Datasetpublic

tripolskypetr/btcusdt_feb_2026_candles_news_forecast

BTCUSDT Multi-Timeframe GARCH Volatility Dataset Symbol: BTCUSDTPeriod: February 2026 (backtest simulation)Records: ~40 320 (1-minute resolution)Source tool: backtest-kit + garch Dataset Description Each row is a 1-minute snapshot of GARCH-predicted volatility (sigma) for BTCUSDT across 8 timeframes, computed during a backtest run. The reliable flag indicates whether the model had enough historical candles to produce a statistically stable estimate.… See the full description on the dataset page: https://huggingface.co/datasets/tripolskypetr/btcusdt_feb_2026_candles_news_forecast.

sourceHugging Facemitupdated 5mo agoView on Hugging Face

0likes28downloads

Dataset Card

BTCUSDT Multi-Timeframe GARCH Volatility Dataset

Symbol: BTCUSDT Period: February 2026 (backtest simulation) Records: ~40 320 (1-minute resolution) Source tool: backtest-kit + garch

Dataset Description

Each row is a 1-minute snapshot of GARCH-predicted volatility (sigma) for BTCUSDT across 8 timeframes, computed during a backtest run. The reliable flag indicates whether the model had enough historical candles to produce a statistically stable estimate.

Data Format

The dataset is stored as JSONL (log.jsonl). Each line is a JSON object with the following structure:

json

{
  "id": "uuid-v4",
  "type": "info",
  "timestamp": 1769904000000,
  "createdAt": "2026-02-01T00:00:00.000Z",
  "methodContext": {
    "exchangeName": "ccxt-exchange",
    "strategyName": "feb_2026_strategy",
    "frameName": "feb_2026_frame"
  },
  "executionContext": {
    "when": "2026-02-01T00:00:00.000Z",
    "symbol": "BTCUSDT",
    "backtest": true
  },
  "topic": "position idle",
  "args": [{
    "symbol": "BTCUSDT",
    "volatility_1m":  { "sigma_1m":  0.00097, "reliable_1m":  false },
    "volatility_5m":  { "sigma_5m":  0.00198, "reliable_5m":  true  },
    "volatility_15m": { "sigma_15m": 0.00403, "reliable_15m": true  },
    "volatility_30m": { "sigma_30m": 0.00541, "reliable_30m": true  },
    "volatility_1h":  { "sigma_1h":  0.00716, "reliable_1h":  true  },
    "volatility_4h":  { "sigma_4h":  0.01698, "reliable_4h":  true  },
    "volatility_6h":  { "sigma_6h":  0.01620, "reliable_6h":  true  },
    "volatility_8h":  { "sigma_8h":  0.04188, "reliable_8h":  true  }
  }]
}

Fields

Field	Type	Description
`timestamp`	integer	Unix timestamp in milliseconds
`createdAt`	ISO 8601 string	UTC datetime of the candle
`executionContext.symbol`	string	Trading pair (`BTCUSDT`)
`executionContext.backtest`	boolean	Always `true` — backtested data
`sigma_Xt`	float	GARCH-predicted 1-step-ahead volatility for timeframe `Xt`
`reliable_Xt`	boolean	Whether the GARCH model had sufficient history to converge

Timeframes

Timeframe	Candles used for fit	Typical sigma range
1m	1 500	~0.001 (often unreliable early)
5m	1 500	~0.002
15m	1 000	~0.004
30m	1 000	~0.005
1h	500	~0.007
4h	500	~0.017
6h	300	~0.016
8h	300	~0.042

Generation Code

typescript

import { listenIdlePing, getCandles, Log } from "backtest-kit";
import { predict } from "garch";

listenIdlePing(async ({ symbol }) => {
  const candles_1m  = await getCandles(symbol, "1m",  1_500);
  const candles_5m  = await getCandles(symbol, "5m",  1_500);
  const candles_15m = await getCandles(symbol, "15m", 1_000);
  const candles_30m = await getCandles(symbol, "30m", 1_000);
  const candles_1h  = await getCandles(symbol, "1h",  500);
  const candles_4h  = await getCandles(symbol, "4h",  500);
  const candles_6h  = await getCandles(symbol, "6h",  300);
  const candles_8h  = await getCandles(symbol, "8h",  300);

  const { sigma: sigma_1m,  reliable: reliable_1m  } = await predict(candles_1m,  "1m");
  const { sigma: sigma_5m,  reliable: reliable_5m  } = await predict(candles_5m,  "5m");
  const { sigma: sigma_15m, reliable: reliable_15m } = await predict(candles_15m, "15m");
  const { sigma: sigma_30m, reliable: reliable_30m } = await predict(candles_30m, "30m");
  const { sigma: sigma_1h,  reliable: reliable_1h  } = await predict(candles_1h,  "1h");
  const { sigma: sigma_4h,  reliable: reliable_4h  } = await predict(candles_4h,  "4h");
  const { sigma: sigma_6h,  reliable: reliable_6h  } = await predict(candles_6h,  "6h");
  const { sigma: sigma_8h,  reliable: reliable_8h  } = await predict(candles_8h,  "8h");

  Log.info("position idle", {
    symbol,
    volatility_1m:  { sigma_1m,  reliable_1m  },
    volatility_5m:  { sigma_5m,  reliable_5m  },
    volatility_15m: { sigma_15m, reliable_15m },
    volatility_30m: { sigma_30m, reliable_30m },
    volatility_1h:  { sigma_1h,  reliable_1h  },
    volatility_4h:  { sigma_4h,  reliable_4h  },
    volatility_6h:  { sigma_6h,  reliable_6h  },
    volatility_8h:  { sigma_8h,  reliable_8h  },
  });
});

Use Cases

—Training volatility prediction models across multiple timeframes
—Studying cross-timeframe GARCH sigma correlations on crypto data
—Benchmarking volatility estimators against GARCH baselines
—Feature engineering for crypto trading ML pipelines

License

MIT