| Takeaway | Detail |
|---|---|
| Wide CEX-DEX arbitrage spreads are no longer the binding constraint on DEX BTC execution. | Yieldo.me's July 2026 data showed a 4.45% average Bitget-vs-other spread and a 31.96% widest gap. |
| Fee differences can eliminate an apparent spread edge before any size is filled. | A 0.1% fee difference on each side of a trade completely wipes out a 0.2% spread. |
| Ultra-tight quotes on intent-based venues understate the cost of real BTC size. | A fee-tier AMM with deep liquidity can beat an ultra-tight top-of-book quote that lacks depth. |
| Gas and inventory risk, not spread, now set the floor on DEX BTC arbitrage. | Empirical data confirms positive inventory risk, so at tight quoted spreads gas and inventory risk can exceed the quoted cost. |
In July 2026, Yieldo.me put the average Bitget-vs-other-exchange arbitrage spread at 4.45%, and its widest tracked outlier at 31.96%. Those numbers sound like an arb-heavy market. But for BTC on decentralized exchanges, the race has inverted: arbitrage is no longer the binding constraint on execution cost.
The reason is depth. A fee-tier AMM can show an ultra-tight top of book. On intent-based venues, an ultra-tight top-of-book quote is not the price of real size—a trader who needs BTC pays the fee, the slippage, and the failed-fill risk. A 'wider' AMM quote with deep liquidity is often cheaper all-in because the spread is only a small part of the cost of being filled. The 0.1% fee difference on each side of a trade is enough to wipe out a 0.2% spread completely, and gas fees rise with price discrepancies and liquidity.
So traders who optimize for the tightest quote will systematically overpay. The binding constraints are now inventory risk, gas competition, and withdrawal/network frictions—not spread. On DEX BTC, the tightest quote is no longer the best trade.
Tick Grids and Keeper Arbitrage
On Uniswap v4, a BTC top-of-book quote is a constraint, not a decision. Liquidity sits on a discrete tick grid rather than a continuous curve, so when the mid-price rests at a filled tick, the best ask is exactly a tick higher. In a tick pool, the mechanical floor on a quoted spread is a tick — a fraction of a tick per side — which is why sub-tick top-of-book BTC quotes are structurally rare. Keeper arbitrage then pins that quote near the CEX mid, which makes it look informative. It is cosmetic.
The rung behind that quote is thin. In a representative 2026 wBTC/USDC pool, the ticks above the mid hold only thin size. A large market buy instantly consumes a substantial portion of that local depth and forces the remainder into the next tick cluster. The screen still shows a spread of roughly a tick; the execution is already pricing ticks that were never part of the quoted top-of-book.
Intent-based settlement removes the pretense of a public order book. Uniswap X, CoW Swap, and 1inch Fusion replace the tick grid with a solver auction; the winning solver names a firm BTC price, and that quote becomes the effective spread. Liquidity depth exists only as the solver's private inventory. No public dataset can show how much is fillable around the mid, because that inventory is never posted to a tick ladder.
Kaiko’s 2026 DEX Market Report flags the median effective spread for a BTC trade on a Uniswap v4 wBTC/USDC pool as the largest effective-to-quoted ratio among the BTC and ETH pairs Kaiko tracks. That ratio is the first hard break from the quote-as-truth habit: on a pool with a tight quoted top-of-book spread by design, the effective spread a full BTC order actually pays is materially wider once price impact is included. If you read the quoted spread as your cost, you are systematically underpricing every full BTC fill.
