I began by examining a clear hypothesis: if QQQ stayed within a range through expiration, a short iron condor—a call credit spread paired with a put credit spread—could benefit from time decay. The structure offered a concrete way to study that idea in short-dated options. It was a research question, not a conclusion drawn from a personal trade or evidence of an edge.
The range-bound idea has a boundary
An iron condor is often discussed as a range-bound, net-credit position. In a quiet session, with the underlying remaining away from the short strikes, time decay can work in its favor. But “the market looks range-bound” is not a durable condition. It is an observation that can stop being true while the position is still open.
If QQQ begins moving decisively in one direction after entry, one side of the position can come under pressure quickly. The other side does not necessarily offset that pressure in a useful way: its value, liquidity, and sensitivity can change differently. A position that looked balanced when opened can develop meaningful directional exposure as the underlying and option prices move.
For a fully intact four-leg position, the familiar maximum-loss calculation describes the expiration payoff under its assumptions: the width of the tested wing less the net credit, before fees. It is not a blanket limit on every operational outcome. Early assignment or expiration uncertainty can leave stock exposure or an unbalanced position; partial fills, closing legs separately, gaps, fees, and slippage can also make the realized result differ from the simple payoff calculation.
What changes after entry
With options close to expiration, gamma deserves particular attention. As an option approaches expiration—and especially as the underlying approaches a strike—delta can change more sharply for a given move in the underlying. The position's net directional exposure can therefore shift while I am managing it. A delta reading at entry is only a snapshot, not a promise about the rest of the session.
The displayed mark is not the same thing as an executable exit. Bid-ask spreads can widen, quotes can move between observation and submission, and a multi-leg order may be difficult to close at a single attractive price. Slippage on an exit can change the realized result materially. These are execution and liquidity risks, not details that can be safely ignored because the opening order received a credit.
I want my research to treat the entry and the exit as parts of one decision. A setup should not be judged only by the credit collected or by how often a historical path appears to finish inside a range. The possible loss, the cost and uncertainty of closing, and the conditions that would invalidate the original premise all belong in the question.
What the records can—and cannot—say
The local read-only audit for the U.S. trading date September 11, 2026, reports 4,652 QQQ decision records. Those are decision rows—not completed trades, orders, or a count of trading outcomes. The audit also found candidate enumeration incomplete across those records, with zero complete Shadow Outcomes and zero formal training decisions. These are data-quality facts only; they say nothing about whether the strategy made or lost money.
I keep three things separate:
- A historical backtest is an estimate from historical inputs and explicit modeling assumptions. It is not a live fill record.
- Paper Trading uses a simulated brokerage environment. It can help inspect a workflow, but it is not live-trading performance.
- Live trading involves real orders and capital. This note does not report a live trade or a live result.
This note does not present a trade log, account statement, or verified performance study. Where the data is incomplete, the conclusion should remain incomplete too.
Why I stopped the QQQ AI Training plan
I have decided to stop the QQQ AI Training plan and will not continue investing in sample collection for it. My research attention is moving to a more immediate question: what happens after an iron condor is open, when the market starts moving in one direction and exposure, exit costs, and risk decisions can change together. That shift in focus—not a claimed result from a model—is the reason to stop this line of work.
The data-quality issues described above are an additional research limitation, not the sole reason for the decision. More rows would not, by themselves, answer the questions I now want to investigate about post-entry behavior and exits.
The next questions are about the open position
My next research focus is narrower: how to recognize a meaningful change in the underlying's behavior after entry, how directional exposure and gamma change while a position is open, and how to evaluate exit choices, transaction costs, and risk controls under realistic spread and fill assumptions. These are questions to investigate, not controls I am claiming have already been validated.
Any future test needs clear data provenance, complete candidate and outcome records, and a separation between modeled fills and observed Paper or live executions. It should also state what it cannot establish. Until that work is done, I do not have a verified conclusion that a particular entry filter or exit rule improves the strategy's risk/reward profile.
Rethinking the strategy does not mean I have found a better one. It means the research question has changed: I want to understand the position after entry, not just the conditions that appear attractive before it.
This article is a research journal, not investment advice or a recommendation to trade options. Trading options involves substantial risk, including the risk of loss. Historical analysis does not guarantee future results.