A trade that met its setup rules should still be rejected when the live bid-ask spread exceeds a limit set in advance. The wider spread changes the entry cost, position size, and loss calculation, so the original approval case no longer applies.
On August 1, 2012, Knight Capital began sending a flood of unintended orders into U.S. equity markets. The company had deployed software for a new trading function, but one server still contained obsolete code. When the new system activated, that old code started buying and selling shares.
Knight’s staff knew something was wrong while the orders were still running. The outcome remained unresolved for roughly 45 minutes. By the time the activity stopped, Knight had accumulated positions that produced a loss of more than $460 million, according to the U.S. Securities and Exchange Commission’s 2013 order concerning Knight Capital Americas. Thomas Joyce, Knight Capital Group’s chief executive at the time, was left leading a company whose survival was suddenly uncertain.
The event was far larger than one retail order on a quiet Friday. The mechanism still matters at retail scale: a condition changed, the original assumptions no longer held, and execution needed a hard stop.
A valid setup can become an invalid order
The Friday signal may still look intact. Price remains near the planned entry. The stop still sits below the same technical level. The profit target has not moved.
Then the spread widens.
Suppose the plan assumed an entry near the displayed midpoint, with a maximum acceptable spread written into the rules before August liquidity thinned. By Friday, the ask has moved far enough from the bid to exceed that limit. Buying at the ask now means starting the position with a larger immediate cost.
That cost affects more than the first line of the trade confirmation. It changes the distance from the actual fill to the stop. If position size was calculated using the earlier entry, the order may risk more money than intended. If the size is reduced to preserve the risk limit, the expected reward may no longer justify the trade.
The chart can remain unchanged while the executable trade becomes worse.
This distinction matters because a signal describes an opportunity under stated assumptions. An order is the price and size available now. Approval belongs to the order, not the earlier signal.
The August rule exists for the inconvenient moment
A spread limit written before summer serves as a precommitment. It records what counts as acceptable while there is no open trade, no Friday deadline, and no urge to salvage a setup after spending time on it.
The useful rule is measurable:
- Record the maximum spread permitted at entry.
- Calculate risk from the expected fill, not the chart’s midpoint.
- Reject the order when the live spread exceeds the limit.
- Reassess later only as a new decision, with a fresh entry and position-size calculation.
The rejection can feel overly strict when the spread misses the threshold by a small amount. That discomfort is the point. A risk rule that applies only when convenient provides little protection.
A trader might be tempted to use a limit order and assume the problem is solved. A limit order can cap the entry price, but it cannot guarantee a fill. A partial fill can also leave the trader with a position different from the one evaluated. The approval decision still needs to account for those outcomes.
For a related example of how execution cost can consume the original trade economics, see How Much Profit Can Slippage Consume in Quiet August Markets?.
Approval gates separate analysis from permission
An approval-gated trading process keeps two decisions distinct.
First: does the market setup satisfy the strategy?
Second: does the order available now satisfy the execution and risk rules?
The first answer can be yes while the second is no. That is not a contradiction. It is evidence that the approval gate is checking current conditions instead of repeating the signal’s conclusion.
For TraderCoach, the AI can generate and queue a trade signal, but a human approves or rejects it before execution. A spread breach is the kind of observable fact that should appear at that checkpoint: planned entry, current bid, current ask, expected fill, revised stop distance, revised position size, and the rule that failed.
No forecast is required. The trader does not need to decide whether the spread will narrow in five minutes or widen further into the close. The present order has already failed the written condition.
This approach also produces a more useful trading journal. “Rejected because the spread exceeded the prewritten maximum” can be reviewed later. “It felt too expensive” cannot be tested consistently.
Keep the rejection in the record
After the market closes, preserve the signal and the reason it was rejected. Record the spread observed during review, the maximum allowed spread, and how the expected fill changed planned risk. Do not replace the original figures with cleaner closing data.
Then review the rule across a meaningful sample. If the limit rejects too many otherwise sound trades, change it between trading sessions and document why. Do not widen it for the order currently waiting for approval. That converts a risk rule into a negotiation.
Knight Capital’s obsolete code caused damage because activation reached live markets without an effective control stopping the resulting orders. A retail trader faces smaller numbers, but the operational lesson is direct: when a live condition crosses a boundary, the control must act before the order does.
On the next Friday review, place the live bid, ask, spread limit, expected fill, and revised risk on one screen. If the spread is beyond the August rule, reject the order and keep the evidence.
Educational content, not financial advice.
Comments
No comments yet.