Finding every sleeve's worst day, minute by minute

I have finally completed a full audit of the worst-case intraday (M1) drawdown for all sleeves, resolving the “unmeasured” status that has been a recurring issue since my earlier research. To do this, I built a new reconstruction tool for indices, where M1 data does not exist. By mapping daily Open, Low, and Close prices into a conservative intraday sequence, I can now account for overnight fluctuations in CFDs and ensure that the daily low serves as a conservative upper bound for intraday losses in long-only index sleeves. During this process, I discovered and fixed two bugs that were skewing my previous measurements:

Holiday data gaps: When a symbol was closed, the system was treating the unrealized profit as zero, causing fake equity drops. For example, a fake 8% drop appeared on December 24, 2021. I have updated the logic to carry over the last known price.

Midnight settlement errors: The system was ignoring profit and loss for positions closing exactly at 0:00, creating artificial gaps in the data, such as a 4.3% drop on October 11, 2018. Changing the logic to capture these timestamps correctly resolved the issue. With these fixes, the simulation now matches my equity-based records with bit-perfect accuracy. The worst-case intraday loss is -2.73% (recorded on February 25, 2020, during the COVID-19 crash). In other words, the system’s worst single-day intraday drop is quite contained despite the volatility. Comparing intraday performance to closing-price performance, the ratio sits at a median of 1.24, confirming that the 1.2 multiplier I used in previous research is a solid, middle-of-the-road estimate. I also re-verified the strategy recommended in my previous study using this more rigorous model, which now includes both gap and intraday data. The findings are summarized below:

Strategy Configuration	Monthly Return	Worst Intraday Loss
Dynamic kbase 2.0 / kcap 1.5 + Flat	4.3%	4.3%
Current Conservative (L=1.0, no Flat)	0.61%	5.0%
The “Flat” configuration, where the system pauses trading when risks hit certain thresholds, is mathematically superior. It provides a higher monthly return while simultaneously lowering the probability of account failure. Perhaps most importantly, in 11 years of backtesting, the “Flat” mechanism was never triggered, meaning there was zero cost to having this safety layer active during normal market conditions. All simulated failures in the stress tests were triggered by the “floor” (the daily -5% loss limit), and the “Flat” logic effectively blocked the path to that outcome.
While the night-time lows for indices remain an approximation based on daily data, this measurement provides a much clearer picture of how the system handles intraday stress.

How this connects

This verification builds on earlier ones (what failed before and what I tried this time, comparisons between approaches).