explains

On-Chain Settlement Finality

In trad-fi, settlement finality is a legal moment. On-chain, it is probabilistic - defined by economic guarantees, not legal ones.

By Anton Titov, Founder · Plexo Institute

Published 2026-05-17

For treasurers choosing settlement assets, finality is the most misunderstood property of stablecoin clearing because blockchains replace legal certainty with economic thresholds.

Reader Brief

In trad-fi, settlement finality is a legal moment. On-chain, it is probabilistic - defined by economic guarantees, not legal ones.

What's Inside

Four concepts that connect legal settlement finality, chain security budgets, reorg risk, and clearing design.

Trad-fi finality - the legal moment that does not exist on-chain

Four settlement regimes, RTGS, CLS, DNS, and SFD, share one property: a legally defined, statute-backed moment of irrevocability backed by central bank money. This is the baseline blockchain settlement must match.

On-chain finality - economic guarantees replace legal ones, and the cost curve is value-dependent

Reversal difficulty scales with confirmations and consensus design: Ethereum finalized checkpoints require corrupting/slashing a large validator share; Bitcoin six-confirmation settlement requires sustained majority hash power. The threshold is an operator policy decision, not a universal constant.

Reorg risk - not theoretical, ETC saw multi-thousand-block reorganizations in 2020

Probabilistic chains can reorganize when an alternative valid history becomes canonical. BFT-style finality changes the failure mode: reversal requires violating validator or slashing assumptions, not merely waiting for a longer branch.

Clearing design - three architecture patterns, chosen per corridor

Wait-for-confirmations, simple and slow; network-level guarantee, fast with credit risk; or atomic PvP, complex with no one-sided exposure. Corridor characteristics determine the choice.

What Finality Means in Trad-Fi

Every trad-fin system defines a specific legal moment when a transfer becomes irreversible.

Fedwire originated an average daily transfer value of about $4.59T in 2025. CLS reported $2.49T in average daily traded volume submitted to CLS in June 2025. Both are institutional settlement systems with legally defined finality, not probabilistic confirmation thresholds [3][4][5][7][8].

Four regimes, one principle: finality is a legally defined moment backed by central bank money or its equivalent

**RTGS, including Fedwire, TARGET2, and CHAPS:** real-time gross settlement in central bank money. Finality at the moment of central bank ledger update. Legal basis: Settlement Finality Directive in EU, Federal Reserve Operating Circular 6 in US. **CLS, FX:** payment-versus-payment finality. Both legs of an FX trade settle simultaneously or neither does. Finality at the moment CLS confirms both pay-ins. **DNS, including ACH and SEPA:** deferred net settlement. Provisional during the day, final at end-of-day RTGS settlement window. **CLS-style legal finality plus zero-hour rule abolition:** Settlement Finality Directive, EU 98/26/EC, explicitly removed the historical zero hour rule for system payments to ensure intraday irrevocability. The common feature: finality is a legally defined moment, supported by central bank money or equivalent, with explicit statutory protection.

What Finality Means On-Chain

On a public blockchain, finality is not a legal concept. It is an economic property.

A transaction is final when reversing it would cost more than any rational attacker would pay. Certainty grows with each block, but 100% certainty is asymptotic. Blockchain gives a cost curve, and each operator draws its own line on it.

Finality threshold diagram with rising reversal-cost bars, an operator release line, policy inputs, and downstream release outcome. — Finality is an operating threshold: confirmations matter because they raise reversal cost above the payment value and risk the operator is willing to release.

~12.8 min Ethereum economic finality, about 64 blocks at PoS [1]
~60 sec Tron operational threshold at about 20 blocks [2]

No legal moment exists: certainty grows with each block, but 100% is asymptotic.

Reversal difficulty scales with confirmations. On Ethereum PoS, reverting a finalized checkpoint requires corrupting or slashing more than one-third of the validator set [1]. On Bitcoin, a six-confirmation reorg requires sustained majority hash power long enough to replace the accepted chain. The threshold is value-dependent, but it should be expressed as policy rather than a universal number. Low-value retail transfers can tolerate shorter confirmation windows; high-value institutional settlement should wait for deterministic finality where available or for a stronger probabilistic threshold. This creates a sliding scale absent in trad-fi. Fedwire gives a binary legal moment: final or not. Blockchain gives a cost curve - and each operator draws its own line on it.

Probabilistic vs Deterministic Finality

Not all blockchains use the same finality model, and the model changes clearing-network design.

Two architecturally distinct approaches exist: probabilistic accumulation of confirmations and deterministic/BFT commitment. Hybrid systems combine a fast probabilistic head with slower deterministic finality.

