Research standard
Evidence first. Scenarios second. Risk always.
This methodology explains how Basis turns market data into calculations and analysis while keeping uncertainty, source limits and invalidation visible.
Last reviewed 2026-08-22
Observation, calculation and interpretation stay separate
A reliable analysis distinguishes what a venue reported from what the platform calculated and what a researcher inferred. Basis keeps raw observations such as candles, trades and funding separate from derived indicators and narrative conclusions. A strong conclusion still remains a hypothesis until price confirms or invalidates it.
- Observation: venue-supplied prices, volume, trades, funding, open interest and positioning.
- Calculation: indicators, formulas, deltas, ratios, regimes and risk sizing derived from observations.
- Interpretation: scenarios, directional bias, confidence and invalidation levels assembled from the evidence.
Multi-timeframe analysis
The higher timeframe defines structure and market regime; the execution timeframe defines the setup; the lower timeframe may refine entry timing. A lower-timeframe trigger is not allowed to erase a conflicting higher-timeframe condition without explaining the conflict. Day-trading workflows generally prioritize daily or four-hour context, one-hour or fifteen-minute setup structure and five-minute or one-minute execution evidence.
Freshness, coverage and degraded data
Basis evaluates candle validity, gaps, freshness and provider fallback. A feed can be technically available yet unsuitable for a specific conclusion. When a venue does not publish a field, the platform should leave it unavailable rather than synthesize a plausible number. One-second bars are offered only where supported by the underlying public trade tape.
Backtests are experiments, not promises
Historical results depend on signal timing, bar resolution, fees, slippage, liquidity and fill assumptions. Basis backtests are intended to compare rules under declared assumptions. They do not prove that future orders will receive the same fills or that a strategy will remain effective after market conditions change.
- Avoid look-ahead data and evaluate signals only after their required inputs exist.
- Treat parameter searches and repeated trials as overfitting risk.
- Prefer walk-forward or out-of-sample checks when deciding whether a rule generalizes.
- Review trade distribution, drawdown and regime dependence—not only headline return.
AI analysis and confidence
The Basis agent can gather evidence and operate supported terminal tools, but model output can still be incomplete or wrong. Confidence is a calibrated summary of evidence quality and agreement, not a probability of profit and never a guarantee. High-confidence language must not replace an explicit invalidation level, risk limit or description of missing inputs.
Risk comes before the target
A position plan should define entry logic, invalidation, stop distance, position size and maximum loss before targets are considered. Basis presents research and planning tools; the user remains responsible for execution, suitability and compliance with local rules.
See the methodology in context.
Open a live market or follow the documented research workflow.