# Detection Engine Hardening Plan

## Objectives
Ensure the detection pipeline produces actionable opportunities with production-grade heuristics and hands them off to the execution stack automatically. Current implementation primarily logs opportunities and uses simplified scoring parameters.

## Action Items

1. **Opportunity Scoring & Filtering**
   - [ ] Define canonical scoring function (profitability, confidence, liquidity, latency) aligned with business thresholds.
   - [ ] Replace ad-hoc ROI/price-impact checks with configurable policies (pull from config/env).
   - [ ] Add regression tests covering borderline cases (low confidence, high slippage, gas-constrained opportunities).

2. **Execution Handoff**
   - [x] Implement direct handoff to `ArbitrageExecutor` via callback (see `SetOpportunityHandler`).
   - [ ] Ensure backpressure handling (bounded queue, metrics when saturated).
   - [ ] Introduce retry/backoff for transient execution failures.

3. **Concurrency & Worker Pools**
   - [ ] Review `WorkerPool` sizing, ensure graceful shutdown and error propagation.
   - [ ] Add instrumentation (metrics/logs) for scan time, queue depth, dropped opportunities.

4. **Historical Validation**
   - [ ] Replay recorded blocks/datasets to validate hit rate and false positives.
   - [ ] Store misclassified opportunities for manual review.

5. **Configuration & Docs**
   - [ ] Centralise detection thresholds in config files (`config/*.yaml`) and document in `docs/5_development/CONFIGURATION.md`.
   - [ ] Update runbooks to describe detection tuning knobs and alerting thresholds.

## References
- `pkg/arbitrage/detection_engine.go`
- `pkg/arbitrage/service.go` (execution integration)
- `pkg/metrics` for instrumentation