Monte Carlo
mc@1 defines the probabilistic execution semantics of the Forge Pool execution platform.
Monte Carlo workloads are used when outcomes are better represented as distributions, confidence bands, tail surfaces, or scenario ranges than as single deterministic values.
Like every primitive family, mc@1 executes through the same public execution contract, deterministic execution doctrine, replay model, and verification framework.
Purpose
Monte Carlo execution enables deterministic distributed simulation over uncertain systems.
Typical use cases include:
- credit loss distribution modeling
- IFRS 9 expected credit loss estimation
- insurance capital and solvency risk
- portfolio stress distributions
- climate and catastrophe scenario simulation
- compound risk modeling
The primitive family defines the execution class.
The selected profile defines the domain-specific simulation behavior.
Operation Identity
Monte Carlo workloads are selected through the op section of the canonical execution contract.
{
"op": {
"name": "mc",
"version": 1,
"profile": "finance.credit.loss_distribution_surface.v1"
}
}mc@1 is not a dedicated API route.
It is executed through:
POST /api/v0/ops/executeExecution Characteristics
Monte Carlo execution is characterized by:
- deterministic seed propagation
- independent or partitioned sampling
- execution-unit reproducibility
- convergence-aware execution where supported
- deterministic reduction
- replay-compatible execution evidence
The execution runtime may distribute work across different infrastructure, but the computational meaning of the workload is governed by the execution contract.
Representative Profiles
Representative mc@1 profiles include:
finance.credit.loss_distribution_surface.v1finance.credit.ifrs9_ecl.v1insurance.scr.surface.v1
Additional profiles may exist or be introduced over time without changing the public execution contract.
Execution Parameters
Monte Carlo profiles typically define parameters such as:
- iteration count
- distribution assumptions
- portfolio or exposure inputs
- scenario assumptions
- severity and frequency models
- confidence or percentile targets
- artifact preferences
There is no universal Monte Carlo argument schema.
Each profile defines its own parameter contract while inheriting the shared mc@1 execution semantics.
Deterministic Execution
Monte Carlo workloads are deterministic by contract.
Equivalent execution contracts must preserve equivalent computational meaning when evaluated under the same:
- primitive version
- execution profile
- canonical arguments
- deterministic seed
- execution policy
- reduction semantics
Execution routing, infrastructure placement, runtime identity, and execution timing must not alter the final distribution meaning.
Reduction Semantics
Monte Carlo execution produces distributed partial results that are reduced into a canonical probabilistic output.
Reduction may produce surfaces such as:
- mean and variance estimates
- percentile bands
- tail-risk surfaces
- loss distributions
- scenario summaries
- convergence indicators
- workload-specific statistical artifacts
Reduction is part of execution truth.
It must remain deterministic, versioned, and replay-compatible.
Execution Evidence
Successful Monte Carlo execution may produce:
- execution metrics
- execution-unit summaries
- deterministic seed references
- result hashes
- reduced probabilistic output
- artifact references
- replay metadata
- verification outcomes
Execution evidence is part of the result surface, not a diagnostic afterthought.
Verification
Monte Carlo verification focuses on validating that the probabilistic result was produced under the declared execution contract.
Verification may inspect:
- seed discipline
- execution-unit consistency
- reduction behavior
- artifact integrity
- repeated execution equivalence
- replay metadata
Verification confirms execution trust without exposing internal runtime topology.
Replay
Replay is especially important for probabilistic workloads.
A Monte Carlo result is only defensible when the system preserves enough execution evidence to reproduce the computational meaning of the run.
Replay depends on:
- operation identity
- profile version
- canonical arguments
- deterministic seed behavior
- reduction semantics
- execution evidence
See:
Relationship to Examples
Concrete Monte Carlo capabilities are documented in the Examples section.
Start with:
Examples explain capability behavior.
This document defines primitive-family execution semantics.
Related Documentation
Continue with:
Continue in Forge Studio
To explore Monte Carlo execution interactively:
- Open Capability Explorer
- Filter capabilities by the
mc@1primitive family - Inspect executable blocks in Block Registry
- Execute representative Monte Carlo workloads
- Review execution evidence, artifacts, and replay metadata
Trust should be established through independent verification rather than documentation alone.
Final Note
Monte Carlo is not a special endpoint.
It is a probabilistic execution family implemented through the same public execution contract as every other Forge Pool primitive.
This allows probabilistic workloads to inherit deterministic execution, replay compatibility, execution evidence, and verification guarantees while preserving a stable API surface.