What Tissue is
Tissue prices live football matches from first principles and refuses to trade on information it doesn’t have. It never chases the market. Instead, it freezes a pre-match fair-value model from TxLINE’s opening consensus odds, then reprices continuously as the match unfolds — using verified score state, match minute, red cards, and match-phase context. When its own fair value diverges from TxLINE’s live market price by more than a configured threshold, it publishes a two-sided quote. When the market moves in a way its own model can’t explain, it pulls its quotes and halts rather than guess. Every one of those decisions is hash-chained, Ed25519-signed, and replayable byte for byte. The desk’s opening priced markets are anchored on Solana before kickoff, and its ledger is anchored again at intervals through the match — so nothing about what Tissue believed, and when, has to be taken on trust.Why independent pricing matters
Most in-play trading tools are reactive: they watch a market move and try to react faster than someone else. That’s a latency race, not an edge, and it has nothing to do with whether the market’s new price is actually right. Tissue does the opposite. It computes what a Poisson/Dixon-Coles goals model — built from verified score state and match time, not from copying the market — believes the match is worth right now, independent of where the market currently sits. The gap between those two numbers is the only thing that triggers a quote. If there’s no gap, there’s no trade, and that’s treated as a correct outcome, not a missed opportunity.What’s actually verifiable, today
- Deterministic replay.
replay(corpus) === ledgeris asserted in CI on every build — the exact same feed always produces the exact same hash chain. - Ed25519-signed decisions. Every ledger record is signed with the operator’s real Solana keypair. Anyone with the public key can verify authorship independent of trusting the process that produced the JSONL file.
- Merkle inclusion proofs. Any past decision can produce an O(log n) proof that it’s included in a specific on-chain-anchored root — not just “trust the whole file.”
- Real on-chain anchoring. The desk’s pre-kickoff fair-value snapshot, and periodic checkpoints through the match, are anchored via real, confirmed Solana transactions.
- Verified source proofs. Every score and odds message is checked against TxLINE’s Merkle proof endpoints and validated on-chain before it’s allowed to influence a decision. A message that fails verification is never admitted.
- A/B proof, not assertion. The Strategy Arena runs the same feed through the same deterministic engine twice — once with every heuristic regime enabled, once fully neutralized — and grades both with the same CLV/Brier grader. The regime-ablation matrix goes further and isolates each regime individually.
What Tissue does not claim
To keep this documentation honest: Tissue never invents a fill, a counterparty, or PnL within TxLINE’s own on-chain program —txoracle has no intent-book or order-matching
instruction of its own, so live mode’s quote-publication book only ever publishes
risk-approved recommendations there. Simulated fills exist only in replay/research mode
and are always labeled simulated. Real execution does exist, on a separate venue
(Slip) — see Verifiability for how a decision turns into a real signed
transaction, gated by its own stricter, off-by-default risk check. See
Limitations for the full list of what depends on TxLINE capabilities
that don’t exist yet.
Where to go next
Quickstart
Run the desk against a real devnet fixture in a few commands.
Architecture
See how a TxLINE message becomes a signed, anchored decision.
Core Concepts
Fair value, the five regimes, and the discipline engine.
Verifiability
How to independently check any decision Tissue has made.
Demo video, GitHub repository, and live dashboard links are withheld while the project
is under competition embargo. They will be added here once publication is authorized.