Model	How finality works	Examples	Implication for clearing
Probabilistic	Confirmations accumulate; certainty grows asymptotically	Bitcoin, classic Ethereum PoW, Tron, Litecoin	Operators define internal threshold, such as 12 blocks
Deterministic / BFT	Validators commit blocks via Byzantine consensus; final on commit	Ethereum PoS, Casper FFG; Cosmos chains; Solana; Algorand	Single block confirmation equals final
Hybrid	Probabilistic head plus deterministic finality after N blocks	Ethereum, where Casper FFG finalizes about every 64 blocks	Either fast head or slower deterministic; pick

Ethereum has two finalities: head at 12 seconds, probabilistic, and Casper FFG at 12.8 minutes, deterministic.

Ethereum post-Merge runs two finality layers simultaneously. **Head, 12 sec:** the latest block. Probabilistic - a short reorg can happen before checkpoint finalization. **Finalized, about 12.8 min:** the most recent checkpoint justified and finalized through Casper FFG. Reverting it requires violating the protocol security assumptions and slashing more than one-third of validators [1]. The gap is not academic: head-of-chain blocks can reorg before finality. For institutional settlement, waiting for finalized checkpoints materially reduces reorg exposure compared with accepting the latest head.

Chain-by-Chain Comparison

For a clearing network operator, the practical question is which threshold to use.

Each chain has a different block time, security model, and operational threshold. A single confirmation policy across all chains is structurally wrong.

Chain	Block time	Operational threshold	Time to threshold
Ethereum (head)	12s	12-32 blocks	~2.4 - 6.4 min
Ethereum (Casper finalized)	12s	1 epoch	~12.8 min
Tron	3s	20 blocks	~60 sec
Solana	0.4s	32+ blocks or super-majority confirmation	~13 sec
Polygon PoS	2s	~256 blocks	~8.5 min
Arbitrum / Optimism	variable	L1 finalization, Ethereum	~12.8 min + dispute window
Bitcoin	10 min	6 confirmations	~60 min

Security budget determines threshold: Bitcoin, Ethereum, and Tron have structurally different models.

Bitcoin and Ethereum have structurally different security budgets, but both make large reversals expensive in different ways: Bitcoin through sustained majority hash power, Ethereum through validator corruption and slashing. Tron is structurally different: 27 super-representatives, elected, control block production. A 20-block threshold is conventional, but the security model depends on the super-representative set rather than on Bitcoin-style industrial hash-rate competition [2]. Smaller PoW chains illustrate the gap: Ethereum Classic experienced 3,600+ block and later multi-thousand-block reorganizations during 2020 attacks [9]. Security budget is not a label - it is a chain-specific risk model.

Reorg Risk and Operational Thresholds

The gap between probabilistic and deterministic finality is not theoretical; it has been exploited.

Ethereum Classic experienced multi-thousand-block reorganizations during 2020 51% attacks, and exchanges responded by suspending or reviewing deposits. That history is the reason probabilistic confirmation thresholds cannot be copied mechanically across chains [9].

ETC multi-thousand-block reorgs, BCH fork-era reorgs, and short head reorgs show why thresholds are operational policy.

**Ethereum Classic 2020:** documented 3,600+ block and later multi-thousand-block reorganizations during 51% attacks. Exchanges reviewed or suspended ETC deposits while the network stabilized [9]. **Bitcoin Cash 2018-2020:** fork-era reorgs demonstrated that minority-chain and split-community dynamics can affect operational settlement assumptions. **Ethereum head reorgs:** short head reorgs can happen before Casper finality; the institutional question is whether a workflow waits for finalized checkpoints or accepts head risk. The operational lesson is narrower than "all chains are risky." Probabilistic-only confirmation needs a chain-specific threshold. Deterministic or BFT-style finality changes the reversal condition and can support a different settlement policy.

Three inputs set the threshold: transaction value, counterparty risk profile, and chain security model.

**Transaction value:** higher-value transfers should wait for stronger finality thresholds, because the attacker incentive and operational loss both rise with value. **Counterparty risk profile:** regulated FI with bilateral netting agreement can support a lower threshold if reorg loss is recoverable via legal claim. Unknown or non-regulated counterparty requires a stricter threshold and no credit extension during the confirmation window. **Chain security model:** Ethereum Casper finality, Tron delegated super-representatives, Bitcoin proof-of-work confirmations, and L2 dispute windows are different models. Each has a different failure mode: Ethereum = validator corruption/slashing; Tron = elected-validator collusion or governance failure; Bitcoin = sustained hash-rate majority; optimistic L2s = L1 finality plus withdrawal/dispute mechanics. Pragmatic defaults should be calibrated per corridor and reviewed quarterly: deterministic where available, probabilistic with explicit thresholds where not.

What This Means for Clearing Network Design

Evidence And Sources

This raw HTML export preserves source visibility for crawler and contractor review. Indexing decision: index, follow.