# Finality The question this note answers: when can I treat a block as irreversible? Notice that this is a different question from "is it in the chain?" — chains reorganize, so mere inclusion doesn't settle anything. What you actually want to know is: what would it cost, and who would have to misbehave, for this block to disappear? Different systems give very different answers, and it's worth knowing which one you're getting. ## The types | **Type** | **Meaning** | |---|---| | Probabilistic finality | Reversal gets less likely over time | | Economic finality | Reversal requires slashable economic loss | | Deterministic finality | Protocol commits once a quorum signs | | Social finality | Humans decide after catastrophic failure | Bitcoin offers probabilistic finality — hence the folklore about "waiting for 6 confirmations." Each block built on top makes reversal exponentially more expensive, but there's no moment where it becomes strictly impossible. You're choosing a risk tolerance, not receiving a guarantee. Ethereum proof-of-stake offers economic finality: a finalized block can only revert if a large fraction of validator stake knowingly violates the protocol and gets slashed for it. The guarantee comes with a price tag attached, and the price is public. BFT systems (Tendermint-style) offer deterministic finality: once two-thirds of validators sign, the block is committed, full stop. It's faster and cleaner — the tradeoff is that the validator set is usually smaller and more permissioned. And underneath all of these sits a social layer that's easy to forget. Even "code is law" systems need humans to choose which software version to run, and when something breaks badly enough, people decide. Every chain has this layer; the honest ones admit it. ## Finality is layered Here's where it gets practically interesting. On a rollup, a single transaction passes through several finality stages, each with a different trust assumption: | **Stage** | **Meaning** | |---|---| | Seen by sequencer | Fast but weak | | Included in L2 block | Better | | Batch posted to L1 | Stronger | | Challenge window passed / proof verified | Strongest | | L1 finalized | Stronger still | So when someone says "confirmed," a useful habit is to ask: confirmed where, by whom, and under what trust assumption? That question decides real product behavior — when an exchange credits a deposit, when a bridge releases funds, when a wallet shows the checkmark. ## Where this matters - [[04 Consensus]] — how each consensus family produces its flavor of finality - [[09 Scaling - Rollups and Data Availability]] — finality across layers - [[Reorgs]] — what happens before finality arrives - [[Trust Assumptions]] — finality is always a claim about who you trust