About Our Forex Sentiment Data
Where every number on this site comes from, how often it updates, and what it will not tell you.
Sentiment: how it is collected
Collect
Retail positioning for 29 instruments, read every 15 minutes from aggregated retail broker positioning.
Two paths
A browser scrape runs first. If it is blocked, the same run falls back to the official API, so a block costs seconds rather than a gap.
Store
Raw readings are kept unmodified, then rolled into M15, M30, H1, H4 and D1 candles for the charts.
Publish
Straight to the site, the API and the embeddable widgets. No smoothing, no editing.
Coverage
- Instruments: 28 currency pairs plus gold (XAU/USD)
- Frequency: every 15 minutes
- History: continuous since January 2026, nothing deleted or revised
- Timezone: all candles and filters use Asia/Jerusalem; weekends are excluded because the FX market is shut
The other data on this site
Sentiment is one of four datasets here. Each comes from a different place and updates at a different pace.
- Central bank policy rates — current levels hourly from TradingEconomics; the change history from the Bank for International Settlements. Two sources, so where they disagree it is visible. See central bank rates.
- Institutional positioning — the CFTC's weekly Commitment of Traders report, taken from the regulator directly, with three years of history. See COT report.
- Correlation — computed here from our own sentiment series, not imported.
- Price — used internally to check whether the signals on this site actually work. It is not republished, because the licence it comes under does not allow that.
What sentiment does and does not tell you
Retail sentiment is the share of retail traders holding long versus short positions in a pair. It is a genuine measurement of where a particular crowd is positioned.
It is widely claimed that this works as a contrarian indicator — that when 80 % of retail is long, price tends to fall. We used to repeat that claim. Then we tested it against our own data across twelve combinations of threshold and holding period, and every one of them lost money. We tested the signals we publish too, in both directions, and found no tradable edge either way once costs are counted.
So we no longer tell you that extreme sentiment predicts a reversal. What it does tell you is real and useful on its own terms:
- Where the crowd is concentrated, and therefore where stop orders are likely to be clustered.
- How positioning is changing — a crowd moving from 40 % to 80 % long over a week is a different situation from one that has sat at 80 % for a month.
- Context for other analysis, particularly next to the interest rate differential and what institutional money is doing.
Known limitations
- One broker's clients. This is a large sample of retail traders, not the whole retail market, and certainly not the whole market.
- Percentages, not size. The headline figure counts traders, not money. One large position and a hundred small ones on the other side read as 99 % one-way.
- Seven months of history. Enough to see behaviour, not enough to draw conclusions across market regimes.
- Gaps happen. When collection fails, the series has a hole rather than an interpolated guess. We would rather show a gap than invent a number.
Corrections
When we get something wrong, we say so on the site rather than quietly editing it. The contrarian claim above is the clearest example: it was stated as fact here for months before we tested it.
More detail in the FAQ, or see what the site does.