Metrics

This is the deepest endpoint in the API: the complete output of the metrics engine for one pair, organized into seven families. Everything the screener can rank by, every number on the chart page's info panels, comes from here. The engine sweeps the entire market every 30 seconds; windowed families are computed per timeframe, so the same request with a different interval answers a different question (is this pair unusual this minute vs this week).

Request

curl "https://osiristrader.com/api/v1/metrics/SOLUSDT?interval=5m" \
  -H "Authorization: Bearer osk_YOUR_KEY"

curl "https://osiristrader.com/api/v1/metrics?symbols=BTCUSDT,ETHUSDT,SOLUSDT&interval=5m" \
  -H "Authorization: Bearer osk_YOUR_KEY"

Response

{
  "symbol": "SOLUSDT",
  "interval": "5m",
  "data": {
    "volatility":  { "NATR": 0.214, "RR": 0.0011, "PK": 0.0013, "GK": 0.0012, "RS": 0.0012, "YZ": 0.0014, "SURGE": 1.21 },
    "volsurge":    { "SCORE": 0.71, "VOL_RATIO": 1.10, "TR_RATIO": 1.42, "RET_ZSCORE": 0.80 },
    "volume":      { "USDT_VOLUME": 5391077, "AVG_VOLUME": 449256, "VOL_STDDEV": 112340, "VOL_GROWTH": 1.08, "VWAP": 183.42, "OBV": 120443 },
    "trades":      { "TC_GROWTH": 1.05, "QV_GROWTH": 1.12, "TC_ZSCORE": 0.4, "QV_ZSCORE": 0.7, "FLOW": -0.04, "BUY_RATIO": 0.48, "SIZE_GROWTH": 1.30, "AVG_TRADE_SIZE": 842.1 },
    "correlation": { "retCorr": 0.87, "realTimeVol": 1.7, "natrRatio": 1.9, "surgeRatio": 1.2 },
    "funding":     { "RATE": 0.0000410, "PREMIUM": -0.0001, "AVG_24H": 0.00005, "TREND": -0.00001, "APR": 4.49, "NEXT": 1783900800000 },
    "trend": {
      "local": "Uptrend", "global": "Sideways",
      "scores": { "LOCAL_SCORE": 0.42, "GLOBAL_SCORE": 0.05, "ALIGNMENT": 0.05, "DIVERGENCE": 0.37, "CONSENSUS": 0.43, "MOMENTUM": 0.11 }
    }
  }
}

volatility: seven estimators

Realized volatility has no single correct formula; each classic estimator uses different parts of the bar and has different strengths. We compute all of them over a rolling window of bars at your interval, so you can pick the one matching your use case (or use their disagreement as a signal in itself: range-based estimators running hot vs close-based ones flags intrabar chop without direction).

volsurge: normalized surge detection

The surge family answers one question: is this pair doing something unusual right now? Three raw ratios live on very different scales (volume pace hits 5-20× during a real surge, range expansion rarely tops 2-4×, return z-scores reach 3-6), so each is divided by its own "clear surge" level before SCORE takes their maximum. The result: SCORE ≥ 1.0 means at least one dimension is clearly surging, whichever one it is. The current bar is pace-adjusted (projected to a full bucket by elapsed time), so a surge is visible seconds into a bar, not only after it closes.

volume

trades: pace-adjusted order flow

The subtlety here is pace adjustment. A 5-minute bar that is 10 seconds old contains ~3% of a normal bar's trades; naive "current vs average" comparisons on live bars mostly measure the clock. All rate metrics project the live bar to a full bucket by elapsed time before comparing, so the values are meaningful at any point inside the bar. (The first 5% of a bucket is too noisy to extrapolate and is clamped.)

correlation: everything vs BTC

Bitcoin is the market's gravity; these four fields measure how much a pair escapes it. All are computed against BTCUSDT over the same window at your interval.

funding

The perpetual funding state, refreshed every 60 seconds. RATE is the current predicted rate (decimal; 0.0001 = the 0.01% baseline); PREMIUM is the live mark-vs-index basis, which is funding's leading indicator (premium moves first, the rate follows); AVG_24H averages the actually-paid fundings of the last 24h; TREND is the current rate minus the last paid one (funding momentum); APR annualizes at the pair's real payment interval; NEXT is the next payment timestamp. The full mechanics and the paid-payments history live on the Funding page.

trend: the trend engine

The trend engine scores every pair on every timeframe by blending EMA spread, MACD histogram and RSI bias, weighted by ADX strength and squeeze state, plus candle-pattern adjustments, squashed through tanh to a −1…+1 score. Lower timeframes aggregate into a local horizon (hours), upper ones into global (days-to-weeks). local/global are the human-readable labels ("Uptrend", "Sideways", "Downtrend"); scores exposes the raw numbers underneath:

Practical notes

• Fields that cannot be computed yet (young listing, insufficient window) are absent, not zero. Code defensively: data.volsurge?.SCORE.
• Windowed hashes carry a timestamp field (ms): the sweep that produced them. Use it to detect staleness instead of assuming freshness.
• The bulk form is the cheap way to sweep a watchlist: 50 pairs per request. For ranking the whole market by one metric, Rankings is one request instead of eleven.
• Not every family covers every interval (e.g. surge skips 4h; volume/trades use curated windows per timeframe). An absent family at an exotic interval is expected, not a bug.