je-suis-tm Parabolic SAR: Faithful Daily EA Long/Cash Port

Hypothesis

Derived by the Research Lead from je-suis-tm's Apache-2.0 'Parabolic SAR backtest': preserve the supplied custom SAR recursion, default acceleration factors and binary long/cash signal on the source's EA daily market, using EA.USEQ. Source: https://github.com/je-suis-tm/quant-trading/blob/611b73f2c3f577ac5b28aaa19ac8c43d3236c7a5/Parabolic%20SAR%20backtest.py. Although the script plots signals rather than submitting orders, its position series defines a testable trading rule. The plotted 'SHORT' marker means closing a long position, not opening a short. Platform completion supplies market execution at the completed daily bar's close, equity-relative sizing and modeled costs. No next-session-open execution, intrabar SAR stop, additional filter or discretionary exit is introduced. The original download interval is 2016-01-01 through 2018-01-01, with the end exclusive; retain it as a reproduction report and separately evaluate the same frozen rules over available history. slicer=450 affects plotting only: calculate signals before slicing and never use it as a trading warm-up.

Preserve the source's EA equity market and daily frequency instead of replacing it with a crypto trend variant. The corpus_concentration object is empty, so no claim about EA's recent share is made. Single-symbol and long-only exposure are overrepresented, but adding hedges, shorts or a basket would alter the requested faithful port. The supplied source's sell markers do not justify a long-short adaptation.

EA.USEQ: delisted 2026-08-04; Yahoo history gone

Implementation

Faithful daily EA long/cash translation of je-suis-tm's custom Parabolic SAR. Buys on a source position transition from zero to one and closes the long on a transition from one to zero. Entries allocate 20% of settled equity in whole shares. Uses completed-bar market execution and platform costs.

Preserves the supplied initialization, signed trend counters, two-prior-bar constraints, separate recursive and real SAR values, extreme-point updates, acceleration factors, strict comparisons and positions.diff() order sequence. The initial undefined difference generates no order. All recursive state uses persistent scalar attributes and constant per-bar work. No shorts, additional filters, stops or discretionary exits are added. Signals begin at row zero; slicer=450 is not a strategy warm-up. The supplied research plan is retained verbatim, but its 450-bar evaluation warm-up conflicts with interpreting the source slicer as plotting only; reproduction must reset at the first EA session on or after 2016-01-01 and end before 2018-01-01, without a 450-bar trading exclusion. Separate available-history evaluation uses the same frozen rules. Reports and verification remain pipeline responsibilities. The required outbox file could not be written because this workspace permits read-only filesystem access; the complete task-bound JSON is returned here.

Outcome Summary

Before coding a single-ticker port of a source strategy, check that the instrument is still listed and its history is still available, or plan a substitute data source for delisted tickers.

The run was abandoned on its first iteration because its only instrument had no data: 'EA.USEQ: delisted 2026-08-04; Yahoo history gone'. No backtest review, analyst verdict or risk review took place.

A faithful port of je-suis-tm's Apache-2.0 Parabolic SAR backtest on EA.USEQ daily bars: it kept the source's custom SAR recursion and default acceleration factors and went long when close was above SAR, returning to cash otherwise (no shorts), with execution at the daily close.

No performance figures exist: the backtest report is empty and no optimization report was produced, so there are no return, Sharpe or trade-count numbers.

Analysis

Parabolic SAR backtest