Signal Methodology

Crypto signals are computed by the MarktBit DataHub from mature market events — real trade ticks collected from supported exchanges and stored with a maturity delay. Stock signals use historical daily prices from our aggregated stock feed (Yahoo/Finnhub-class sources with a gate overlay).

We do not generate signals from fixtures, random values, or simulated prices. If the upstream data is unavailable for an instrument, the engine reports "insufficient history" instead of emitting a signal.

The crypto engine evaluates a rolling window of recent ticks per symbol (default: the last 60 events, minimum 3). It computes momentum (percentage change across the window), range position (where the current price sits between the window low and high), plus RSI(14), MACD histogram, SMA(20)/SMA(50) and realized volatility where the window supports them.

A BUY signal requires momentum ≥ +1% AND a range position ≥ 0.65. A SELL signal requires momentum ≤ −1% AND a range position ≤ 0.35. Confidence scales with the strength of the measured momentum. Signals outside these thresholds are not emitted.

The indicator values shown alongside each signal are the same values the engine computed at emission time — they are part of the stored signal record, not re-computed or filled in at display time.

The performance endpoint compares each emitted signal against the latest mature price for that symbol, once the signal has aged past a minimum window (default 60 minutes). A signal counts as a "hit" when the price moved in the signaled direction beyond a small noise floor; it counts as a "miss" when it moved the opposite way, and "flat" inside the noise floor.

The hit rate is reported together with its sample size. A small sample is not evidence of edge, and we display it accordingly. Signals younger than the minimum age are reported as pending, not counted.

MarktBit signals are a measurement of recent market behavior — they are not predictions, not financial advice, and not a promise of profit. Past hit rates, including the ones we publish on the performance page, do not guarantee future results.

Momentum-style strategies can underperform sharply in ranging or reversing markets. Always size positions for the possibility that any individual signal is wrong.

Where AI-generated commentary appears, it is produced by a provider fallback chain and is labeled as generated content. When no provider is available, the UI shows the unavailability instead of fabricated analysis. AI output never overrides the measured indicator values.