ProOrder live entries differ from backtest entries…..

Hi JS

Thanks again for your help with this. Your suggestion to use the diagnostic graphs has been useful, and I think we’ve narrowed the question down a little further.

I’ve now run a new diagnostic copy of the 5/1/10 strategy using the same parameters and 20 October 2026 date. The order list still matches the original backtest.

I’ve attached the complete diagnostic script, the relevant backtest order report, the live order report, and the two graph screenshots around the trade in question.

I’ve added graphs for ShortSignalNow, FullShortEntryCondition, PositionState and ShortOrderCommandSeen to help us understand the strategy’s behaviour.

The graphs show the following:

• 18:08:40: ShortEntryScore = 15, ShortSignalNow = 1, FullShortEntryCondition = 1, ReadyToTrade = 0 and PositionState = 0.
• 18:08:50: ShortEntryScore = 6, ShortSignalSeen = 1, ShortOrderCommandSeen = 1, ReadyToTrade = 0 and PositionState = 0.

The backtest records the short entry at 18:08:50, whereas the live order report records the corresponding short entry at 18:10:40 — a difference of approximately 110 seconds.

I appreciate that the graph values may not reveal the exact sequence of internal strategy evaluation and order processing, so I don’t want to jump to conclusions.

Could I ask for your opinion on two points?

1. Does this pattern look consistent with ProOrder evaluating the completed 18:08:40 bar and submitting the order for the next bar, despite ReadyToTrade plotting as zero? Or does it suggest that we should investigate the re-arming logic or another aspect of the code?
2. Having seen the script and these results, is there another diagnostic test you would recommend to help establish why the live entry occurred approximately 110 seconds later?

I’m keen to understand whether the discrepancy originates in the strategy’s signal evaluation, order processing, or execution conditions before considering any changes to the code or moving to another platform.

I realise this is a fairly specific issue, and I appreciate the time you’ve already spent looking into it. Your suggestion to use the graph functions has given us a more concrete way to investigate, so thank you.

Any thoughts or suggestions would be very welcome.

Screenshot-2026-10-21-213239.png Screenshot-2026-10-21-213215.png Screenshot-2026-10-21-213254.png

this is the script I ran, what in your opinion is the reason for the discrepancy between them in order to get different results?

Screenshot-2026-10-22-014110.png diagnostic-test-5-1-10-on-20-10-26.itf

Good afternoon to you both.

JS you are correct, a UK DFB account is traded.

Nicholas the live tests are done on a live autotrading system.

So before posting my strategy script and live-versus-backtest results, I thought it would be helpful to explain how I develop and select the strategy I trade each session.

I’m researching two multi-timeframe versions of a US Tech 100 (DFB) swing-based trading system in ProRealTime v12:

• Strategy 1: 05/01/10 — 5-minute, 1-minute and 10-second timeframes.
• Strategy 2: 10/02/20 — 10-minute, 2-minute and 20-second timeframes.

Both use the same general approach to directional alignment and entry scoring, with a 12-point stop loss and a 10-point profit target.

My pre-trading selection process

Before each trading session, I optimise both strategies over six consecutive, overlapping one-hour windows. The windows move forward in ten-minute increments.

For each strategy, I examine two separate sets of results:

• Six GAIN-optimised results: the variable combinations selected to maximise points gained.
• Six %WIN-optimised results: the combinations selected to maximise the percentage of winning trades.

This gives me 12 optimisation results for each strategy.

The three variables being optimised are:

• A: Minimum entry score.
• B: Minimum EMA spread.
• C: Minimum EMA slope.

I assess the results for consistency rather than simply choosing the best individual result or automatically favouring the latest window. I look for repeated variable combinations, stability between windows, changes in trade frequency, and whether the GAIN and %WIN results support the same settings.

I also calculate weighted averages of the GAIN-optimised variables as an additional research reference. These are not automatically used for live trading: the final settings must be justified by the evidence, and any live selection must be clearly distinguished from the original optimisation results.

What I’m trying to establish

The objective is to identify which of the two strategies offers the strongest evidence for the next session, select defensible settings, and then compare the actual live results with ProBacktest.

The strategies are still under development, so I’m not claiming that the selection process guarantees profitable results. I’m trying to make the process systematic, repeatable and evidence-based.

My current concern is that live trading does not always appear to reproduce the entries shown in ProBacktest, even with the same strategy script and settings being used.

I’ll post the script and corresponding results for the strategy selected for the next trading session tomorrow. I’d particularly welcome feedback on the reliability of the pre-trading process, as well as any technical explanation for differences between ProBacktest and live execution.

Thanks for taking the time to look at this.

Please find attached the script used for todays (22/10/26 17:00 – 19:30 UK time) trading session.