Data basis: Imbalance messages of the NYSE Arca core opening auction for SPY (via Databento), only 09:20 to 09:31 New York. Prices: US500 CFD, 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 during the cash session (on average 2.3 bps, 1 bp = 0.01%), 2.5 times that before the open. Benchmarks: randomly swapped direction (2,000 times), always long on the same days, random markets. Pre-registered on 19 October 2026 before the data was bought (SHA-256 57177605…), plus one addendum before any evaluation (SHA-256 14450c37…). No trading recommendation.
The open has an auction too. Anyone who must buy or sell at the official opening price sends market-on-open orders, and the exchange publishes in advance which side has more waiting. The question is the same as for the closing auction: does the imbalance predict the move around the cash open?
Two rules in the direction of the imbalance. Rule A tries to catch the auction print, rule B bets on follow-through in the first half hour.
1. Rule
Each day the last message of the core opening auction until 09:29:00 counts. A buy imbalance means long, a sell imbalance short. Size is imbalance shares times reference price, compared with the previous 60 days (all days, from the median, from the 75th percentile). Rule A: entry 09:29, exit 09:31, with pre-open costs. Rule B: entry 09:31, exit 10:00. Primary rule: B from the median. Gate to the holdout: net t ≥ 2 and q ≤ 0.10 across six tests.
One addendum before the first evaluation: Arca runs no opening auction for QQQ, which is listed on Nasdaq. That was a mistake in our pre-registration. The primary rule therefore applies to SPY alone. We then tested QQQ via Nasdaq: QQQ opening auction.
2. Result
| Rule | Size | n | net bps | t | gross bps |
|---|---|---|---|---|---|
| A (print) | all days | 1,166 | −5.36 | −18.89 | +0.54 |
| A (print) | from median | 543 | −4.94 | −10.34 | +0.87 |
| A (print) | from 75th pct. | 286 | −4.03 | −5.68 | +1.83 |
| B (follow-through) | all days | 1,175 | −1.11 | −1.12 | +1.25 |
| B (follow-through) | from median | 550 | +0.41 | +0.25 | +2.74 |
| B (follow-through) | from 75th pct. | 292 | +2.53 | +0.98 | +4.89 |
Gross (grey) and net (orange) per trade, net with one standard error. Rule A mainly loses the pre-open costs. For rule B gross rises with size, while every variant stays below t 1 net.
The primary rule made +0.41 bps at t 0.25. Some directional information is there: against randomly swapped directions (mean −2.25 bps) it sits right at the threshold (p 0.052), and always long on the same days made only +0.91 bps gross. On two random markets the same rule lost −3.10 and −4.77 bps. It is not enough to trade: costs of 2.3 bps eat almost the entire gross advantage.
3. The observation we followed up
For rule B the gross result grows with the size of the imbalance: +1.25, +2.74 and +4.89 bps. That looks like a real mechanism, but it was not the pre-registered primary rule and with 292 trades and net t 0.98 it is not reliable. We therefore tested it on a second market, the QQQ opening auction via Nasdaq. There the sign flips: large imbalances lose −4.78 bps gross. The size effect does not replicate.
What it means
The opening-auction imbalance may carry a little information about the first half hour, but clearly less than the cost of a CFD trade. Trading the print itself with CFDs is hopeless, because the spread is wider before 09:30.
Limits
- CFD instead of auction. The CFD follows the futures, which trade before 09:30. Part of the imbalance may be priced in earlier.
- One market. Only SPY, because Arca runs no auction for QQQ.
- Cost model. Fixed points per round trip, a flat 2.5 times that before the open.
- Not tested: single stocks, messages earlier than the last one before 09:29, holding beyond 10:00.
All pattern families of the scan in the overview. Related: SPY closing auction, QQQ opening auction via Nasdaq and a strong pre-open hour means a bigger open.
📄 Full study as PDF: 6 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.