Choosing Realistic Entry Prices for Intraday Backtests

Summary

The document asks how to model trade entries for a strategy that generates signals on one-minute data. The author describes shifting entries from the signal bar’s close to the next minute’s open, adding a stop-loss, and accounting for brokerage and other charges. Despite strong in-sample and out-of-sample results, live fills often differ from the prices assumed in testing, possibly because of the delay between generating and placing an order.

The author considers using the midpoint of each bar’s high and low, but the document does not establish that this is a reliable fill model or report a comparison of alternatives. It raises the practical question of how to align historical and forward tests with actual execution. The evidence is limited to the author’s reported experience; there are no fill records, market or order details, or analysis showing how large the discrepancies are. The proposed midpoint should therefore be treated as a question for investigation, not a validated recommendation.

Key ideas

• One-minute signals can produce live entries that differ from prices assumed in a backtest.
• The author shifts the assumed entry from the signal bar’s close to the next minute’s open.
• Stops and brokerage or other charges are included in the described testing process.
• The high-low midpoint is proposed as a possible entry assumption, but is not evaluated in the document.
• The reported test results do not establish that the live execution model is realistic.

Tags

# Entry price to consider while doing a forward/backtest test?


# Entry price to consider while doing a forward/backtest test?












1 min timeframe signal generation.

Thing I have considered:

1 - `Close` price can be predictive in nature

Counter: shifted the `entry` price to the open of next minute.

2- Added stop-loss.

3- Brokerage/other charges

I did in-sample fitting and then out-sample test. Both the results are very good.

The strategy is now live and I am facing certain issues.

The entry price almost never matches, I assume this is because of the `seconds` gap between order generation and placement.

I realise that this will happen and I can't change much about it at this time. I am trying to understand:

What Entry price should I consider in the backtest and forward test(after market I run the strategy on closed data)?

I am thinking of this:

`(high + low)/2`

Is there a universally accepted/recommended approach?

TIA

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.