We try to break your strategy before the market does.
Upload a trade list from TradingView, NinjaTrader, MetaTrader or your own code. Fifty tests look for luck, overfitting, a few lucky trades, thin cost buffers, regime changes and path risk, and tell you in plain words what survives.
The free autopsy needs only your trade list and a free account. No card.
Reads the exports you already have
Eight ways a backtest lies to you
Every report checks the same eight failure modes, each with its own tests and thresholds. Pick one to see what it looks like in a real report.
From export to verdict in about a minute
No column mapping, no settings you have to understand first. The engine reads the file, checks what it can prove, and says what it cannot.
- 1
Upload what you have
The trade list is enough for the free autopsy. Add the chart data and your script, and the engine can re-run your strategy itself.
Trade listChart dataPine or Python script - 2
We attack it
Thousands of reshuffled histories, permuted markets, parameter grids, walk-forward splits and cost stress, inside a 60-second budget.
- 3
Read the verdict
Pass, warning or fail for each failure mode, the evidence behind it, and what would have to change for the strategy to hold up.
Cheaper than a blown account
Most prop firm evaluations end on a drawdown limit, not on a missed target. Every failed attempt is a fee, and most traders buy the next one within weeks.
- ~$250
- average price of an evaluation
- 5–14%
- of evaluations end in a funded account
- ~7%
- of all buyers ever receive a payout
- 60–70%
- of failed attempts end on a drawdown breach
Industry figures for 2026 compiled by Track360 from firm disclosures and platform data.
A backtest rarely contains the day that ends accounts
Run the unchanged strategy through a synthetic crash market, calibrated on how real crashes behaved: the depth of 2008, the speed of 2020, the minutes of 6 May 2010.
- −34% in 23 trading days, February to March 2020
- −57% over 17 months, 2007 to 2009
- −9% within minutes on 6 May 2010
Not only what breaks. What to change.
After the verdict, a few concrete changes that follow from the measurements: the contract size your account can carry, what better execution is worth, which group of trades loses beyond chance, when the track record is long enough to size up. None of them tunes a parameter to the history — that would be the mistake we test for.
- Trade 2 contracts instead of 4The drawdown that 1 in 20 histories reaches stays within a quarter of the account.
- Plan with the recent edge$21 per trade since the break, not the $48 the full history suggests.
- Buy back the cost bufferEvery tick saved per side is worth $1,850 over these trades.
We test the tests
An engine that paints every strategy red passes any demo. Ours is measured against strategies whose truth we built in: broken ones it has to catch, honest ones it must leave alone.
checks passed on synthetic strategies with planted flaws, 40 seeds each
failed verdicts on honest control strategies across all methods
of 3,273 trades of a TradingView export reproduced to the cent by our re-run
planted failure modes found in eight strategies the engine had never seen
From the validation reports that ship with every release. The five misses are listed there with their causes.
Know before you fund it.
Run the free autopsy on your last backtest. If it survives, you will know why. If it does not, you will know that too.
Prop Firm Check
Your trades against the rules of a funded-account evaluation: the chance to pass, why attempts fail, and what getting funded costs you on average. Cheaper than a blown account.
1Your trades
2The evaluation rules
3Position size
How likely is your next challenge to pass?
Thousands of simulated evaluations, built from your own trading days in their real order: profit target, trailing drawdown, daily loss limit, minimum days. Then the number that matters: what getting funded costs you on average, in fees.
Crash Test
Your unchanged strategy on a synthetic crash market, calibrated on how real crashes behaved: the minutes of 2010, the month of 2020, the half-year of 2008. What it would have lost, and whether the account survives.
1Your strategy
2The crash
3Account
A backtest rarely contains the day that ends accounts.
The Crash Test builds it: the market falls, volatility bursts, sessions open with gaps, and your script trades through it bar by bar.
What happens to your strategy
The engine never grades a number it cannot support. It reads your file, decides what the data can prove, and runs every test that fits into the next minute.