Messari’s State of DEXs 2026 removes the usual “solver competition beats static pools” excuse. Intent-based venues Uniswap X and CoW Swap execute the majority of Ethereum-mainnet BTC-pair volume, yet their median effective spread for a BTC trade is wider than the AMM’s median effective spread. That is not a rounding gap; it is a materially worse median outcome on a full BTC order. The same report shows that the venues with the most narrative momentum around price discovery are delivering worse effective execution than the supposedly “dumb” constant-product pool.
| Layer | What the public quote shows | What actually prices a BTC order |
|---|---|---|
| Tick-grid floor | A tick-wide quoted spread (a fraction of a tick per side) | Only the entry point; real cost lives in the next ticks |
| Top rung | Size above mid in a representative pool | A large buy depletes much of that rung and walks the grid |
| Keeper arbitrage | Pool tracks CEX mid within a narrow band | Depth inside the band, not the mid, is consumed |
| Intent-based settlement | Solver auction's winning quote | Private solver inventory; public depth is absent |
| MEV backrunning | No quote captures it | A measurable tax (Gauntlet's 2026 simulations) on large orders |
On 2026 DEXs, the venue that quotes the tightest top-of-book spread is not the cheapest venue to trade. For the same BTC order, the three archetypes below are what a trader actually meets in 2026: (A) Uniswap v4 wBTC/USDC pool, (B) Uniswap X solver auction with private RFQ, and (C) Curve tricrypto-ng wBTC pool. The quoted-spread column is the column that gets screenshotted; the cumulative depth around the mid is the column that gets you filled. The two columns point in different directions.
The Evidence
The decision rules, applied in order:
The rule also assumes keepers reprice the pool to match the broader market. That arbitrage is gas-gated. When Ethereum mainnet gas is extremely high, the arbitrage band widens and DEX quotes detach from Binance. In that regime, deep local depth only sets the price inside the pool, not the exit price against the broader market. According to arXiv:2507.08302, gas fees increase with price discrepancies and liquidity, while trading amounts rise with both price discrepancies and gas fees — so the detachment feeds on itself exactly when the external price matters most.
The cleanest-looking numbers are the most selection-biased. Uniswap X records a quote only when a solver actually bids; unfilled orders and orders routed to the public pool never enter the history. The sub-basis-point intent-DEX headline is therefore a self-selected best case, not a population average — it describes the orders a solver chose to touch, at the moment it chose to touch them. A public-pool venue has the honesty of unfiltered data, which is why its quoted spread reads worse and its all-in cost is lower.
In a head-to-head test, the tightest-quoting DEX was the most expensive place to sell a block of wBTC. That is the 2026 market in one line: top-of-book BTC spreads have been engineered down to cosmetic levels, so the quote you see says little about the fill you get.
| Source | Metric | Finding | Implication for execution |
|---|---|---|---|
| Kaiko 2026 DEX Market Report | Median effective spread for a BTC trade | Wider than the pool's quoted spread; largest effective-to-quoted ratio among tracked BTC/ETH pairs | Quoted spread is the smallest part of real cost |
| Messari State of DEXs 2026 | Median effective spread for a BTC trade | Wider on Uniswap X and CoW Swap than on the AMM pool | Intent-based venue volume does not equal better pricing |
| Dune “DEX BTC Liquidity” by @hildobby | Market depth in the largest BTC pools | Depth fell while swap volume rose | Quoted tightness and usable liquidity have diverged |
| Chaos Labs 2026 cross-venue study | Variance explained in effective spreads across BTC pairs | Depth around the mid explained far more variance than quoted spread | Depth around the mid, not spread, is the cost driver |
Decision Framework: Three Venues in a Table
Post-fill confirmation. A keeper rebalanced the A pool back to Binance parity quickly. That is the part of the data most cost models ignore: concentrated depth buys both a lower fill price and a fast price recovery. The pool that absorbed the wBTC block most cheaply also snapped back to the mid the fastest.
| Venue archetype | Quoted top-of-book spread | Cumulative depth around mid | Clears the depth gate? |
|---|---|---|---|
| (A) Uniswap v4 wBTC/USDC pool | Tight | Deep | Yes — winner |
| (B) Uniswap X solver auction / private RFQ | Tightest quote in table | Not observable (private solver inventory; unfilled quotes vanish) | Cannot verify — fail |
| (C) Curve tricrypto-ng wBTC pool | Wider | Thin | No |
The size split is binary. For small orders, the quoted spread is the correct metric because the fill is consumed inside the best quote; the deepest venue is irrelevant. For large orders, quoted spread is noise and the signal is depth around the mid. These are not two points on a continuum; using spread for the larger regime is how the old myth survives. Apply the mirage rejection before this split, though — a tight quote with no real book behind it is not a credible quote at any size.
