How gold reacts to CPI: 50 releases, measured.

What the data shows

1. Almost all of the move comes in the first 15 minutes. The first candle after 08:30 New York time was a median 3.5 times the size of the same candle on ordinary days. The whole New York session was only 1.35 times its normal range, so CPI concentrates volatility into one candle rather than creating a wild day.
2. The first direction tends to stick. It was still in place after an hour in 39 of 50 releases and at the New York close in 37 of 50. The average move gives back a little after the first candle, from +0.47% to +0.35% by the close, but it doesn't reverse.
3. Big first moves are more reliable than small ones. When the first candle moved more than the median (0.34%), the direction held to the close in 21 of 25 cases. Smaller first moves held in only 16 of 25; a small first candle is often just noise before the real move.
4. One day in four is a whipsaw. On 13 of 50 release days, price later broke both the high and the low of the first candle. Stops placed just beyond the first candle are the ones that get hit.
5. There is no edge in guessing early. In the 90 minutes before the release, gold's range was 0.96 times normal. The market waits for the number, and so should you.
6. The dollar moves are growing, the relative moves aren't. With gold at more than twice its 2022 price, a typical first candle is now worth far more in dollars, but compared with an ordinary day it has shrunk from about 6× in 2022 to 2.0× in 2025. Size your trades from the dollar range, not from habit.

What this means for trading CPI

Protect or close open trades before 08:30 New York time, don't place market orders at the release, and let the first 15-minute candle close. If it is a large candle, the data says its direction is more likely than not to hold, so a pullback toward it is the higher-probability trade. If it is small, expect the real move to come later and keep stops outside the first candle's range. Gold's reaction is the same idea we use in our displacement and fair value gap entries.

By year

All 50 releases

Download the data (CSV). Free to use and republish with a link to this page.

How we measured it

Releases: all 50 US CPI releases published by the Bureau of Labor Statistics between 13 July 2022 and 11 September 2026, each at 08:30 New York time. The October 2025 report was never published because of the government shutdown, and the September 2025 report came out late, on 24 October.

Prices: XAUUSD 15-minute candles from a retail broker feed (about 100,000 candles), converted from the broker's server clock to New York time with daylight saving handled for both Europe and the US. The release price is the open of the 08:30 candle; "first 15 minutes" is that candle's close; "after 1 hour" is the close of the 09:15 candle; "NY close" is the last candle before 17:00 New York time.

Normal day: each release is compared with the median of the 60 previous weekdays without a CPI release, measured over the same time window, in % of price so that changes in gold's price level don't distort the comparison.

Limits: 15-minute candles hide the spikes inside them, and other brokers' prices differ slightly. The period includes a strong gold bull market, which is why more first moves were up than down. We didn't sort releases by whether CPI beat or missed forecasts; this study is about how gold moves, not which way. Past reactions don't guarantee future ones.

Frequently asked questions

Educational research, not financial advice.