Backtesting Methodology and Assumptions | QANTERION

Last reviewed: 2026-08-09 · Owner: QANTERION engineering, [email protected]

A backtest is only as trustworthy as the assumptions behind it, and most of those assumptions are invisible in the number that gets advertised. This page states which assumptions QANTERION controls, which belong to the strategy execution core, and which are not modeled at all.

Where a value is not disclosed, this page says “not published” rather than describing something plausible. An invented assumption is worse than a missing one.

Who computes what

QANTERION is not a self-contained backtesting engine. It is a terminal in front of a separate strategy execution core, and the split matters for how you should read any figure it displays.

ConcernOwner
Market dataStrategy execution core
Strategy computationStrategy execution core
Simulated fillsStrategy execution core
Tenant-level trade ledgerStrategy execution core
Authentication and account scopeQANTERION platform
Capital pre-checks and allocation reservationQANTERION platform
Settlement receipts and double-entry postingQANTERION platform
Trade-tape and position projection shown in the terminalQANTERION platform
Displayed risk metricsQANTERION platform

Anything in the “strategy execution core” column is supplied by a system QANTERION integrates with rather than one it operates. We do not restate its internal assumptions as our own.

Metric formulas

These are computed by QANTERION and are fully disclosed. Percentage inputs stay in percentage points, so the resulting ratios are dimensionless.

Sharpe ratio

Sharpe = (annualized return − risk-free rate) / volatility

Requires volatility strictly greater than zero. A zero or negative volatility input produces no value rather than a division-by-zero artifact or a silently substituted default.

Maximum drawdown

Max drawdown % = ((peak − trough) / peak) × 100

Requires peak greater than zero and trough not above peak. A trough above its peak is rejected as invalid input rather than returned as a negative drawdown.

Calmar ratio

Calmar = annualized return / max drawdown %

Requires a max drawdown strictly greater than zero.

Invalid inputs return no value and an explicit error for the affected field. Nothing is coerced to a default. A metric that quietly becomes 0 when its input is missing is indistinguishable from a real result of 0.

You can run these on your own numbers with the Sharpe ratio and drawdown calculator.

What the platform guarantees about recorded activity

These are properties of QANTERION’s own accounting, enforced in code:

  • No synthetic fills. If the strategy execution core is unreachable, the platform reports the failure. It never generates local fills to fill a gap. At most it briefly serves the last successfully retrieved activity.
  • Run-scoped activity. Displayed trades are filtered to those after the current run’s start, so a previous run’s residual activity cannot inflate the current one.
  • Deterministic settlement receipts. A receipt identifier is derived from the tenant, strategy, instance, and a stable start boundary. A retry or a restart produces the same identifier rather than a second receipt.
  • Idempotent ledger posting. Entries post against the receipt, so a replayed settlement does not double-count.
  • Acknowledgement is not execution. A platform-side accepted means the command was taken, not that it ran. Execution state comes only from the strategy execution core’s observation.

Simulation, paper, and live are different things

  • Simulation / paper is the default. Live trading is refused unless a deployment explicitly enables it with a broker and market-data provider configured.
  • Paper and live are separated at the account scope, resolving to different tenant scopes at the strategy execution core rather than sharing state behind a display toggle.
  • Paper results are not live results. Paper runs do not experience real queue position, partial fills against real liquidity, venue outages, or the funding and borrow costs of a real position.

Not modeled, not published

The following are standard backtest concerns. They are listed because their absence is the single most useful thing this page can tell you.

Not currently published by QANTERION. These are the strategy execution core’s assumptions, and we will not restate them as ours:

  • historical data source, vendor, and coverage period
  • time zone and bar-timing convention
  • corporate actions and symbol-change handling for equities
  • survivorship-bias and look-ahead controls
  • spread, slippage, and market-impact assumptions
  • liquidity and fill assumptions, including partial fills
  • leverage, margin, and liquidation behavior in simulation
  • in-sample and out-of-sample separation

Not modeled:

  • Trading fees, funding rates, borrow costs, and financing costs are not applied by the QANTERION platform to displayed strategy figures. A fee schedule exists in the data model but is not currently applied to any displayed result. Treat displayed figures as gross of these costs unless the strategy execution core states otherwise.

No reproducible worked example is published yet. One is planned. We are not publishing a fabricated example in the meantime.

Limitations

Historical simulation describes what a rule would have done on data that has already resolved. It cannot establish what it will do next. Past and simulated performance does not indicate future results, and a strategy that performed well in simulation can lose money in live markets.

Simulated results carry the further limitation that they were not achieved with capital at risk. They do not reflect the effect of an order on a real market, and they do not reflect how a person behaves during a drawdown they are actually living through.

Nothing on this page is investment advice. See the risk disclosure.

Corrections

If a figure or an assumption on this page is wrong, write to [email protected] naming the claim. Corrections are made on the page with the reviewed date updated.