I am running a live Market-On-Open trend-following system with a 10-position cap. Because my system often generates more than 10 signals, I use a weight priority to ensure the backtest/live trading remains deterministic.
However, I’ve noticed that my open positions are shifting unexpectedly between days sometime. For example, on Jan 5, I noticed that LRCX (which had been open for some days and is a live position I'm holding) was missing from the list entirely—neither closed nor open—while LLY appeared in the open list. This change was not caused by any exit signals.
Essentially, the "Open Positions" state on Jan 2 did not match the state on Jan 5, even though no trades should have occurred. The results are consistent when run multiple times on the same day, but they seems to diverge when run on different days.
Is this expected behavior ?
However, I’ve noticed that my open positions are shifting unexpectedly between days sometime. For example, on Jan 5, I noticed that LRCX (which had been open for some days and is a live position I'm holding) was missing from the list entirely—neither closed nor open—while LLY appeared in the open list. This change was not caused by any exit signals.
Essentially, the "Open Positions" state on Jan 2 did not match the state on Jan 5, even though no trades should have occurred. The results are consistent when run multiple times on the same day, but they seems to diverge when run on different days.
Is this expected behavior ?
Rename
It could be any number of reasons, including an unstable weight indicator coupled with a shifting start date. It’s not something I’ve personally seen but if you can reproduce it by changing the backtest date range and stepping forward day by day we can have a look. You can email the relevant evidence to support@wealth-lab.com.
Thanks Glitch. At first I thought of that different start date might be a reason so I used a fixed date and tested it. It still happened so it's unlikely to be that. (Later on I figured start date should not be an issue since the system has regime filter so with ~10 years at a few points all positions would be reset)
I'll try to see whether I could reproduce this tonight.
Assume stable weights and fixed start dates, could "Retain NSF position" cause this kind of issue?
I'll try to see whether I could reproduce this tonight.
Assume stable weights and fixed start dates, could "Retain NSF position" cause this kind of issue?
I don't see how.
If the strategy uses Stop and Limit orders that could both execute on the same bar, WL will randomize which order is filled if the selection in Preferences > Backtest > Other Settings > Exit Prioritization is Neutral (Random). That's a possible source of randomness in a backtest.
@Cone
1. Entry is MOO.
2. No limit orders. Exit is either MOO, or trailing stop based when market assessment is not good. But for the case I mentioned it shouldn't be stop order since LRCX is ripping for the last few days.
I'll see what I can find and provide for further debug tonight.
1. Entry is MOO.
2. No limit orders. Exit is either MOO, or trailing stop based when market assessment is not good. But for the case I mentioned it shouldn't be stop order since LRCX is ripping for the last few days.
I'll see what I can find and provide for further debug tonight.
Your Response
Post
Edit Post
Login is required