Deterministic Solana slot replay, plus Monte Carlo simulation.
ReplayState takes any historical Solana transaction, replays its slot using the real Agave runtime against archival ground truth, then runs thousands of stochastic trials to produce the full outcome distribution — inclusion probability, slippage bands, MEV exposure, and causal attribution. It answers: “Was my transaction outcome fair, and how do I make it better?”
6
archival mainnet blocks
real data, not mocks
3
AMM adapters
Raydium · Orca · Meteora
428
passing tests
0 failures, 0 warnings
100K+
trials per run
reproducible seeds
RPC Providers
Turn your archival data advantage into a premium analytics product.
You already have the blocks and snapshots. ReplayState is the replay and simulation layer that converts that infrastructure into execution intelligence your users will pay for.
Market Makers & Traders
Stop guessing at fee strategy. Get empirical distributions.
Run any historical transaction through thousands of stochastic trials to see the p5/p50/p95 range of outcomes — inclusion rate, slippage, MEV exposure — under real network conditions.
DeFi Protocols
Measure MEV and inclusion risk before deploying capital.
Understand how your protocol's transactions behave across congested, calm, and adversarial blocks. Quantify sandwich attack probability and optimize execution parameters with data.
How it works
Pull a historical block from archival RPC. Boot a snapshot of Agave bank state at that slot. Replay every transaction in exact on-chain order — field-for-field parity against ground truth.
Run 1,000–100,000 stochastic trials. Each trial re-samples network latency, arrival order, MEV activity, and scheduler contention from calibrated distributions. Seeds are SHA-256 deterministic — every run is reproducible.
Aggregate p5/p50/p95 outcome distributions, causal attribution (network vs contention vs MEV), and fee optimization recommendations. Manifests are SHA-256 signed with Sigstore for auditability.
Model Accuracy Scorecard
Predictions validated against real on-chain outcomes across 6 archival blocks
Production Architecture
Everything above is our fast interactive preview. Here is what the production engine does.
Capability Comparison
| Capability | Demo (Fast Preview) | Production Engine |
|---|---|---|
| Runtime | Off-chain AMM math | Full Agave BPF (v3.1.8 + v3.0.14) |
| Data Source | Archival RPC getBlock | LaserStream gRPC (Helius) + CAS |
| Sandboxing | None | Linux namespaces + seccomp + cgroup v2 |
| Provenance | None | SHA-256 manifest, Sigstore signing |
| Trials | 100–10K, ~200/s | 100K+, distributed workers |
| Validation | Ground truth badges | FR-DET-001 parity (err, fee, balances) |
| Monitoring | API rate limits | Prometheus /metrics, stage histograms |
| Batch | Single scenario | NDJSON batch orchestration |
Helius Integration
- ●LaserStream / Yellowstone gRPC for real-time block streaming with sub-second latency
- ●RPC credit budget enforcement with per-run cost limits and automatic throttling
- ●ReplayState is the analysis layer on top of Helius infrastructure — Helius provides the data pipeline, ReplayState provides the simulation engine