You cannot select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
2.1 KiB
2.1 KiB
Performance Attribution and Reporting
Objective
Performance reporting in Phase 7 must separate alpha quality from operational quality. A winning week is not evidence of healthy production if reconciliation is failing. A losing week is not evidence of broken alpha if execution matched plan and market conditions were adverse.
Reporting Layers
- Strategy-level performance
- Trade-level attribution
- Feature-family attribution
- Operational attribution
- Stage/ramp governance reporting
Daily Report
Required sections:
- positions opened and closed
- realized and unrealized PnL
- gross and net exposure
- slippage summary
- incidents and alerts
- reconciliation outcome
- candidates rejected by guard reason
- top winners and losers
Weekly Report
Required sections:
- expectancy, hit rate, payoff ratio
- attribution by event type
- attribution by hold horizon bucket
- attribution by score bucket
- attribution by approval path
- execution quality summary
- parser/manual review summary
- drift summary
- release/change summary
Monthly Report
Required sections:
- cumulative return and drawdown
- risk-adjusted performance
- capital stage history
- configuration changes and their impact
- recurring incident themes
- operational debt backlog
- recommendation: hold, ramp, freeze, or rollback
Attribution Dimensions
Every trade should be attributable by:
- event_type
- direction
- score bucket
- overlay on/off
- sector
- approval path
- config version
- parser version
- feature version
- live stage
Minimum Metrics
- gross pnl
- net pnl
- expectancy
- win rate
- average win
- average loss
- profit factor
- average holding period
- turnover
- realized slippage
- percent of days with incidents
- percent of trades requiring manual intervention
Interpretation Rules
- Do not attribute alpha improvement to a release without sufficient sample.
- Do not attribute operational issues to market conditions without evidence.
- Compare live vs simulation using matched assumptions where possible.
- Maintain a stable definition of score buckets across review periods.