Where can I get high-quality Level 2 or tick data for forex to run order book simulations?

[FTtrader]

Hi guys,

Getting high-quality Level 2 (Depth of Market/Order Book) tick data for forex is notoriously difficult compared to crypto or equities.
Because forex is a decentralized, over-the-counter (OTC) market, there is no single "National Best Bid and Offer" (NBBO) or central limit order book.
Liquidity is fragmented across various electronic communication networks (ECNs) and interbank platforms.
If you want to run realistic discrete event simulations for scalping, you need order book data that shows not just the top-of-book (Level 1) prices,
but the volume queuing at multiple price levels. Here is a breakdown of the best sources for L2 forex data, categorized by accessibility and institutional grade.

1. Free & Developer-Friendly Options

If you are just building your simulation framework or learning how to parse L2 data, start here before spending thousands of dollars.Dukascopy: This Swiss broker is famous in the quant community for providing high-quality, free historical tick data. While their standard historical data export is excellent for tick-by-tick Level 1 (top of book) data, getting full L2 depth historically requires pulling from their API or specialized platforms. TrueFX: A project by Integral (a major FX technology provider), TrueFX offers free, clean, institutional-grade top-of-book tick data. While primarily L1, it is excellent for testing your engine's parsing speed and handling millisecond timestamps. Crypto Exchanges (Alternative Sandbox): If your goal is simply to build and test the mechanics of an order book simulator (queue modeling, latency arrays, limit order fills), use crypto data first. Binance, OKX, and Coinbase offer free real-time WebSocket feeds of full L2/L3 order books, and platforms like Tardis.dev offer cheap, pristine historical order book data for crypto. The microstructure dynamics are different from FX, but the engineering challenges of handling the data are identical.

2. Institutional ECNs (The Gold Standard)

If you have a budget and need the most accurate representation of interbank liquidity, you need data sourced directly from the major ECNs. These platforms are where the real volume trades.EBS (CME Group): EBS is one of the two dominant interbank platforms (along with Reuters/Refinitiv). CME Group acquired EBS and now sells its historical market data. Their EBS Spot FX data provides up to 10 levels of order book depth, delivered in 100-millisecond time-slices. This is the institutional standard, but it requires enterprise licensing. Cboe FX (formerly Hotspot FX): Cboe operates a massive institutional FX ECN. They sell historical tick data that reflects bids, offers, and executions from major banks, hedge funds, and high-frequency trading (HFT) firms. Their data is excellent for backtesting institutional strategies because it shows a true aggregated view of diverse market participants. LSEG (London Stock Exchange Group) Tick History: LSEG (which now owns Refinitiv/Reuters, the other major interbank FX platform) offers one of the most comprehensive tick databases in the world. They offer normalized Level 1 and Level 2 data reaching back decades. You can access it via cloud platforms like AWS S3 or Google BigQuery.

3. Data Aggregators and Vendors

These vendors purchase data from multiple venues (like EBS, Cboe, and various banks) and package it for quants and hedge funds.Algoseek: Known for providing incredibly detailed "as-is" data collected directly from Equinix co-location servers (where the matching engines live). They are highly regarded for institutional research. FactSet Tick History: Powered by BMLL Technologies, this service provides normalized Level 2 data and handles the heavy lifting of standardizing formats across different venues. Databento: While famous for disrupting the US Equities and Futures data market with an affordable "pay-as-you-go" API model for L2 and L3 data, their FX coverage is currently limited, though they are rapidly expanding. They are worth watching.What You Must Verify Before BuyingBefore purchasing any historical FX L2 dataset, you must confirm exactly what format you are getting:Snapshot vs. Event-Driven: Are they giving you a snapshot of the order book every 100 milliseconds (like EBS Spot FX), or are they giving you the raw ADD, MODIFY, CANCEL, and TRADE messages? To run a true discrete event simulation with accurate queue position tracking, you need the individual messages (MBO/Level 3). Snapshots force you to guess what happened between intervals. Timestamp Granularity: Are the timestamps in milliseconds, microseconds, or nanoseconds? For scalping simulations, you need microsecond precision at a minimum.Aggregation: Because FX is OTC, is the vendor showing you the book from one specific ECN, or a consolidated book from multiple feeds? A consolidated book might look deep, but you might not actually be able to execute against all of it from a single broker connection.

Or do you have any other secret way how to get properly data for your back testing?
Ideally tick by tick?

Thanks for reading,
have a nice day.

[LondonScalper]

Getting high-quality Level 2 (Depth of Market/Order Book) tick data for forex is notoriously difficult compared to crypto or equities.

Agreed — and the hard bit is usually not “finding a vendor,” it’s deciding what the book is supposed to represent.

FX L2 is venue-specific. A Dukas-style historical book, a broker DOM, and an EBS/Reuters feed are not interchangeable universes. If your discrete-event sim assumes one continuous LOB the way equities do, you’ll get beautiful fills that never exist on the retail path you actually trade.

Desk-practical split I use:

• Research / microstructure toys: paid tick + multi-level where available (TrueFX-style L1 is fine for many scalp questions; L2 only if you’re studying queue effects)
• Execution realism: your own broker’s tick + reject/slip logs beat a perfect foreign book
• Never mix venues in one sim without tagging the source

For pure M1 discretionary work I rarely need full depth — spread regime, spike behaviour, and session liquidity dominate. Order-book sims pay off when you’re testing limit-queue or spoof-sensitive logic.

Curious whether you’re simulating maker/taker behaviour or just trying to stress-test stop placement against adverse selection?

[PropScalpDesk]

Forex L2 is messy — define the research question

True shared level-two like equities is not what most retail FX feeds are. Before buying tick or depth data, I ask what decision it will change: execution-model validation, markout studies, or just prettier charts that do not alter skips and flats.

Practical path I have seen work:

• Broker-exported tick and deal history for your venue — limited, but relevant to the book you actually trade.
• Paid historical tick from reputable vendors for research, knowing it is not your broker’s book.
• Be sceptical of anyone selling “the” FX order book as universal truth across venues.

For day-to-day scalping from Frankfurt, my own deal log beat fantasy depth more often than the reverse. Data without a question is just storage cost.

Start with a single hypothesis written in one sentence. Buy data only if that sentence requires it. Otherwise you will drown in ticks and still not know when to skip a Frankfurt open print.

What are you trying to build — a filter, EA logic, or a microstructure study that must generalize beyond one broker?