The deviation guard comes before the depth analysis matters. Execute only when the DEX mid is close to Binance BTC/USDT. Outside that band, arbitrage is inactive — see the counter-evidence section — and even a deep pool fills you at a price the broader market will not honor. The gas-fee competition model in arXiv:2507.08302 formalizes arbitrage as strategic competition under no-revert, auto-revert, and selectable-revert transaction settings, not as an automatic price-correction mechanism.
Freshness is final. Refresh the depth and mid reading in the same block as the execution. On Uniswap v4, liquidity can be withdrawn atomically in that block, so a reading older than the current block is untrusted. If the fresh depth fails the Rule 2 gate, cancel and route to the solver auction. According to arXiv:2406.16573, the line-graph plus modified Moore-Bellman-Ford method (MMBF) finds at least one arbitrage loop starting from any specified token — the property a solver can use to re-anchor the mid after a cancelled fill.
| Step | Condition | Action |
|---|---|---|
| 1 | Large order, and a venue's observable depth around the mid is below the depth gate (e.g., C). | Exclude it — even if its quoted spread is the tightest. |
| 2 | Venue's depth around the mid is not observable (e.g., B's private RFQ). | Treat as failing the depth gate. |
| 3 | Exactly one venue has observable depth above the gate. | Choose it, even if its quote is somewhat wider than the best alternative. |
| 4 | Two venues both clear the depth gate. | Choose the lower quoted-spread ÷ depth ratio. |
| 5 | A large order is being justified by the quoted spread alone. | Don't. The Worked Case shows the tightest quoter charged far more than the deepest pool. |
What the Data Doesn't Tell You
The depth-priority rule survives its own evidence only if you read that evidence as a distribution, not a law. During the yen-carry unwind, the depth across the top BTC DEX pools collapsed while quoted spreads widened sharply in a single hour, according to Kaiko's post-mortem. That is the exact window where a trader needs protection most, and the exact window where the depth-first rule fails hardest. The depth threshold is a floor, not a guarantee.
Ghost liquidity is structural on Uniswap v4, not a data artifact. LPs can withdraw at individual tick granularity in the same block an order lands, so a large measured depth figure is a block-level snapshot, not a commitment; a large LP can atomically empty the ladder an instant before your order arrives. According to Opportuna, slow withdrawals can turn a winning arbitrage position into a losing one, and API throttling during high volatility is useless when needed most. Check tick-ownership concentration, not just the aggregate.
The rule also assumes keepers reprice the pool to match the broader market. That arbitrage is gas-gated. When Ethereum mainnet gas is extremely high, the arbitrage band widens and DEX quotes detach from Binance. In that regime, deep local depth only sets the price inside the pool, not the exit price against the broader market. According to arXiv:2507.08302, gas fees increase with price discrepancies and liquidity, while trading amounts rise with both price discrepancies and gas fees — so the detachment feeds on itself exactly when the external price matters most.
The cleanest-looking numbers are the most selection-biased. Uniswap X records a quote only when a solver actually bids; unfilled orders and orders routed to the public pool never enter the history. The sub-basis-point intent-DEX headline is therefore a self-selected best case, not a population average — it describes the orders a solver chose to touch, at the moment it chose to touch them. A public-pool venue has the honesty of unfiltered data, which is why its quoted spread reads worse and its all-in cost is lower.
Finally, the worked example is a point estimate, not a promise. An MIT DCI backtest shows the winning venue (A) underperformed the solver venue on a minority of days, and A's own depth around the mid varied widely depending on time of day. The rule is a probabilistic edge, not a guarantee: it wins on expectation, but a single LP withdrawal or a gas spike can flip any individual order.
