fix(MarketResilience): require the depleted side to come back at its price, not just its size

[silahian]

The defect

Immediacy-weighted depth is measured from the touch of the frame being measured, so the reference price moves with the book. A bid that was swept and rebuilt three ticks lower therefore read as a full recovery — the size was back, but the price that had been bid for was gone.

The score published 0.3125 and the bias tile said nothing, on a book that had just given way.

The rule

A depleted side now counts as recovered only when both hold inside the window:

1. its immediacy-weighted depth is back to 90% of the way from its trough to its baseline (unchanged), and
2. its best price is back within one spread baseline of where it was quoted on the frame before the depletion was detected.

Where the anchor comes from

The frame before the depletion — not either of the two obvious alternatives:

• Not the depletion frame itself. On a sweep that frame already shows the moved touch, so the event would be measured against its own damage.
• Not the anchoring print. A print admits events for the whole shock window — five seconds by default — so that price can be seconds stale. Ordinary drift between the print and the depletion would then read as a failure to redeploy and publish a directional arrow from nothing.

Why one spread baseline

It is the only distance unit already in the file, it is the instrument's own typical distance, and it carries no tick size and no price scale — this plugin serves crypto, equities, futures and FX. Routine requoting inside the spread still counts as a recovery; a touch that moved further than the instrument's own spread does not.

The alternatives were each rejected for a concrete reason: zero tolerance fires on ordinary quote competition, a fraction of a spread is a second constant with no basis, and a tick count is a hardcoded multiplier.

The test is one-sided. A side that came back at a better price came back, and is credited. The tile reports which side FAILED, and nothing failed there.

Skipped, crediting the recovery

When there is no anchor (the print arrived before any book) or no measurable spread baseline. Neither can support the comparison, and silence is the safe direction for a tile that only speaks about failures.

Also in this change

ActivateDepthEvent's cold-start spread baseline floored at 1.0 regardless of the book, while IsLOBDepleted prefers the book's own spread when it has one. That measured the event on a different scale than the detector that admitted it, and on an instrument quoting below 1.0 it is a tolerance larger than the whole price. The two now agree.

Both tile tooltips are corrected: for the new rule, for the arrow holding its last reading between the hysteresis levels rather than resetting to neutral, and for the first-recovery rule, which had lost the qualifier that a first-of-session non-recovery does publish.

Deliberately NOT changed

The plugin Description strings. They are hashed into the plugin id that keys saved settings, so editing one orphans every existing user's symbol and provider selection and leaves the tile dead on launch. Verified byte-unchanged in this PR. The corrected wording lives in the tooltip, which is not hashed.

Evidence

dotnet test tests/Unit/Studies.MarketResilience.Test → 68 passed, 0 failed, with no test modified by this PR.

dotnet build VisualHFT.Plugins/Studies.MarketResilience → 0 errors.

The rule was mutation-checked three ways before this port:

mutation	result
price test removed entirely	4 tests red
written as a symmetric band abs(touch - anchor) <= tolerance	1 test red — the better-price case
anchor read off the depletion frame instead of the frame before	2 tests red

Reviewer note

The zero-tests problem called out in #79 is gone — this test project now executes and reports its 68 results normally.

[Copilot]

Copilot review overview

🟡 Changes recommended

The documented “skip anchor check when no measurable spread baseline” behavior isn’t achievable with the current state model because the anchor tolerance is conflated with the depth-normalization baseline.

Get a fresh assessment by requesting another Copilot review.

Review effort: Lite
Findings: 1

Open (1)

• Separate anchor tolerance from depth normalization baseline · New

What changed in this PR

Updates the Market Resilience study so a depleted side is only credited as “recovered” when depth returns and the best price returns close enough to the pre-depletion touch, preventing false recoveries after repricing.

Changes:

• Anchor the recovery price check to the previous book frame (pre-depletion) and require the touch to return within one spread-baseline.
• Align cold-start spread-baseline behavior between depletion detection and depth-event activation.
• Update both tiles’ tooltip text to reflect the new recovery rule and clarified hysteresis behavior.

File	Description
VisualHFT.Plugins/​Studies.MarketResilience/​Model/​MarketResilienceCalculator.cs	Adds previous-touch anchoring and enforces price-return gating for depth recovery.
VisualHFT.Plugins/​Studies.MarketResilience/​MarketResilienceStudy.cs	Updates Market Resilience tooltip to describe price-return requirement.
VisualHFT.Plugins/​Studies.MarketResilience/​MarketResilienceBiasStudy.cs	Updates Bias tile tooltip to reflect price-aware recovery and hysteresis persistence.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.