Data basis: Closing-auction imbalance messages from NYSE Arca for SPY and QQQ (via Databento), only 15:45 to 16:01 New York. Prices: US500 and NAS100 CFDs, Dukascopy minute data. Search period 1 May 2018 to 30 December 2022. The holdout from 2023 stayed untouched because the gate was missed. Costs per round trip 0.8 index points (US500) and 2.5 points (NAS100), on average 2.4 and 2.5 bps (1 bp = 0.01%). Benchmarks: randomly swapped direction (2,000 times), always long on the same days, random markets. Pre-registered on 17 October 2026 before the data was bought (SHA-256 e30f1c15…). No trading recommendation.
Index funds that need the official closing price trade in the closing auction. From 15:50 the exchange publishes continuously which side has more market-on-close volume waiting. Because the direction of this flow is public before the event, the imbalance is widely seen as a hint of where the market goes in the last minutes. We test this on the largest index ETF, SPY, and trade the US500 CFD.
The rule: first message from 15:50:00, entry at the open of the 15:51 minute in the direction of the imbalance, exit at the open of 16:00 (before the auction print) or 16:01 (after it).
1. Rule
Each day the first closing-auction message from 15:50:00 counts. A buy imbalance means long, a sell imbalance short. Size is imbalance shares times reference price in USD, compared with the previous 60 days: all days, from the median, from the 75th percentile. The primary rule was fixed in advance: from the median, exit 16:00, SPY and QQQ averaged per day. Gate to the holdout: net t ≥ 2 and q ≤ 0.10 across 14 tests (q is the error probability adjusted for the number of tests).
2. Result for SPY
| Size | Exit | n | net bps | t | gross bps |
|---|---|---|---|---|---|
| all days | 16:00 | 1,167 | −3.34 | −5.09 | −0.98 |
| all days | 16:01 | 1,167 | −2.90 | −4.49 | −0.53 |
| from median | 16:00 | 564 | −3.86 | −3.61 | −1.51 |
| from median | 16:01 | 564 | −2.94 | −2.86 | −0.59 |
| from 75th pct. | 16:00 | 300 | −4.33 | −2.76 | −1.96 |
| from 75th pct. | 16:01 | 300 | −2.77 | −1.98 | −0.41 |
Left: net per trade with one standard error, gross as a diamond. Right: the pre-registered primary rule against the same rule on two random markets.
Every variant is negative even before costs, and larger imbalances make it worse rather than better. The primary rule (SPY and QQQ averaged, 839 days) lost −3.31 bps per trade at t −4.20. With randomly swapped direction the mean was −2.39 bps, so the real direction was no better than chance (p 0.88). Always long on the same days made +1.03 bps gross. On two random markets without any direction the same rule lost −4.78 and −2.53 bps. After costs, what remains is exactly what a rule without information loses.
3. Our mistake with QQQ
The pre-registration took SPY and QQQ from NYSE Arca. SPY is listed there, QQQ is not: QQQ is a Nasdaq listing, and the Arca closing auction for QQQ is practically empty (median paired volume four shares). The QQQ rows of this test therefore measure no real auction volume. It does not change the result, SPY alone is clearly negative. We then tested QQQ correctly via Nasdaq, also without an edge: QQQ closing auction via Nasdaq.
What it means
The imbalance is public as soon as it appears. Trading in its direction from 15:51 finds no move that survives costs. Whether the market prices it in before the message, we did not measure. The slightly negative gross value points to a small counter-move rather than continuation, but it is too small to make anything of after costs.
Limits
- CFD instead of auction price. The US500 CFD follows the S&P futures, not the ETF price in the auction.
- First message only. We did not use how the imbalance changes until 15:59, nor the paired volume.
- Period. The imbalance data starts in May 2018. Half trading days without a 15:50 message drop out.
- Not tested: single NYSE stocks, trading in the auction itself, continuation beyond 16:01. We tested the single-stock level for the Nasdaq 100: 20 Nasdaq heavyweights.
All pattern families of the scan in the overview. Related: SPY opening auction, index event days in the FTSE 100 and charm, double tops and the 15:50 effect.
📄 Full study as PDF: 5 pages incl. all tables, figures and limits (in German).
Disclaimer: Historical statistics are no guarantee of future market behaviour. This study is not investment advice. Trading carries a risk of loss up to total loss.