The Entry Price Your Backtest Assumed, and What Approval Could Cost

A weekend backtest that assumes instant fills can make a strategy look cleaner than a trader could execute. Add the time needed to read, assess, and approve each signal, then measure the entry price and risk again.

In 1999, NASA lost the Mars Climate Orbiter as it approached Mars. The investigation found that one team used imperial units while another expected metric units. The spacecraft’s navigation calculations had a hidden mismatch between the model and the real system. NASA documented the failure in its Mars Climate Orbiter Mishap Investigation Board report.

A trading backtest can carry a smaller version of that same problem. The chart may say an entry happened at the instant a condition appeared. A real approval-gated workflow has a queue, a person reading the setup, a decision, and then an order reaching the market. If the model assumes away that gap, it has tested a different system from the one you intend to use.

The fill in the spreadsheet is often the best possible fill

A candle closes. Your rule triggers. The backtest records an entry at that closing price.

In practice, you may need time to check position size, remaining daily risk, the stop location, current exposure, and whether the signal still fits your written rules. Even a quick review creates delay. During a quiet session, that may barely matter. During a fast move, the price can move before approval.

For an illustration, imagine a strategy records a long entry at 100 with a stop at 98. If approval and order placement happen after the price moves to 101, the same stop now risks more per share. Keeping the original position size changes the dollar risk. Reducing position size changes the potential result. Moving the stop to preserve the original structure may mean rejecting the trade altogether.

The original test did not answer which choice you would make. It assumed the choice never existed.

Re-run the test with an approval window

Start with the rules you would actually follow. For every signal, add a delay that represents review and approval. Use several windows rather than one flattering estimate: a short delay for a simple setup, a longer one for a trade that requires checking exposure or market conditions.

Then record what happens to each trade:

• The price at the original signal.
• The price when an order could realistically be approved.
• Whether the stop distance still fits the position-sizing rule.
• Whether the trade would be rejected because it exceeds the risk limit.
• Whether the order would fill at the assumed price, a worse price, or not at all.

This is not a prediction of future results. It is a way to expose an assumption before capital is at risk.

A delayed-entry version may produce fewer trades. That can be useful information. If a setup only works when the trader acts at a speed no human review can match, the strategy may belong in a fully automated system with different controls, or it may need rules designed for slower execution.

Approval changes the decision, not only the timestamp

The approval step has a second effect: it introduces judgment. That judgment can be a source of inconsistency if you have no written rules. It can also prevent a trade that no longer fits the plan.

Before approving, a trader can ask:

• Has price already moved far enough that the planned stop no longer supports the intended risk?
• Does this order add exposure to a position already open in the same market or sector?
• Has the daily drawdown limit changed what the next loss can cost?
• Is this signal based on information available at the time, or does the test accidentally use data revealed later?

Those checks matter because a backtest is an estimate, not a record of what you personally could have done. The model-validation guidance flags unrealistic instantaneous execution alongside look-ahead bias, survivorship bias, overfitting, and weak trade-sizing rules. Each can make a historical result appear more certain than it was.

For a related example of where historical testing ends and a human decision begins, see Backtest confidence versus approval-gated execution.

Treat rejected trades as part of the record

A rejected signal is data. Record the reason: price moved beyond the entry range, position risk exceeded the limit, exposure was already too high, or the setup no longer met the rules.

Over time, that journal shows whether the strategy needs a wider entry range, smaller initial size, fewer trades during volatile periods, or a different signal design. It also shows whether approval decisions are disciplined or emotional.

NASA’s Mars Climate Orbiter did not fail because calculation itself was useless. It failed because the calculation did not match the system carrying it out. Your Sunday backtest needs the same discipline: test the signal, the delay, the sizing rule, and the approval decision together.

Educational content. Not financial advice.