Included in every report is the run with the settings in your thesis: your contribution schedule, your account type with after-tax results and the short-term and long-term gain split, dividend handling, rebalancing rules, and a glide path if your rules specify one. The studies that go further, full tax modeling across both account types, a glide path compared against fixed weights, and withdrawal sustainability, are separate checks priced on request.
Every report runs the same battery
Every report runs the same set of checks, the same way. A free public report gets exactly the same treatment as a paid private one.
The independent rebuild matches the claim within tolerance and holds up across the battery, windows, eras, delays, and costs.
The headline number reproduces, but something narrows it: one era carries it, one ticker carries it, or it sits on a parameter cliff.
The claim does not survive point-in-time data, honest execution, or the out-of-sample splits. The report names what explains the gap.
Free published reports you can read first
Traders post strategy claims in public every day: a CAGR, a Sharpe ratio, an equity curve. We take some of the complete, specific ones, replicate them, and publish the full report, whether the claim holds up or not. Anyone can read them, and they show you exactly what a paid report looks like before you buy one.
That openness applies only to strategies that are already public. A report you commission is different: it is 100% confidential, and the strategy and the report are your intellectual property. We never publish it, share it, or reuse it for anyone else. It becomes public only if you decide to make it public.
One strategy. One rebuild. One verdict.
$999 launch price per strategy: $499 to start, $500 when the report is ready, and the start payment refunded in full if we can't run the test. Read a free published report first if you want to see the work before you pay.