Benchmark Trust Model¶
euroflex_bess_lab separates benchmark modes deliberately.
Oracle benchmarks¶
Oracle benchmarks use perfect foresight. They are useful as:
- an upper bound
- a regression target
- a way to measure the cost of imperfect information
They are not auditable trading workflows.
Examples:
belgium.da_only.perfect_foresight.single_assetnetherlands.da_plus_fcr.perfect_foresight.portfoliobelgium.da_only.perfect_foresight.single_asset.baseline
As-of-safe workflows¶
As-of-safe workflows only use information visible at the decision timestamp.
Built-in auditable providers:
persistencecsv
These runs are the right baseline for forecast-aware public benchmarking.
Revision-aware workflows¶
schedule_revision stays auditable only when:
- the base workflow is itself auditable under the chosen provider
- checkpoints use only data visible at that checkpoint
- locked commitments remain locked
Revision benchmarks produce paired benchmark names:
{market_id}.{base_workflow}.{provider_name}.{run_scope}.baseline{market_id}.{base_workflow}.{provider_name}.{run_scope}.revision
This separates:
- the original D-1 plan
- the revised expected plan
- the final realized outcome
Reserve-lite assumptions¶
Reserve-aware benchmarking in the current release line is intentionally simplified.
- symmetric FCR-style capacity reservation
- asymmetric Belgium-first aFRR with expected-value activation
- capacity-only settlement by default
- expected activation settlement for aFRR only
- optional non-delivery penalty proxy
- no qualification or bid-size granularity logic
Benchmark-grade vs market-grade¶
Benchmark-grade:
- due diligence
- sensitivity analysis
- cross-market comparison
- portfolio screening
- plug-in testing for private forecasts
- checkpoint revision and expected-vs-realized audit
Not yet market-grade:
- live bid submission
- qualification-aware reserve operations
- continuous intraday re-trading
- operational reconciliation against real settlement statements
- EMS-integrated dispatch control