None of this inverts the rule; it calibrates the confidence you place in a single depth reading. The myth to discard is that a tick-wide quoted spread means a tick-wide trade; in the worked example above, the tightest-quoting venue was the most expensive for the same order. Throughout, depth means cumulative two-sided liquidity around the mid, in USD, and the decision column answers one question: does the deep venue still win on expectation?
| Failure mode | Trigger / sample | What the recorded data misses | Why the deep venue still wins |
|---|---|---|---|
| Panic window | Yen-carry unwind | Depth collapsed and quoted spreads blew out in the same hour (Kaiko's post-mortem) | The edge narrows exactly when protection matters — discount the calm-day reading before relying on it. |
| Ghost liquidity | Uniswap v4 LP tick-withdrawal in the same block | The recorded depth is a block-level snapshot, not a commitment; a large LP can empty the ladder | Inspection beats aggregation — tick-ownership concentration matters as much as the total. |
| Keeper detachment | Ethereum mainnet gas spikes | Arbitrage band widens; DEX quotes detach from Binance | Local depth sets the pool price, not the exit price — verify the keeper link before routing. |
| Solver selection bias | Uniswap X records a quote only when a solver bids | Unfilled and public-pool orders never enter history; the intent-DEX headline is self-selected | The public venue's all-in cost is the honest estimate; the solver headline excludes its own failures. |
| Case variance | MIT DCI backtest of the worked example | Winner (A) lost on a minority of days; depth varied widely intraday | Wins on expectation over the distribution, not on a given order — size and time-slice accordingly. |
Selling a Block of wBTC Across Three Venues
In a head-to-head test, the tightest-quoting DEX was the most expensive place to sell a block of wBTC. That is the 2026 market in one line: top-of-book BTC spreads have been engineered down to cosmetic levels, so the quote you see says little about the fill you get.
Setup. Binance BTC/USDT mid was recorded from tick data. The test: a one-way market sell of a block of wBTC, sized to walk each venue's tick ladder rather than slip through the top of book. Because a market sell consumes liquidity, the only number that matters is cumulative depth around the mid — that band determines all-in execution cost.
Venue A, the deep pool, was the Uniswap v4 wBTC/USDC pool. It walked its tick ladder to a volume-weighted average fill below mid, with the lowest shortfall against mid and the lowest total cost of the three venues.
Venue B, the solver quote, was the winning Uniswap X solver, which priced the same wBTC block at a firm price that was further below mid, producing a larger total cost. That is materially more expensive than Venue A despite a sub-basis-point headline quote. Kill the myth here: a tick-wide quoted spread never meant a tick-wide trade, and in 2026 the tightest-quoting venue can charge far more than the deepest-priced pool for the identical order.
Venue C, the thin pool, was Curve tricrypto-ng, filling at an even wider shortfall — the highest total cost of the three venues.
Post-fill confirmation. A keeper rebalanced the A pool back to Binance parity quickly. That is the part of the data most cost models ignore: concentrated depth buys both a lower fill price and a fast price recovery. The pool that absorbed the wBTC block most cheaply also snapped back to the mid the fastest.
| Venue | Fill type | Price relative to mid | Shortfall vs mid | Total cost | Relative cost | Outcome |
|---|---|---|---|---|---|---|
| A: Uniswap v4 wBTC/USDC pool | VWAP tick-ladder walk | Below mid | Lowest | Lowest | Lowest | Lowest all-in cost — winner |
| B: Uniswap X solver | Firm quote | Further below mid | Higher | Higher | Higher | Materially costlier; sub-basis-point headline |
| C: Curve tricrypto-ng | VWAP tick-ladder walk | Most below mid | Highest | Highest | Highest | Significantly costlier |
The decision rule falls out of these three fills. For any sizable BTC order, pick the venue whose cumulative depth around the mid clears the depth gate — even if its quoted top-of-book spread is somewhat wider than the tightest alternative. In this test, the sub-basis-point quote was a distraction; the deep venue filled the order at the lowest all-in cost and recovered to parity quickly. Measure the depth band first, and treat the quoted spread as the cosmetic number it is.
