fix(polymarket): verify executable top against depth

[proerror77]

Change contract

Verify each Polymarket top bid/ask against the unique best executable level in retained full depth using the collector's strict 0.02 < price < 0.98 predicate; accept an absent top only when that side has no executable level.

Closes #302.

Out of scope

Collector or producer behavior; evidence bytes, manifests, digests, or republishing; snapshot sampling; research projection; evaluator/MCTS; execution policy; Paper or Live.

Dependency / merge order

Independent focused verifier fix on current main. Merge this PR, publish its exact-main research image, then rerun #232 against the unchanged evidence anchor. No stacked PR.

Focused validation

• cargo test -p ploy-market-data — 63 passed
• cargo test -p ploy-research checked_in_btc_and_sol_templates_pin_current_brief_and_rust_policy -- --nocapture — passed
• cargo fmt -p ploy-market-data -- --check — passed
• cargo clippy -p ploy-market-data --all-targets --no-deps -- -D warnings — passed
• git diff --check — passed
• Exact-boundary positive fixture covers ask depth 0.02/0.03 with top 0.03 and complementary bid depth 0.98/0.97 with top 0.97.
• Counterexamples cover missing executable top, non-executable top, wrong price/size, duplicate best level, and a crossed executable book whose top still matches depth.
• Independent spec and standards reviews: APPROVE, no blockers.

Cloud reproduction / pending acceptance

snapshot-issue232-main-0724b failed at the old absolute-depth predicate. Pinned read-only diagnostic snapshot-issue232-book-tradeable-proof-0724b checked all 532 book rows, retained 2,128 boundary levels, and found zero top mismatches under the collector predicate. Post-merge exact-main cloud acceptance remains pending by design.

Rollout / rollback impact

One verifier predicate plus targeted tests and mechanically required governed mission pins. Rollout is the exact-main research image; rollback is a one-commit revert. Immutable evidence and successful artifacts are unchanged.

Workflow

Matt diagnosing-bugs established the cloud counterexample; Matt TDD reproduced it before the surgical fix; Matt implement and two-axis code-review completed the focused issue. A PRD was unnecessary because this is one reproduced producer/verifier contract drift.

Summary by CodeRabbit

• Bug Fixes

  • Improved Polymarket order book validation by focusing on executable market depth, while correctly preserving non-tradeable boundary levels.
  • Added safeguards to reject crossed or inconsistent executable top-of-book data.
  • Improved handling when only boundary depth is available.

• Configuration

  • Updated search policy snapshots for BTC and SOL five-minute research missions.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 0b82abc0-2cac-450c-a100-ed3408527d57

📥 Commits

Reviewing files that changed from the base of the PR and between 888c3fe and 3fd723c.

📒 Files selected for processing (3)

• rust_hft/prediction-markets/config/research_missions/polymarket-btc-5m.example.json
• rust_hft/prediction-markets/config/research_missions/polymarket-sol-5m.example.json
• rust_hft/prediction-markets/crates/ploy-market-data/src/polymarket_evidence/verified.rs

---

📝 Walkthrough

Walkthrough

The PR updates two Polymarket research mission snapshot identifiers and changes orderbook verification to compare top-of-book values with executable depth, including boundary-level handling, crossed-book rejection, and targeted tests.

Changes

Polymarket orderbook verification

Layer / File(s)	Summary
Executable depth matching and validation rust_hft/prediction-markets/crates/ploy-market-data/src/polymarket_evidence/verified.rs	validate_book now matches bids and asks against the best tradeable depth level, preserves boundary levels, updates rejection text, and adds coverage for executable, crossed, and boundary-only cases.

Research mission policy snapshots

Layer / File(s)	Summary
Mission snapshot references rust_hft/prediction-markets/config/research_missions/polymarket-{btc,sol}-5m.example.json	The BTC and SOL five-minute mission examples reference new search_policy_snapshot_id SHA-256 values.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Possibly related PRs

• proerror77/monday#145: Updates the same BTC/SOL research mission snapshot identifiers, but changes a different substantive pipeline.
• proerror77/monday#146: Updates the same mission snapshot identifiers, while its substantive logic concerns repricing and fillability.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly summarizes the main verifier fix and matches the changeset.
Description check	✅ Passed	The description is mostly complete and matches the template; only the optional Scope exception section is not explicitly written as None.
Linked Issues check	✅ Passed	The verifier now checks the unique executable top against depth and preserves the required edge-case rejections, matching #302.
Out of Scope Changes check	✅ Passed	The extra research mission snapshot pins appear to support the verifier rollout, and no clearly unrelated changes stand out.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch codex/polymarket-tradeable-top-302

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3fd723c61d

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Filter non-executable levels before projecting depth

When a side contains a boundary level plus an executable top (for example, the new 0.02/0.03 ask fixture), this accepts the row but retains the 0.02 level. project_verified_polymarket_evidence copies every retained level into ResearchPmBookSnapshot (polymarket_evidence_projection.rs:155-179), while sweep_buy_to_stake_with_config and the sell sweep accept every price strictly between 0 and 1 (factors_v2.rs:8204-8207, 11126-11128). Consequently the executable-PnL evaluator consumes liquidity that this verifier explicitly classifies as non-executable, overstating fills and corrupting settlement/PnL labels whenever boundary depth is present. Apply the same strict predicate to projected/swept depth, rather than only when selecting the top.

Useful? React with 👍 / 👎.