Strategy Backtest Auditor
Tells you what your backtest actually proves, which is usually less than you think
- 9hard rules
- Things it must never do, enforced in every run
- 5handover rules
- Conditions where it stops and gives the work to a person
- 10output fields
- A structured result, not a paragraph you have to re-read
- ≤6steps per run
- Bounded work: at most 4 tool calls, enforced by the runtime
Read-only by design. This agent has no path to sending, spending, publishing or committing anything. Where that limit is the product, removing it would remove the reason to trust the output.
Overview
Every platform will run you a backtest. None of them will tell you the result is meaningless, and most of them are, for reasons that are mechanical and never checked.
This agent audits the result rather than producing another one. It checks whether costs and slippage were included, and re-states the outcome with them if they were not — because a great many edges are smaller than the spread. It reports how much of the return came from the single best trade, whether the sample contains enough independent events to mean anything, whether the rule could only have been chosen after seeing the data, and whether the universe quietly excludes what failed.
It reaches no conclusion about whether you should trade the strategy. It reports what the evidence supports and what it does not, and the honest answer is often that the sample is too small to say. That answer is worth more than another equity curve, and it is the one nothing else in this category will give you.
Benefits
You stop paying for edges that are not there
The cheapest trade you will ever make is the one an audit talked you out of.
You learn which test to run next
Every finding names what would actually settle it.
It is honest when the answer is boring
"Your sample cannot support this" is a real result, and nothing else will tell you.
What it does
Re-states the result with costs
Fees, spread and slippage, applied. A great many published edges are smaller than the cost of taking them.
Shows how much was one trade
If the best trade is most of the return, there is no strategy — there is one trade.
Counts independent events, not trades
Two hundred trades on correlated instruments in one trend is closer to one observation than to two hundred.
Looks for look-ahead and survivorship
A rule that used information unavailable at the time, or a universe that quietly drops what failed.
Flags parameter fitting
How many variants were tried before this one, and what that does to the result.
Concludes nothing about trading it
It reports what the evidence supports. Whether to risk money on it is not its decision.
What this changes
What this changes about your week
- It takes over
- Checking whether a backtest actually proves anything before money goes near it.
- Done today by
- Nobody. Every platform produces the backtest and none audits it.
- How often
- Before trading a strategy, and before buying someone else's.
- Typical manual effort
- Doing it properly needs statistics most traders were never taught.
Why Finance & Accounting in particular. Most published edges are smaller than the costs of taking them, or are one lucky trade. The cheapest trade you will ever make is the one an audit talked you out of.
Runs unattended
Started by you or by an event, and it finishes on its own. Nothing waits for someone to be at a desk.
The same standard every time
The two-hundredth item is held to the bar the first one was. Consistency is the part people cannot sustain.
It cannot act on its own
Strategy Backtest Auditor has no path to sending, spending or committing. That limit is why its output is safe to act on.
How it runs
This agent runs server-side through the PROMIVO runtime. Each run is logged step by step and every tool call is permission-checked before it executes.
- 1InputValidated against the agent input schema
- 2Modelclaude-opus-5 with automatic fallback
- 3Toolsextract_structured_data
- 4PermissionsRuns READ actions automatically; anything higher waits for approval
- 5OutputStructured result matching the output schema
- 6LoggingTokens, duration and every step recorded
Demo
Demo dataIllustrative sample output, abridged.
Input
{
"rules": "…",
"period": "2023–2026",
"universe": "Top 50 by volume",
"variantsTried": 0
}Output
{
"escalate": true,
"findings": [
{
"issue": "costs",
"detail": "Costs were not applied. Three quarters of the headline return is consumed by them.",
"severity": "fatal",
"whatWouldSettleIt": "Re-run with your venue's actual fee tier and realistic slippage."
},
{
"issue": "fitting",
"detail": "Variants tried was given as 0, which is rarely true. If the threshold was chosen after looking at this data, the result cannot be read at all.",
"severity": "fatal",
"whatWouldSettleIt": "State how many thresholds were tested, then re-test the chosen one on data never examined."
}
],
"worstRun": "Eleven consecutive losses; deepest drawdown 34% over four months.",
"disclaimer": "An audit of the evidence you supplied. Not investment advice, not an opinion on whether to trade this strategy, and not a forecast of any kind.",
"concentration": {
"comment": "Excluding the best five trades, the strategy loses money. This is five observations, not two hundred and twelve.",
"bestFiveShare": "104% of net return",
"bestTradeShare": "58% of net return"
},
"whatItDoesNot": [
"That it has an edge. After costs, almost all of the return is five trades.",
"Anything about future performance. Nothing in a backtest can establish that."
],
"escalationReason": "Two fatal findings: costs were omitted and the fitting history is unknown.",
"whatThisSupports": [
"That this rule was profitable on this data before costs."
],
"independentEvents": "212 trades, but 184 opened during two trending periods. Closer to a handful of observations.",
"restatedWithCosts": "Headline +142%. Applying the 0.1% taker fee and one tick of slippage per side across 212 trades: +31%."
}Requirements
Integrations
No integrations required.
Before you start
- The strategy rules as stated
- Its results — trades or an equity series
- The costs assumed, if any
- Requires the All-Access plan or higher
Use cases
Before risking capital
Find out what the backtest actually established.
Before buying a strategy or bot
Audit the seller's published results.
A live strategy underperforming
See whether the backtest ever supported it.
Pricing
All-Access
Billed monthly through your PROMIVO subscription. Cancel at any time.
Runs consume your plan allowance for agent executions and tokens. See plan limits
FAQ
Will it tell me if the strategy is good?
It will tell you what your evidence supports. Whether to trade something is a decision about your own money and risk, and that is investment advice — not something this agent gives or is permitted to give.
Can it improve my strategy?
No, and be suspicious of anything offering to. Adjusting a rule until the backtest looks better is precisely how a strategy becomes fitted to the past and useless in future.
Does it run backtests?
No. It audits results you already have. A second engine producing a second curve would add another number to argue about rather than an answer.
Reviews
Reviews open once customers have run this agent.
Need this agent to work differently?
Tell us what to change and our team will scope a customised version for your business.
Customize this agent