Backtesting comparison methodology
The unit of comparison is a named workflow or edition, not a claim that one platform is universally best.
What was verified
We reviewed official public documentation on 17 October 2026 for 10 selected workflows. Records distinguish coded engines, visual strategy generation, manual replay and no-code options testing. We did not open trial accounts, buy licences, install providers, execute provider imports or benchmark their speed. Suggested verification tasks are clearly labelled tasks you can run, not tests we completed.
Asset-class evidence
A true asset field means the sources document the workflow or instrument model. It does not mean data is included or every broker supports it. False means outside the selected workflow, such as share trading in a selected options-only workflow. Null means not established. A generic array library, a chart of an underlying or a symbol picker is not evidence of contract multipliers, Forex margin, historical option chains, expiry, assignment or multi-leg fills.
Costs, licences and exports
Software licence, market data, cloud computation, exchange fees and trading commissions are separate categories. A perpetual licence includes a stated update period; it is not a perpetual data subscription. An annual-billing monthly equivalent is labelled as such. Unknown prices remain null rather than zero; no mixed-billing cheapest ranking is calculated. Vectorbt uses fair-code CommonsClause terms; it is not described as unrestricted open source.
Result export means documented access to result records or reports, including programmatic access. It does not imply a one-click CSV download. Code export is distinguished from result export; current format, plan entitlement and implementation remain explicit in each profile. We do not claim native export acceptance.
Our original backtest laboratory
The local SMA tool has its own explicitly bounded execution and rounding model. It uses a completed-close target and next available daily open, fixed whole units, cash checks, buy/sell commission and adverse price impact. The final position is marked, not forcibly sold. It provides full order/equity CSVs and a resumable setup, separately from provider research.
Testing covers hand ledgers, independent Decimal calculations, prefix invariance (adding future bars does not change earlier fills), missing/invalid OHLC, dates, costs, insufficient cash, no-trade cases and actual browser exports. A synthetic example is invented data, not a market performance record or an optimized strategy.
Search research and updates
Topic demand uses DataForSEO US/English estimates retrieved 16 October 2026. These are not traffic forecasts, and nearby variants are not summed. A zero keyword-difficulty value does not prove no competition. Version 2026-10-17.1 is this initial snapshot; changing provider terms should be rechecked before use. Corrections can be submitted with the edition and official source through our contact page. No automatic monitoring is promised.