ReplayState

Solana Backtesting Engine

Simulate your transactions against real historical block data using Monte Carlo analysis. Test fee strategies across thousands of stochastic trials. Quantify MEV risk before committing capital. ReplayState runs your transactions through real network conditions, scheduler contention, and sandwich attack models to answer: “What fee should I use? Am I getting sandwiched? How reliable is my inclusion rate?”

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

Helius

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.

Alpha

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.

Risk

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

1Fetch & Replay

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.

2Monte Carlo Shadow Run

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.

3Signed Report

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.

📡
Fetch
LaserStream gRPC / archival RPC
Sub-second block ingest
💾
Boot
Snapshot archive loading
Full bank state restoration
▶️
Replay
Agave BPF execution workers
Deterministic parity (FR-DET-001)
🎲
Shadow Run
Monte Carlo trial engine
100K+ trials, reproducible seeds
📊
Report
Provenance-signed manifests
Sigstore + SHA-256 verification

Capability Comparison

CapabilityDemo (Fast Preview)Production Engine
RuntimeOff-chain AMM mathFull Agave BPF (v3.1.8 + v3.0.14)
Data SourceArchival RPC getBlockLaserStream gRPC (Helius) + CAS
SandboxingNoneLinux namespaces + seccomp + cgroup v2
ProvenanceNoneSHA-256 manifest, Sigstore signing
Trials100–10K, ~200/s100K+, distributed workers
ValidationGround truth badgesFR-DET-001 parity (err, fee, balances)
MonitoringAPI rate limitsPrometheus /metrics, stage histograms
BatchSingle scenarioNDJSON 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
427+ automated tests
100K-trial stochastic regression
Sigstore-signed releases