---
license: cc-by-4.0
tags:
  - forex
  - finance
  - ohlcv
  - historical-data
  - dukascopy
---

# Forex Historical Data

Historical OHLCV+Volume data for 30 forex instruments sourced from DukasCopy tick data via [forexsb.com](https://forexsb.com/historical-forex-data).

## Data Coverage

- **30 instruments** across majors, crosses, and commodities
- **7 timeframes**: M1, M5, M15, M30, H1, H4, D1
- **210 Parquet files** total
- **Data source**: DukasCopy real tick data compiled into bar data
- **Timezone**: UTC

## Instruments (30 total)

### Major Pairs
EURUSD, GBPUSD, USDJPY, USDCHF, AUDUSD, NZDUSD, USDCAD

### Cross Pairs
EURJPY, EURGBP, EURAUD, EURNZD, EURCHF, EURCAD,
GBPJPY, GBPAUD, GBPNZD, GBPCHF, GBPCAD,
AUDJPY, AUDNZD, AUDCHF, AUDCAD,
NZDJPY, CHFJPY, CADJPY, CADCHF,
NZDCAD, NZDCHF

### Commodities
XAUUSD (Gold), XAGUSD (Silver)

## File Structure

```
{symbol}/{SYMBOL}_{TIMEFRAME}.parquet
```

Example: `EURUSD/EURUSD_M1.parquet`

## Columns

| Column | Type | Description |
|--------|------|-------------|
| Time | timestamp[ns, tz=UTC] | Bar open time |
| Open | float64 | Open price |
| High | float64 | High price |
| Low | float64 | Low price |
| Close | float64 | Close price |
| Volume | int64 | Tick volume |
| timestamp_ms | int64 | Unix timestamp in milliseconds |

## Timeframe Ranges

- **M1**: ~200,000 bars (~6 months)
- **M5**: ~200,000 bars (~2.5 years)
- **M15**: ~200,000 bars (~7 years)
- **M30**: ~200,000 bars (~14 years)
- **H1**: ~100,000 bars (derived from M30)
- **H4**: ~25,000 bars (derived from M30)
- **D1**: ~5,000 bars (derived from M30)

## Usage

```python
from datasets import load_dataset

# Load a specific symbol/timeframe
dataset = load_dataset("Yllvar/fx-historical-data", data_files="EURUSD/EURUSD_M1.parquet")

# Load all EURUSD data
dataset = load_dataset("Yllvar/fx-historical-data", data_files="EURUSD/*.parquet")
```

## Source

Data collected from DukasCopy tick data and compiled by [Forex Software Ltd](https://forexsb.com).