趋势类
判断趋势的方向和强度
| Indicator | Primary Use | Pros | Cons |
|---|---|---|---|
| Moving Averages (SMA/EMA/HMA) | Trend direction and dynamic support/resistance. | Simple, fast, universal quantitative baseline. | Significant lag; severe whipsaws in chop. |
| MACD | Identifying momentum shifts and crossover signals. | Combines trend and momentum in one oscillator. | Double-lag; prone to false divergences. |
| ADX / DMI | Regime detection (trending vs. ranging filter). | Quantifies trend strength independent of direction. | Heavy lag; cannot predict breakout starts. |
| Aroon Oscillator | Detecting early trend inception. | Tracks time elapsed rather than price magnitude. | High false breakout rate in sideways channels. |
| Ichimoku Cloud | Complete trend bias and dynamic S/R framework. | Holistic framework; projects future equilibrium levels. | Over-parameterized; Cloud bounds lag heavily. |
| Parabolic SAR | Dynamic trailing stop-loss and trend exits. | Rule-based exit logic that accelerates with trend. | Triggers lethal whipsaws in range-bound markets. |
| TRIX | Macro trend-cycle filtering and divergence. | Extreme noise filtration against liquidation wicks. | Large phase delay; late entry on fast moves. |
| Linear Regression | Statistical fair value and mean-reversion modeling. | Lower lag than MAs; provides mathematical slope. | Highly distorted by outlier spikes and wicks. |
| KST | Major multi-cycle momentum reversals. | Synthesizes four timeframes for cleaner cycle turns. | Slow response to abrupt market regime shifts. |
| Coppock Curve | Pinpointing macro cycle bottoms. | Highly reliable for spotting secular bull turns. | Strictly macro-scale; useless for swing or intraday setups. |
- 灵敏度排序(由高到低): HMA > EMA > SMA
- 滞后性排序(由大到小): SMA > EMA > HMA
- HMA兼具极高的灵敏度(低滞后)与出色的平滑度(抗噪声)
动量/振荡类
衡量价格变化的速度,以及超买超卖的程度。
| Indicator | Primary Use | Pros | Cons |
|---|---|---|---|
| RSI | Mean reversion and divergence detection. | Bounded (0–100); stable, standardized baseline. | Saturates at extremes during strong trends. |
| Stochastic | Timing range turns via %K/%D crossovers. | Highly responsive to short-term cycles. | High noise and whipsaws in volatile regimes. |
| Williams %R | Spotting immediate market exhaustion. | Fast detection of local peaks and troughs. | Causes premature exits during sustained breakouts. |
| CCI | Detecting extreme deviations and breakouts. | Unbounded scale captures explosive rallies cleanly. | Subjective thresholds; distorted by crypto wicks. |
| ROC | Cross-asset momentum ranking. | Scale-independent percentage calculation. | Distorted by the rolling window drop-off effect. |
| Momentum | Measuring raw price acceleration. | Simplest unweighted time-series velocity metric. | Scale-dependent; unsuited for cross-asset scans. |
| CMO | Capturing unsmoothed directional velocity. | Faster than RSI; cleanly bounded (-100 to +100). | Generates high turnover and false breaks. |
| Ultimate Osc. | Multi-timeframe divergence filtering. | Blends 3 timeframes to reduce single-period noise. | Slower response; higher overfitting risk. |
| StochRSI | Scalping triggers and micro-reversals. | Maximum sensitivity to early momentum shifts. | Severe whipsaws; pins at 0 or 1 during trends. |
| SROC | Smoothed cyclical trend-following. | Eliminates standard ROC drop-off noise. | Double-smoothing introduces execution phase lag. |
| DPO | Identifying cycle lengths and periodicity. | Removes trend bias to isolate pure cycles. | Inherent displacement lag makes real-time trading hard. |
| Twiggs Mom. | Smoothed zero-line crossover timing. | EMA-weighted to prevent abrupt drop-off spikes. | Delayed reactions during sharp flash moves. |
成交量/资金流类
用量价关系判断吸筹或派发。
| Indicator | Pros | Cons | Primary Use |
|---|---|---|---|
| OBV | Simple, fast; reliable for spotting macro divergences. | Binary volume assignment; easily skewed by flash spikes. | High-timeframe trend confirmation and divergence detection. |
| A/D Line | Weights volume by candle close location to avoid binary bias. | Cumulative and unbounded; ignores inter-bar gaps. | Identifying hidden accumulation during range consolidations. |
| CMF | Bounded ( to ); provides a clean normalized ML feature. | Suffers from cliff drops when volume spikes exit the window. | Regime classification (net buying vs. selling dominance). |
| Chaikin Osc | Adds momentum to A/D; leads price turns rapidly. | High noise; generates frequent whipsaws in sideways chop. | Timing short-term reversals and mean-reversion execution. |
| MFI | Volume-weighted RSI; bounded (–) with built-in volume filtering. | Pins at extremes in parabolic trends; distorted by wash trading. | Spotting liquidity exhaustion and overbought/oversold boundaries. |
| Force Index | Combines price delta and volume to model true kinetic thrust. | Unbounded; requires rolling z-score normalization for models. | Measuring breakout velocity and filtering momentum entries. |
| EMV | Directly quantifies market friction and order-book thinness. | Spikes erratically during illiquid micro-ranges or flash events. | Detecting liquidity vacuums and paths of least resistance. |
| PVT | Scales volume by percentage price change rather than binary ticks. | Cumulative drift complicates cross-sectional pair ranking. | Tracking relative volume accumulation with proportional weighting. |
| NVI / PVI | Isolates smart money positioning (NVI) from retail FOMO (PVI). | 24/7 continuous crypto volume makes raw signals noisy. | Filtering for institutional accumulation on quiet, low-volume days. |
| Volume Osc | Directly tracks volume expansion/contraction cycles. | Direction-agnostic; provides no directional bias on its own. | Breakout confirmation and volatility-based position sizing. |
| Twiggs MF | Fixes CMF gap errors using True Range and exponential smoothing. | Slightly higher lag; rarely built into standard quant libraries. | Low-noise capital flow filtering in volatile trend strategies. |
| Williams A/D | Factors in True Range gaps without needing exchange volume feeds. | Uses synthetic price-range flow rather than real traded volume. | Flow proxy for new listings or illiquid pairs lacking clean volume data. |