Upload README.md with huggingface_hub

README.md

@@ -1,35 +1,116 @@
 ---
-dataset_info:
-  features:
-  - name: timestamp
-    dtype: timestamp[ms]
-  - name: open
-    dtype: float64
-  - name: high
-    dtype: float64
-  - name: low
-    dtype: float64
-  - name: close
-    dtype: float64
-  - name: volume
-    dtype: float64
-  - name: quote_volume
-    dtype: float64
-  - name: num_trades
-    dtype: int64
-  - name: taker_buy_base_volume
-    dtype: float64
-  - name: taker_buy_quote_volume
-    dtype: float64
-  splits:
-  - name: train
-    num_bytes: 203327360
-    num_examples: 2541592
-  download_size: 136104558
-  dataset_size: 203327360
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
+license: mit
+task_categories:
+  - time-series-forecasting
+  - reinforcement-learning
+tags:
+  - finance
+  - cryptocurrency
+  - bitcoin
+  - trading
+  - ohlcv
+  - perpetual-futures
+  - derivatives
+  - binance
+size_categories:
+  - 1M<n<10M
 ---
+
+# BTCUSDT Perpetual Futures 1-Minute OHLCV (1 2021 - Mar 2026)
+
+![Price History](price_history.png)
+
+## Overview
+
+1-minute OHLCV candlestick data for the **BTC/USDT USDT-margined perpetual futures** contract on Binance, covering **May 1, 2021** to **February 28, 2026**.
+
+- **Rows**: 2,541,592
+- **Completeness**: ~100% (only 7 single-bar month-boundary gaps from data packaging)
+
+## Why futures OHLCV?
+
+Perpetual futures are the most actively traded crypto instruments, with volume typically **3-10x higher** than spot. This dataset provides several signals not available in spot data:
+
+- **Futures volume**: Higher liquidity and more representative of actual trading interest
+- **Taker buy/sell ratio**: `taker_buy_base_volume / volume` shows whether aggressive buyers or sellers dominate each candle (>0.5 = buyer-dominated)
+- **Number of trades**: Trade count as a proxy for market activity independent of order size
+- **Quote volume**: Notional value traded, useful for comparing activity across price levels
+
+## Columns
+
+| Column | Type | Description |
+|--------|------|-------------|
+| `timestamp` | `datetime64[ns]` | Candle open time (UTC) |
+| `open` | `float64` | Opening price (USDT) |
+| `high` | `float64` | Highest price in the candle |
+| `low` | `float64` | Lowest price in the candle |
+| `close` | `float64` | Closing price (USDT) |
+| `volume` | `float64` | Trading volume (base asset) |
+| `quote_volume` | `float64` | Trading volume (USDT notional) |
+| `num_trades` | `int64` | Number of trades |
+| `taker_buy_base_volume` | `float64` | Taker buy volume (base asset) |
+| `taker_buy_quote_volume` | `float64` | Taker buy volume (USDT notional) |
+
+## Statistics
+
+| Metric | Value |
+|--------|-------|
+| Start price | $57,550.00 |
+| End price | $66,937.10 |
+| Min price | $15,502.00 |
+| Max price | $126,086.80 |
+| Return | +16.3% |
+| Avg daily volume | 345,284 BTC |
+| Mean taker buy ratio | 0.4971 |
+| Mean trades/candle | 2,600 |
+
+## Data Quality
+
+The only gaps are single missing bars at month boundaries (23:59 -> 00:01, missing the 00:00 candle). This is a Binance data packaging artifact, not actual missing trading data. Completeness is effectively 100%.
+
+## Joining with spot OHLCV
+
+This dataset complements the spot OHLCV dataset [`Torch-Trade/btcusdt_spot_1m_05_2021_to_03_2026`](https://huggingface.co/datasets/Torch-Trade/btcusdt_spot_1m_05_2021_to_03_2026). To join at training time:
+
+```python
+from datasets import load_dataset
+import pandas as pd
+
+# Load both
+spot = load_dataset("Torch-Trade/btcusdt_spot_1m_05_2021_to_03_2026")["train"].to_pandas()
+spot["timestamp"] = pd.to_datetime(spot["timestamp"])
+
+futures = load_dataset("Torch-Trade/btcusdt_perp_1m_05_2021_to_02_2026")["train"].to_pandas()
+futures["timestamp"] = pd.to_datetime(futures["timestamp"])
+
+# Merge — rename futures columns to avoid clash
+futures = futures.rename(columns={
+    "open": "fut_open", "high": "fut_high", "low": "fut_low", "close": "fut_close",
+    "volume": "fut_volume", "quote_volume": "fut_quote_volume",
+    "num_trades": "fut_num_trades",
+    "taker_buy_base_volume": "fut_taker_buy_volume",
+    "taker_buy_quote_volume": "fut_taker_buy_quote_volume",
+})
+df = spot.merge(futures, on="timestamp", how="left")
+
+# Derive taker buy ratio
+df["taker_buy_ratio"] = df["fut_taker_buy_volume"] / df["fut_volume"]
+```
+
+## Usage
+
+```python
+from datasets import load_dataset
+import pandas as pd
+
+ds = load_dataset("Torch-Trade/btcusdt_perp_1m_05_2021_to_02_2026")
+df = ds["train"].to_pandas()
+df["timestamp"] = pd.to_datetime(df["timestamp"])
+
+print(df.shape)  # (2541592, 10)
+print(df.head())
+```
+
+## License
+
+MIT -- data sourced from [Binance Data Collection](https://data.binance.vision/).