Christ1234/queyn-eurusd-regimes

Synthetic EUR/USD 1-Minute Sessions with Regime Labels (Queyn)

100 synthetic EUR/USD trading sessions at 1-minute resolution, generated by the Queyn agent-based market simulator. Each session covers the London–New York overlap (243 candles, 13:00–17:02 UTC). Every candle carries a ground-truth market regime label — something real market data cannot provide.

All data is synthetic. It is not historical market data and must not be treated as such.

Why this dataset

• Ground-truth regimes. Each candle is labeled with the regime that generated it (ranging, consolidation, mild trend, breakout, reversal, volatility spike) and its direction, plus the exact simulation parameters behind it.
• Microstructure included. Bid, ask and spread come from simulated market makers, so spreads widen and narrow with market conditions.
• Statistically validated. Sessions reproduce the stylized facts of high-frequency EUR/USD returns: fat tails, volatility clustering, and weak autocorrelation. The methodology is published as a technical report: doi.org/10.5281/zenodo.21451431.

Files

candles.csv — 24,300 rows, one per 1-minute candle

Column	Description
session_id	Session identifier (S001–S100)
session_type	ny_london (London–New York overlap simulation)
minute	Minute index within the session (0–242)
time_utc	Clock time of the candle (session starts 13:00 UTC). No calendar date: sessions are synthetic
open, high, low, close	Mid price OHLC
tick_count	Number of simulated ticks in the candle
bid_close, ask_close	Best bid / ask at candle close
spread_avg_pips, spread_max_pips	Average and maximum bid–ask spread during the candle, in pips
regime	Ground-truth regime: ranging, consolidation, mild_trend, breakout, reversal, volatility_spike
regime_direction	bullish, bearish or neutral (as scheduled; inside a reversal the trend flips partway, see the weight columns)
trend_up_weight, trend_down_weight	Share of directional order flow driving the price up / down (0–1)
volatility_level	Volatility intensity parameter (0–1)
liquidity_stress	Liquidity stress parameter (0–1); higher values mean wider spreads and more erratic moves

sessions.csv — 100 rows, one per session

Per-session summary statistics: lag-1 return autocorrelation (ac1), autocorrelation of absolute returns (ac1_abs), standard deviation of 1-minute returns in pips, kurtosis, skewness, Jarque–Bera statistic, high–low range, net move, average spread, number of regime switches, and an OHLC integrity check (bad_ohlc, always 0).

validation_summary.csv

Averages across all 100 sessions compared with reference values from real EUR/USD London–New York sessions.

Metric	Queyn (mean of 100)	Real EUR/USD reference
Lag-1 autocorrelation	0.041	0.063
Std of 1-min returns (pips)	1.55	1.45
Kurtosis	4.45	4.75
Autocorrelation of |returns|	0.139	0.105
Session range (pips)	32.3	31.9

Quick start

from datasets import load_dataset
candles = load_dataset("Christ1234/queyn-eurusd-regimes", "candles", split="train").to_pandas()
sessions = load_dataset("Christ1234/queyn-eurusd-regimes", "sessions", split="train").to_pandas()

Example notebook (stylized facts and a regime classifier): Kaggle notebook

Suggested uses

• Regime classification: predict regime from price, volume and spread features.
• Strategy robustness testing: check how a strategy behaves across known regimes.
• Change-point detection: regime switches are exactly known.
• Reinforcement learning execution environments with realistic spreads.
• Teaching: stylized facts and market microstructure.

Limitations

• Synthetic data reproduces statistical properties of real markets. It does not reproduce real events, news, or historical price levels.
• Do not use it alone to train models for directional price prediction on real markets. Always validate on real held-out data.
• Only EUR/USD during the London–New York overlap is covered.
• Full methodology, results and limitations: doi.org/10.5281/zenodo.21451431.

Need more?

Generate your own scenarios — custom trends, volatility, market stress and black-swan events — and export them in CSV or JSON at queyn.com.

Also available

• This dataset on Kaggle: https://www.kaggle.com/datasets/christinadav/synthetic-eurusd-1-m-sessions-with-regime-labels
• Companion dataset with 214 labeled market shocks: Synthetic EUR/USD Black Swan Events

Citation

Davidopoulou, C., & Ouranidis, A. (2026). Statistical Validation of a Hybrid Agent-Based Synthetic Market Data Generator for EUR/USD. Zenodo. https://doi.org/10.5281/zenodo.21451431

Queyn is a simulation platform for education and research. Nothing here is financial advice.