Crypto backtest cost-model worksheet

Blank research template. It records assumptions; it does not recommend a trade, confirm a venue fill, or establish future performance. Keep private account identifiers and credentials out of this file.

1. Freeze the research record

2. Name the instrument and accounting convention

3. Preserve source-backed assumptions

Write rates with their original units. Distinguish a percentage such as 0.05% from its decimal form 0.0005. A rate-on-notional fee can be estimated as executed notional multiplied by the decimal rate only when the instrument's schedule defines that basis. If the fee basis, account tier or contract rule is unknown, record it as unknown or state the explicit assumption.

4. Record executions separately

Record partial, rejected, canceled and unfilled orders separately when the research question depends on whether an order could execute. Do not infer a fill merely because a candle touched a price.

5. Separate spread from additional slippage

If the fill model already moves a midpoint reference to the bid or ask, do not subtract that same spread again as a separate cost. Label modeled price movement and separately charged fee lines so another reader can reproduce each one.

6. Align carry charges with eligible events

Record only settlements for which the position is eligible under the contract. Use the venue's specified interval, sign convention and price basis. Keep a missing rate unknown; do not silently substitute zero.

7. Declare baseline and higher-cost cases before comparison

Replay the complete strategy for each case. Preserve the same strategy and dataset versions unless a change is the declared test. A cost change can alter quantity, fills, exits and later equity; subtracting one charge from a selected trade is not a substitute for a full replay.

8. Reconcile and report each replay

9. Review before interpreting

• Rates, percentage/decimal conversions, contract units and fee-bearing notionals reconcile.
• Entry and exit sides are both included; partial and unfilled orders remain visible where relevant.
• Spread is not duplicated inside slippage or fill prices.
• Funding is aligned to eligible UTC settlement events and missing values remain explicit.
• Quantity, tick/lot rounding, margin, liquidation and gap behavior match the research model's stated limits.
• Baseline and higher-cost cases were declared before interpreting outcomes.
• The full strategy was replayed for each case; failed and inconclusive results are retained.
• Synthetic, imported, public-observation, simulated and actual execution evidence are labeled separately.

Decision and unresolved evidence

For product-model limits, see SuperQuant research methodology. The template contains no current venue rates; verify live documentation for the exact account and contract.