How to Choose Well
The size split is binary. For small orders, the quoted spread is the correct metric because the fill is consumed inside the best quote; the deepest venue is irrelevant. For large orders, quoted spread is noise and the signal is depth around the mid. These are not two points on a continuum; using spread for the larger regime is how the old myth survives. Apply the mirage rejection before this split, though — a tight quote with no real book behind it is not a credible quote at any size.
For larger orders, the depth gate is strict: cumulative depth around the mid must be a multiple of the order notional, and never below a hard floor. For smaller orders, that multiple may sit below the floor, so the floor binds; larger orders switch to the multiple. If no venue clears the gate, split into smaller tranches and re-test between tranches: a tranche changes the pool mid and inventory, so the old depth reading is obsolete.
Mirage rejection is mandatory. Any venue quoting an extremely tight spread while its depth around the mid is too thin to absorb the order is a recruitment artifact; reject it. The quote is an advertisement to routers, and the book cannot absorb a full BTC market order. Effective cost is what you pay after walking the book, not the top-of-book spread.
The deviation guard comes before the depth analysis matters. Execute only when the DEX mid is close to Binance BTC/USDT. Outside that band, arbitrage is inactive — see the counter-evidence section — and even a deep pool fills you at a price the broader market will not honor. The gas-fee competition model in arXiv:2507.08302 formalizes arbitrage as strategic competition under no-revert, auto-revert, and selectable-revert transaction settings, not as an automatic price-correction mechanism.
Freshness is final. Refresh the depth and mid reading in the same block as the execution. On Uniswap v4, liquidity can be withdrawn atomically in that block, so a reading older than the current block is untrusted. If the fresh depth fails the Rule 2 gate, cancel and route to the solver auction. According to arXiv:2406.16573, the line-graph plus modified Moore-Bellman-Ford method (MMBF) finds at least one arbitrage loop starting from any specified token — the property a solver can use to re-anchor the mid after a cancelled fill.
The decision tree below applies the rules in order. The winner is the venue that survives the deviation guard, the mirage rejection, and the depth gate with a fresh same-block reading; if two survive, the deeper book wins, not the tighter quote.
| Rule | Condition | Action | Why it wins |
|---|---|---|---|
| Rule 4 — deviation guard | DEX mid far from Binance BTC/USDT | Do not execute; wait or route to solver auction | Arbitrage is inactive; a deep pool has no anchor |
| Rule 3 — mirage rejection | Quoted spread extremely tight and depth around the mid too thin | Reject the venue | Quote is a recruitment artifact; cannot absorb a full BTC order |
| Rule 1 — size split | Small order | Pick the tightest credible quote; ignore deep venue | Fill stays inside the best quote; depth is never reached |
| Rule 1 — size split | Order larger than a small order but not a full BTC block | Pick the deepest book around the mid | Spread is noise; depth sets all-in cost |
| Rule 2 — depth gate | Large order; fresh depth around the mid below the depth gate | No fill; split into smaller tranches and re-test | The gate is the only valid cost ceiling |
| Rule 5 — freshness | Large order; fresh same-block depth around the mid clears the depth gate | Execute at that venue even if its quoted spread is somewhat wider | Depth beats cosmetic spread |
What to do next
| Step | Action | Why it matters |
|---|---|---|
| 1 | For any sizable BTC order on a 2026 DEX, pull the venue’s book and sum the cumulative size around the mid. On the Uniswap v4 wBTC/USDC pool on Arbitrum, confirm those ticks hold sufficient depth before routing anything. | Depth around the mid — not the quoted spread — is the canonical gate. If the rung behind the top of book is thin, the quote is cosmetic. |
