After assumed trading costs
RESEARCH NOTE / SEPTEMBER 2026
Does the idea
hold up?
We test whether market activity can help predict the next move in Nasdaq futures. Here is what the evidence says so far.
Still in research · No approved trading model
Audited history: January 2024–September 2026. All returns on this page are simulations, not money earned.
01 / THE CURRENT EVIDENCE
A small gain. An uneven journey.
The overall result is positive, but it depends heavily on one good year. That is not enough to call the approach reliable.
The difficult part of the journey
A thin margin above break-even
The simulated balance, trade by trade
Starting at $0 • one Nasdaq E-mini futures contract • US dollars
The trend hypothesis uses an earlier positive US session that closed in the upper third of its range. It was explored using already-viewed history; it has not passed independent validation.
One good year does not tell the whole story
Asia bullish baseline • simulated net after costs
2026 ends in September. Year lengths and trade counts differ.
About half of the original calls were right.
50.6%
Across 1,828 eligible session outcomes, the original bias correctly anticipated the direction to the session close about half the time.
This is a different test from the 300 Asia bullish trades. A correct direction can still lose money if a stop is reached first.
02 / WHERE WE ARE
Learning from what failed.
Our job is to challenge the idea before trusting it. Rejected models are part of that process.
- 01Complete
Collect the history
Downloaded NQ price and volume history, then built five-minute bars for consistent analysis.
- 02Not approved
Test the first models
General, Asia-specific and trade-outcome models showed unstable results. None earned approval.
- 03Current stage
Audit and rebuild
Corrected session timing, incomplete references, contract changes, Sunday sessions and simulated stop fills. This page uses the audited baseline.
- 04Not started
Test a frozen idea on new data
A future experiment must keep its rules unchanged while recording new outcomes. Historical results alone do not complete this step.
What happened to the earlier reports?
Earlier model and strategy reports were exploratory. The data audit changed eligibility and execution assumptions, so their figures are superseded for direct baseline comparison. A previously described “fresh holdout” also overlapped history already examined, and was not an independent final test.
We have not combined those older figures with the current audited results. No model has been approved for live trading.
03 / UNDER THE NUMBERS
Simple rules. Visible assumptions.
These assumptions define the simulation. Real fills and costs may differ.
Follow the auction
Auction Market Theory studies where trading activity concentrates. Our profile groups each candle’s volume around its average price. It is an approximation, not exchange-level volume at every price.
One defined attempt
Take the first bullish Asia call, enter at the next five-minute bar, and use a stop based on recent volatility. The target is 1.5 times the stop distance. Exit any remainder at session close.
Include the friction
One tick of slippage on entry and exit, plus $5 commission per round trip. If both stop and target occur inside a candle, the simulation assumes the stop happens first.
Keep comparable sessions
Exclude incomplete sessions, known degraded dates and conflicting contract context. Use New York session times, including daylight saving, and the most recent completed US session.
A few terms, in everyday language
- Bullish
- A prediction that price will rise.
- Profit factor
- Total simulated profits divided by total simulated losses. Above 1 means a net gain, before any costs omitted from the simulation.
- Drawdown
- The largest fall from a previous peak in the simulated running total.
- Walk-forward testing
- Make choices using earlier history, then evaluate them on later periods. Repeatedly examining the same history can still influence research decisions.
- ATR
- A measure of recent price movement used to set the stop distance.
THE CONCLUSIONUseful evidence.
Useful evidence.
Not a proven edge.
The audit gives us a clearer starting point. The returns remain inconsistent, the margin is thin, and the current models have not demonstrated dependable performance on new data.