This page sets out the research framework for GoAlphaX studies. Individual reviews should identify which checks were completed, which were not possible and what remains uncertain.
1. Define the claim
Record the source and the specific proposition being tested. A statement about win rate differs from a claim about total return, risk-adjusted performance or capital preservation. We distinguish source statements from our own interpretation.
2. Write rules that can be reproduced
Specify the instruments, signal timing, entry and exit rules, order type, position sizing, trading hours and treatment of unfilled orders. Any rule absent from the original source must be identified as an assumption, not presented as something the creator said.
3. Describe and check the data
Identify the provider, coverage dates, timezone and price-adjustment basis. Check missing observations and inconsistent prices. Explain how these are handled. Where relevant, consider corporate actions, delisted securities and whether the tested universe was actually available at the time.
4. Model execution and capital
Include stated assumptions for commissions, bid–ask spreads, slippage and financing where applicable. A signal calculated at a candle’s close cannot justify an earlier fill inside that candle. When both a stop and target are touched within one bar, use finer data or disclose an explicit convention and its effect.
Report the capital committed and leverage assumptions. Options studies require suitable options prices and contract specifications, with treatment of expiry, exercise and assignment. A stock-price signal alone does not establish an options strategy’s profitability.
5. Report more than a headline
Show the trade count, win rate, average win and loss, net outcome, drawdowns and an appropriate benchmark. Explain whether returns are compounded, how idle cash is treated and whether results include open positions. Include losing periods rather than selecting only favourable examples.
6. Check how fragile the result is
Where data permits, test different market periods and reasonable changes to parameters and trading costs. Reserve unseen data for evaluation. Repeatedly modifying a strategy after seeing its test results weakens the independence of that test.
7. State a limited conclusion
Reviews may conclude: supported within this test; sensitive to assumptions; not supported within this test; or insufficient information to test. These are descriptions of the evidence, not buy or sell instructions.
Research records and corrections
Studies should show their analysis date, data cutoff and material revisions. Hypothetical examples, backtests and live results must be labelled separately. A meaningful correction should explain what changed and whether it affects the conclusion.
Historical simulations are hypothetical. They do not reproduce all conditions of actual trading and cannot guarantee future performance.
