FamilyLinks/btc-price-1m-2017-2025

📊 BTC-1m-Micro-Macro: Multimodal Financial Dataset

A research-grade dataset bridging High-Frequency Market Microstructure with Fundamental On-Chain Analytics.

---

📑 Abstract

This dataset provides a granular view of the Bitcoin market by combining 1-minute OHLCV price action (sourced from Binance) with Macro On-Chain indicators (sourced from CryptoQuant).

Unlike standard datasets that focus solely on price history, this repository includes Order Flow dynamics (Delta, CVD) and deep network analytics (Miner flows, Whale activity, Institutional sentiment). It is designed for training Machine Learning models (LSTM, Transformers, XGBoost) to detect market regimes, predict volatility, and simulate high-frequency trading strategies.

• Resolution: 1 Minute.
• Coverage: August 2017 — December 2025.
• Format: Parquet (Snappy Compression).
• Alignment: Raw Time-Alignment (See Methodology).

---

🏗️ Data Structure & Feature Dictionary

The dataset consists of 33 features categorized into three layers: Microstructure, Technicals, and Macro On-Chain.

1. Market Microstructure (Source: Binance Spot)

Reflects the immediate supply/demand dynamics and order book aggression.

Feature Name	Type	Description
date	Timestamp	UTC Timestamp (1-minute intervals).
open, high, low, close	Float	Standard OHLC price data in USDT.
volume_btc	Float	Total trading volume in BTC.
volume_usdt	Float	Total trading volume in USD (Quote Asset Volume). Useful for normalizing volume across years.
trade_count	Int	Number of distinct trades executed. High count + low price change = Bot Wars / Accumulation.
taker_buy_vol_btc	Float	Volume bought via Market Orders (Aggressive Buyers).
taker_sell_vol_btc	Float	Volume sold via Market Orders (Aggressive Sellers). Calculated as Total Vol - Taker Buy.
net_flow_delta	Float	Net Order Flow. (Taker Buy - Taker Sell). Positive values indicate aggressive buying pressure.
cvd_1d	Float	Intraday Cumulative Volume Delta. The running sum of Delta, reset daily at 00:00 UTC. Shows who controls the specific day.

2. Technical Indicators

Pre-calculated features for convenience.

Feature Name	Type	Description
rsi_14	Float	Relative Strength Index. Momentum oscillator measuring the speed and change of price movements.
atr_14	Float	Average True Range. Measures market volatility in absolute USD terms.
volatility_pct	Float	Normalized Volatility. Calculated as ATR / Close. Useful for comparing volatility across different price epochs (e.g., 2017 vs 2024).

3. Macro On-Chain Metrics (Source: CryptoQuant)

Fundamental indicators describing the behavior of Whales, Miners, and Institutions. Note: These metrics are mapped to the 1-minute timeframe via forward-fill.

🏦 Institutional & Market Sentiment

Feature Name	Description	Interpretation
cq_coinbase_premium_index	The % difference between Coinbase Pro (USD) and Binance (USDT) prices.	High: US Institutions buying. Low/Negative: US Institutions selling (Strong Bearish Signal).
cq_funding_rates	Periodic payments to traders that are Long/Short.	High: Market is overheated (Longs pay Shorts). Risk of Long Squeeze.
cq_open_interest	Total number of outstanding derivative contracts.	Rising OI: More leverage/fuel entering the market.
cq_cme_futures_..._stacked	Open Interest specifically on the CME exchange.	Proxy for traditional institutional interest (Wall Street).
cq_esl	Estimated Leverage Ratio. Open Interest / Exchange Reserves.	Indicates how much leverage is being used relative to actual collateral.

🐋 Whale & Exchange Activity (Inflows)

Feature Name	Description	Interpretation
cq_exchange_inflow_total	Total amount of BTC transferred to exchanges.	High inflow usually indicates potential selling pressure.
cq_exchange_inflow_mean	Average size of inflow transactions.	High Mean: Whales are depositing. Low Mean: Retail is depositing.
cq_exchange_inflow_top10	Volume of the top 10 largest inflow transactions.	Specific indicator for "Mega-Whale" deposits.
cq_exchange_inflow_cdd	Coin Days Destroyed (CDD). Weight of inflow based on coin age.	High Spike: "Old Hands" (Long-term holders) are moving coins to sell. Major bearish signal.
cq_exchange_inflow_..._value_bands	Inflow volume categorized by transaction size (e.g., 10-100 BTC).	Shows specifically which size cohort is selling.
cq_exchange_inflow_..._age_bandspct	% of inflow coming from different age groups (e.g., 6m-12m).	If older age bands spike, it confirms long-term holder capitulation.
cq_exchange_reserve_usd	Total value of BTC held in exchange wallets.	Decreasing: Supply shock (Bullish). Increasing: Potential sell-side liquidity (Bearish).

⛏️ Miner Activity

Feature Name	Description	Interpretation
cq_miners_position_index	Ratio of miner outflows to their 1-year moving average.	>2.0: Miners are selling aggressively.
cq_miner_to_exchange_flow	Direct flow from Miner wallets to Exchanges.	Immediate selling pressure from miners covering costs.

💥 Liquidation Data

Feature Name	Description	Interpretation
cq_long_liquid	Value of Long positions forcefully closed.	Indicates "Long Squeeze" / Bottoming events.
cq_short_liquid	Value of Short positions forcefully closed.	Indicates "Short Squeeze" / Topping events.

---

⚠️ Important Methodology Note: Time Alignment

This dataset uses RAW Time Alignment.

• Alignment: The On-Chain metric for 2024-01-01 is aligned with the price candles of 2024-01-01.
• Purpose: This allows for precise correlation analysis (e.g., "Did the price drop exactly when Miner Inflow spiked?").
• WARNING for ML Engineers: If you are building a predictive trading model, YOU MUST APPLY A LAG (Shift) to the cq_ columns.
  • In a live trading environment, daily on-chain metrics are often confirmed only after the day closes (UTC 00:00).
  • Recommendation: Shift macro features by +1440 minutes (1 day) during preprocessing to avoid Look-Ahead Bias.

---

🚀 Usage Example (Python)

How to load the data and prepare it for a Backtest or ML Model:

import pandas as pd

# 1. Load the Parquet file (Efficient & Fast)
# Assuming you downloaded the file or used the HF library
df = pd.read_parquet("BTC_Raw_Micro_Macro_1m.parquet")

# 2. Prepare for ML (Apply Lag to Macro Metrics)
# We shift CryptoQuant metrics by 1 day (1440 minutes) to simulate 
# realistic data availability and prevent data leakage.
macro_cols = [c for c in df.columns if c.startswith('cq_')]
df[macro_cols] = df[macro_cols].shift(1440)

# 3. Filter Data (e.g., Remove NaN from the beginning)
df.dropna(inplace=True)

# 4. Example Signal:
# "Whales are moving old coins (High CDD) AND Retail is panic selling (Negative Delta)"
signal = (
    (df['cq_exchange_inflow_cdd'] > 1_000_000) & 
    (df['net_flow_delta'] < -100)
)

print(f"High Risk Minutes Identified: {signal.sum()}")