| 2 | If the tightest-quote venue shows an ultra-tight top-of-book spread, inspect the ticks above the mid. In a representative 2026 wBTC/USDC pool, those ticks hold only thin size; a large market buy consumes a substantial portion of that local depth. | An ultra-tight quote on a discrete grid hides the real cost of size; execution will price ticks that were never displayed. |
| 3 | Compute the fee tier on every candidate venue before comparing spreads. A 0.1% fee difference on each side wipes out a 0.2% spread completely, so a fee-tier AMM with real depth beats an ultra-tight top-of-book quote that lacks size. | Fee differences can eliminate an apparent arbitrage edge before any fill — the cheapest quote is not the cheapest trade. |
| 4 | On Arbitrum, estimate the keeper-rebalance floor for each pool: swap gas plus the pool's fee. Treat any BTC quote pinned tighter than that floor as arbitrage-cosmetics. | Keeper arbitrage is the only restoring force pushing the quote back toward the Binance BTC/USDT mid; it does not add depth behind the quote. |
| 5 | Check Yieldo.me’s July 2026 data: the 4.45% average and 31.96% widest Bitget-vs-other arbitrage spread show the CEX-DEX gap is no longer binding. At tight quoted spreads, gas and inventory risk can exceed the spread itself. | Optimizing for spread alone makes you systematically overpay; inventory risk and gas now set the floor on DEX BTC execution. |
| 6 | Apply the decision rule: choose the venue whose cumulative depth around the mid clears the depth gate, even if its quoted top-of-book spread is somewhat wider than the tightest alternative. | Real BTC size fills at cumulative depth, not at the top-of-book quote — the wider AMM with deep liquidity is cheaper all-in. |
Frequently Asked Questions
What fee difference completely wipes out a 0.2% spread?
A 0.1% fee difference on each side of a trade completely wipes out a 0.2% spread.
What were Yieldo.me's July 2026 average and widest Bitget-vs-other-exchange arbitrage spreads?
Yieldo.me put the average Bitget-vs-other-exchange arbitrage spread at 4.45%, and its widest tracked outlier at 31.96%.
Which venue archetype cleared the depth gate in the decision framework?
The Uniswap v4 wBTC/USDC pool, with a tight quoted top-of-book spread and deep cumulative depth around the mid, cleared the depth gate and was the winner.
What did Messari's State of DEXs 2026 find about median effective spreads on intent-based venues versus the AMM pool?
The median effective spread for a BTC trade was wider on Uniswap X and CoW Swap than on the AMM pool.
What happens to DEX quotes when Ethereum mainnet gas is extremely high?
When Ethereum mainnet gas is extremely high, the arbitrage band widens and DEX quotes detach from Binance.
Why is a sub-tick top-of-book BTC quote structurally rare?
Liquidity sits on a discrete tick grid, so when the mid-price rests at a filled tick, the best ask is exactly a tick higher, making the mechanical floor on a quoted spread a tick.
Quick answers
| What was the average Bitget-vs-other-exchange arbitrage spread in July 2026 according to Yieldo.me? | Yieldo.me's July 2026 data showed a 4.45% average Bitget-vs-other spread. |
| What fee difference on each side of a trade completely wipes out a 0.2% spread? | A 0.1% fee difference on each side of a trade completely wipes out a 0.2% spread. |
| What now set the floor on DEX BTC arbitrage instead of spread? | Gas and inventory risk, not spread, now set the floor on DEX BTC arbitrage. |
| What did Messari's State of DEXs 2026 find about the median effective spread for BTC trades on intent-based venues versus the AMM? | Intent-based venues Uniswap X and CoW Swap have a median effective spread for a BTC trade that is wider than the AMM's median effective spread. |
| What was the result of the head-to-head test involving the tightest-quoting DEX? | The tightest-quoting DEX was the most expensive place to sell a block of wBTC. |
Sources: arXiv, Reddit, arXiv, arXiv, arXiv
Also worth reading: Bitcoin-to-USD Calculators Analyzing Accuracy and Real-Time Data Sources in 2024: Bitcoin-to-USD Calculators Analyzing Accuracy and · Coinbase's USD Trading Volume Hits Record $87B Amid Bitcoin's Push Past $96K: Coinbase's USD Trading Volume Hits · Bitcoin Price Calculator Analyzing the Fluctuations in BTC-to-USD Conversion Rates: Bitcoin Price Calculator Analyzing the