In traditional fragmented markets, prices across venues converge quickly. If Nasdaq shows ETH at $3,500.10 and a competing venue shows $3,500.25, arbitrage bots close the gap in milliseconds. The premise is that multiple venues share a common pool of limit orders and fast participants.
In Uniswap, this assumption breaks down. Price divergence across fee-tier pools can persist for minutes — sometimes much longer. Understanding why isn’t just trivia; it’s how you read the AMM DOM correctly.
Why AMM pools diverge
Each Uniswap fee-tier pool is an independent contract. There is no shared order book, no central matching engine, and no automatic price synchronization between pools. Each pool updates its price only when a swap actually executes against it.
Three things cause persistent divergence:
1. Slot0 staleness
When the backend first subscribes to a pool, it reads the current price from the contract’s slot0 storage — a single on-chain value that holds the most recent sqrtPriceX96. For actively traded pools like the 0.01% WETH/USDC pool, slot0 is refreshed every few seconds by real swap flow. For the 0.30% pool, which may see only a few trades per hour, slot0 could be hours old.
Until the first live swap arrives in a pool, the price shown is the slot0 snapshot price — real but potentially stale. The Age column in the fee tier table tracks this directly: a — means no live swap has been observed, only the initial snapshot.
2. Unequal swap flow across tiers
Not all fee tiers attract equal flow. On most pairs, the 0.01% and 0.05% pools receive the bulk of volume because they have lower cost. The 0.30% pool trades infrequently. When a large directional move happens, the active pools update immediately via swap events; the inactive pool’s price lags until a trader routes through it or an arbitrageur takes the gap.
3. LP composition differences
Each pool attracts different LPs with different positioning strategies. If the 0.01% pool has recently experienced a large mint at current tick, that LP may have pushed the price slightly during their provision. The 0.30% pool, with its different LP composition, won’t reflect that shift until a swap propagates it.
What Δ Mid measures
The mackinac fee tier table shows a Δ Mid column: the difference between each pool’s mid price and the most recently active pool’s mid price, in basis points.
The reference pool is always whichever pool had a swap most recently. All other pools are measured against it.
| Δ Mid | Interpretation |
|---|---|
ref | This pool is the freshest — the current benchmark |
+8.3 bp | This pool’s mid is 8.3 bp above the reference pool |
−3.1 bp | This pool’s mid is 3.1 bp below the reference pool |
A small, transient Δ Mid — a few basis points that flickers and resolves — is normal. Swap routing naturally equalizes prices as flow moves through the most efficient pool.
A persistent Δ Mid is different. If a pool consistently shows +10 bp or −15 bp relative to the reference, this means either:
- The pool is truly stale (low volume, no recent swaps), in which case the Age column will confirm it
- LP composition differences are creating a structural price bias
- A recent large trade in one pool hasn’t been equalized yet because the spread isn’t wide enough to attract arbitrage after fees
How fast divergence resolves
Resolution speed depends on the divergence size relative to the fee cost of closing it.
A 3 bp divergence between the 0.01% and 0.05% pools won’t attract arbitrage — the fee on a round trip through both pools exceeds the gap. The divergence persists until organic swap flow happens to equalize them.
A 20 bp divergence on the same pair will attract fast arbitrage, typically closing within seconds of appearing. The gap is wide enough to cover fees and gas on Arbitrum (usually under $0.50 per transaction).
The 0.30% pool is almost never arbitraged against the tighter pools except when the gap exceeds roughly 30–40 bp, because the fee cost of routing through the 0.30% pool is high.
What persistent divergence tells you
When you see a persistent non-zero Δ Mid on a pool that has recent swap activity (Age is current), it’s a real signal:
- Price fragmentation: swap routing is not equalizing the pools, meaning the gap isn’t wide enough to attract arbitrage given fee costs
- Directional flow in one pool: a sustained positive Δ Mid on the 0.01% pool relative to the 0.05% pool suggests buyers are preferring the tighter pool and moving its price up faster
- LP composition divergence: different LP positioning between tiers can create persistent biases, especially around liquidity range boundaries
For a trader monitoring AMM mechanics, Δ Mid is one of the cleaner signals of where real price discovery is happening — and which pools are being left behind.