A signal is useful only when its data, rule, timeframe, invalidation, and execution assumptions are explicit.
The anatomy of a defensible signal
A signal should state its observation time, instrument, source, rule, and invalidation. “Bullish” alone omits all five. A 1-hour momentum label can coexist with a negative 7-day trend because the horizons differ.
Signals are classifications of available data, not causal explanations. Rising price and high reported volume do not prove accumulation by informed buyers. Falling price near a daily low does not identify who sold. The same pattern can result from news, liquidations, thin liquidity, or routine rebalancing.
How the home scanner forms labels
The market scanner awards positive and negative points from 1-hour, 24-hour, and 7-day price changes, 24-hour range position, and total-volume-to-market-cap intensity. A score of at least 3 is labeled bullish, at most −3 bearish, and values between are neutral. It can also attach early-reversal and combination labels based on additional price, range, volume-proxy, and volatility conditions.
Its pressure display is clamp(50 + score × 8), and the overall pressure is the average across analyzed assets. This is a heuristic summary, not executed buyer-versus-seller flow. Total volume divided by market cap is a volume proxy, not TVL or order-book depth. The accumulation and distribution names are shorthand for rule matches, not proof of wallet identity or institutional activity. There is no social/news sentiment feed.
Cross-check without changing the claim
Use /charts when you want the separate Ichimoku and Keltner framework: cloud location supplies directional permission and two closes beyond a Keltner boundary supply confirmation. Use /dashboard for DeepBook order-book context. There, bid and ask notionals are price × quantity resting in the book; they are not long and short positions, and visible depth can be canceled before execution.
These tools add context but cannot turn a home signal into order-flow evidence. The /signals UI and server alerts need not share logic, so avoid asserting exact thresholds unless that surface documents them.
Turn a label into a testable plan
Write a conditional plan: if the signal persists on a closed candle and context agrees, consider an entry; at invalidation, exit or reassess. Define maximum loss before size, including fees, spread, slippage, and alert delay.
Review a complete sample rather than memorable winners. Record unchanged rules and timestamps. Separate signal quality from execution quality; a handful of examples does not establish performance.
- Observation: what exactly changed, and over which interval?
- Confirmation: which independent condition must persist?
- Invalidation: what observable fact makes the setup no longer valid?
Worked conditional signal plan
This fictional snapshot illustrates interpretation, not a recommendation or a live app result.
- Scanner evidence
- 1h +1.2%; 24h +3.4%; range position 88%
- Heuristic score
- +4 points, so pressure = 50 + (4 × 8) = 82
- Chart cross-check
- Price above cloud, but only one close above Keltner upper
- Plan and risk
- Wait for rule confirmation; entry $52, invalidation $50, risk $2/unit
The home label is strongly positive under its heuristic, while the separate chart rule is not yet confirmed. With a hypothetical $100 maximum loss, 50 units is the arithmetic ceiling before fees and slippage; choosing not to trade remains valid.
Common interpretation traps
- Treating an alert as an instruction and entering without a timeframe, invalidation, or maximum-loss limit.
- Describing heuristic pressure or named accumulation patterns as verified executed flow or trader positioning.
- Backtesting only remembered alerts, changing rules after outcomes, or ignoring spread, fees, slippage, and alert delay.
Frequently asked questions
Is bullish the same as a buy recommendation?+
No. It is a classification produced by a rule set. Suitability, risk capacity, execution, and invalidation are separate decisions.
Why might the home scanner and chart disagree?+
They use different inputs and rules. The scanner scores multi-window changes, range, and a volume proxy; the chart uses candle-based Ichimoku and Keltner calculations.
Can DeepBook bid notional confirm long positioning?+
No. Bid notional is resting bid price multiplied by quantity. It does not identify leveraged longs, intent, or whether the order will trade rather than be canceled.