These two get compared constantly and they are barely comparable. Dukascopy is a serious historical data archive with tick-level depth going back two decades. Yahoo Finance is a quotes service that happens to expose an API. For daily bars both work; for backtesting anything intraday, only one is a candidate.
| Dukascopy | Yahoo Finance | |
|---|---|---|
| Cost | Free | Free |
| Account / key | None | None |
| Tick data | Yes, with bid and ask | No |
| Intraday depth | Back to ~2003 | Short recent window only |
| Daily bars | Yes | Yes — reliable |
| Spread information | Yes | No |
| Ease of access | Difficult | Very easy |
| Speed | Slow | Fast |
| Reliability per request | Intermittent | Good |
| Suitable for intraday backtesting | Yes | No |
If you are backtesting an intraday forex strategy, the table ends the discussion at the fourth row. Yahoo does not retain deep intraday history, so there is nothing to backtest against. Use it for daily bars, quick quotes and prototyping; use Dukascopy for anything else.
Free, no account, no API key, no request cap, tick-level with bid and ask, back to roughly 2003. On specification alone it beats most paid sources, which is why so many free forex archives — including ours — are built from it.
Getting it out is the problem, and these are hard-won specifics rather than general grumbling:
Rather than describe the reliability problem abstractly, here is a measurement of it. Our own archive is built from Dukascopy, so auditing it audits Dukascopy.
Checked on 17 August 2026 across 1,911,141 EURUSD one-minute bars spanning 3 January 2021 to 30 April 2026, these coverage holes remain:
| Last bar | Next bar | Days missing | Explanation |
|---|---|---|---|
| 2023-12-29 | 2024-01-01 | 3 | New Year — legitimate |
| 2025-06-25 | 2025-06-30 | 4 | Missing history |
| 2025-07-02 | 2025-07-07 | 4 | Missing history |
| 2025-07-22 | 2025-07-28 | 5 | Missing history |
| 2025-09-24 | 2025-09-28 | 4 | Missing history |
| 2025-11-20 | 2025-11-24 | 3 | Missing history |
| 2026-03-06 | 2026-03-12 | 5 | Missing history |
Seven holes in five years, six of them genuine data loss from the empty-response behaviour above. That is the realistic reliability picture: very good data, with occasional silent holes you must actively check for.
Also worth knowing, and not fixable by any amount of retrying: Dukascopy genuinely has no index data for 10–18 March 2021, so index instruments carry an eight-day hole that is upstream of everyone.
Yahoo Finance deserves its popularity. Through the widely used yfinance library it is a couple of lines of Python, it is fast, it needs no key, and for daily bars across equities, indices and currency pairs it is dependable.
Its limits for forex backtesting are structural rather than incidental:
Use it for what it is good at: daily-bar research, correlation studies, quick prototypes, and anything where a few lines of code matter more than resolution. For reference, a daily-return correlation between EURUSD and GBPUSD over 1,536 overlapping trading days comes out at r = 0.77 with the pairs moving in the same direction 77.72% of the time — exactly the kind of study daily bars are perfect for.
| You want… | Use |
|---|---|
| Years of intraday forex history | Dukascopy (or a pre-built archive) |
| Tick data with real spreads | Dukascopy |
| Daily bars for a quick study | Yahoo Finance |
| A few lines of Python and results now | Yahoo Finance |
| Correlation or regime research | Yahoo Finance |
| To skip the pipeline entirely | FXAbsolute |
That last row is worth stating plainly since it is our own tool: if you want Dukascopy-quality one-minute history without building and debugging a download pipeline, it is already assembled. 28,412,683 one-minute bars across 15 forex, index and crypto instruments from 2021 to 2026, free, no account, no key, no cap — and the seven holes above are published rather than hidden, so you know exactly what you are getting.
The honest limitation: it is mid-price OHLC, so unlike raw Dukascopy it carries no bid/ask spread. If spread modelling is what you came for, go to the source.
Dukascopy, for anything involving backtesting. It offers free tick-level data with bid and ask going back to roughly 2003, with no account or key. Yahoo Finance retains only a short window of intraday history and provides no tick or spread data, so it cannot support intraday forex backtesting. Yahoo is better for daily bars and quick studies.
Only for a short recent window. Yahoo does not retain deep intraday history, so you cannot pull years of one-minute forex bars from it. For deep intraday history use Dukascopy, HistData, or a pre-assembled archive such as FXAbsolute's 28,412,683 one-minute bars covering 2021 to 2026.
The data quality is excellent, but the delivery is intermittent. It sometimes returns an empty array for a valid time window, which — if cached by a download pipeline — permanently creates multi-day holes. An audit of 1,911,141 EURUSD one-minute bars derived from Dukascopy found 7 coverage holes of 3 to 5 days across five years, six of them genuine data loss.
Raise the connection timeout well above the default, request roughly four-day chunks rather than month-sized ranges, and always retry windows that return empty rather than caching the empty result as truth. Afterwards, verify by counting bars per trading day and looking for unexplained multi-day gaps.
No. Yahoo Finance provides OHLC bars without bid/ask or spread information, so execution cost is invisible in any backtest built on it. Dukascopy provides tick data including both bid and ask, which is why it is the standard free source for spread-aware forex research.
Using a pre-assembled archive avoids building a download pipeline entirely. FXAbsolute provides 28,412,683 one-minute bars across 15 instruments from 2021 to 2026 free with no account or key, with its coverage gaps published. For bulk files, HistData offers M1 data back to 2000 as monthly ZIP downloads.
Five years of one-minute history across 15 instruments, built from Dukascopy with every coverage hole published — free, no key, no cap.
Open FXAbsolute →