Three levels of evidence
What you upload decides which questions can be answered. Each level adds tests on top of the one before.
Trade list
Free
- Luck against thousands of reshuffled histories
- How many variants you could have tried before this looks lucky
- Profit carried by a few trades
- Cost buffer against your broker's real fees
- Decay, drawdown and ruin
+ chart data
Pro
- Fills the bars never offered, stops inside gaps
- Regimes, exposure against timing, structural breaks
- Capacity, cost regimes, margin inside the trade
- Random entries in the same market as a benchmark
+ your script
Pro and Premium
- Your strategy re-run on permuted markets
- Parameter grid, walk-forward, overfitting probability
- Neighbouring timeframes it was never tuned on
- Premium: synthetic markets, volatility and cost shocks
Inside the minute
- First second
Reading the file
Format, columns, instrument, point value, currency and time zone are recognised from the export itself. Nine formats, no mapping.
- Before any test
The sample gate
How many independent trades the file really holds, after overlapping positions and trades that move together within a day. Too little history means "cannot say", never a pass.
- Seconds 2–15
Resampling and statistics
Block bootstrap, dependence-corrected t-statistics, deflated Sharpe, multiple-testing haircuts, break tests and drawdown distributions.
- Seconds 15–50
Re-running your strategy
With chart data and script, the engine runs your strategy itself: on permuted markets, across a parameter grid, walk-forward and on neighbouring timeframes.
- Last seconds
The verdict
Every failure mode gets pass, warning, fail or cannot say, with the evidence and the method card behind it.
Rules the engine keeps
Every method has a measured false-alarm rate
Before a test ships, it runs against honest strategies. None of the current methods fails an honest control.
Estimates never fail a strategy on their own
A test that has to guess an input, such as how many variants you tried, can warn but cannot deliver a fail by itself.
Short history means no statement
Below nine months or 120 trading days the report says nothing about time stability or regimes. Not a pass, not a fail.
Your files stay with the analysis
Uploads live in memory for the session and are not shared or used to train anything. Remove them with one click.
No trade list? We generate it.
Upload chart data and your script, and the engine runs the backtest itself with TradingView's own execution rules: individual entries, fills along the bar's path, order prices on the tick grid. On our benchmark strategies the re-run matches TradingView's exports trade for trade. If you upload an export as well, the same run checks that the file really comes from this script and this data.
Pine Script and Python
Paste or upload a Pine strategy as it is. Python strategies use a small, Pine-like interface and need nothing beyond math and numpy:
STRATEGY = dict(initial_capital=25_000, default_qty_value=1,
commission_type="cash_per_contract", commission_value=0.62, slippage=1)
INPUTS = dict(fast=dict(type="int", default=20, min=2, max=200, title="Fast SMA"),
slow=dict(type="int", default=60, min=5, max=400, title="Slow SMA"))
def on_bar(s):
fast, slow = s.sma(s.close, s.inputs.fast), s.sma(s.close, s.inputs.slow)
if s.crossover(fast, slow):
s.strategy.entry("L", "long")
if s.crossunder(fast, slow):
s.strategy.close("L")The same strategy written in Pine and in Python produces identical trades in our engine, to the cent.
Fifty tests, eight failure modes
Every test in the engine, grouped by the question it answers. Open a failure mode for its methods; open a method for how it works and where it stops working.
See it on your own backtest.
The free plan runs every trade-list test.
Pick the level of evidence
Every plan runs the full engine on what you upload. The difference is how much you can upload, and how hard we push.
A year of Pro costs less than a single prop firm evaluation fee. Check your chances first.
Questions about plans
Can I try Pro first? Every Pro test is visible in the free report as a locked row, with the question it would answer. Upgrade when your trade list leaves questions open.
Is this investment advice? No. The engine finds weaknesses that are visible in the data you upload. A strategy that passes every test can still lose money.
Terms of Service
The full terms of service will be published here before paid plans open.
StressTestQuant is a research tool. Nothing presented on this website is financial advice.
Questions, answered
About the product, your data, the tests and the plans.