This describes how the live system executes. The daily-bar backtest fills the same entries at that same close, so the simulation and the live track use one execution convention. Execution detail, not a performance claim.
Most retail strategies quietly assume you buy at the open. It is the worst assumption in the whole pipeline. The open is the single most violent, widest-spread, most gap-prone print of the day, and momentum names, the ones a breakout system wants, gap the hardest. So we do not trade the open. Every entry the live system places is a market-on-close order, filled in the closing auction. Here is why the close beats the open for a concentrated momentum book, and why the backtest models exactly that.
The open is where retail gets a bad price
At the bell, the overnight order imbalance clears all at once: spreads are wide, the first prints whip, and a name that closed strong can open several percent higher on nothing but a gap. Send a market order into that and you are filled at a price you did not choose, on the day’s thinnest liquidity. For a strategy whose edge is a few percent of expectancy per trade, paying the open’s spread-and-gap tax on every entry is not a rounding error, it is a meaningful slice of the edge, handed away at the worst possible moment.
The close is the deepest, most orderly print of the day
The closing auction is the opposite of the open: the single largest, most orderly liquidity event of the session, where index funds, rebalancers and institutions all cross at one struck price. A market-on-close order joins that auction, and two things follow that matter enormously for a systematic book. First, full size fills, the auction has the depth to absorb the whole intended position at one clean price, instead of walking up a thin intraday book. Second, and less obvious, you size against a price you can already see forming.
Why the close fixes sizing, not just slippage
A systematic book has to decide how many shares, and that decision needs a price. Size against the open and you are guessing: the fill can land well away from your estimate, so you either over-commit capital you do not have or leave intended exposure on the table. Size against the closing print, which the auction telegraphs in the minutes before it strikes, and the share count and the fill agree. The book deploys the fraction of capital it intended, no more and no less. For a fully-invested strategy with hard capital limits, that precision is the difference between running the tested allocation and running an accidental one.
The signal still has to be true at 3:55
Firing at the close has a second, quieter benefit: a full extra session of confirmation. The system builds its candidate list in the morning, then waits, and commits only the names that still qualify in the near-close window. A stock that triggered at the open but fell apart by the afternoon never gets bought. The close is not just a better price; it is a later, better-informed decision. Entering on confirmed strength, into the day’s deepest liquidity, is the whole point.
The backtest models exactly this
A daily-bar backtest cannot fill inside the bar, so it fills at the one print every daily bar definitively records: the close. That is the same print the live system targets with its market-on-close orders, so the simulation and the live execution use the same convention. There is no fill-assumption gap to reconcile, no favorable open assumed in the model that the live book never gets, and no day-later proxy standing in for the real fill. The tested entry and the traded entry are the same kind of price, struck in the same auction. It is the same principle as everything else here: model the strategy the way it actually trades, and publish the live track alongside so the two can be checked against each other.