DEX arbitrage sounds trivial: buy the same asset where it is cheap and sell where it is expensive.
In major DEX markets, however, the moment a human notices a visible mismatch is often already too late. Bots continuously scan for out-of-sync pools. Uniswap itself explains that price divergence creates arbitrage opportunities and that arbitrage bots constantly search for them on-chain.[1]
So is DEX arbitrage simply over for anyone without institutional infrastructure?
Not quite.
The central question is not whether a spread exists. It is why the spread exists, how long it survives, and who is structurally able to capture it.
1. Drop the shortcut “DEX means unregulated”
In Japan, the word DEX alone does not determine whether an activity is inside or outside financial regulation.
Japan’s Financial Services Agency has said that DEXs may require rules tailored to their technical characteristics rather than simply copying the framework for conventional exchanges.[2]
In a March 2026 FinTech proof-of-concept review, the FSA also stated that developing and deploying a DEX protocol may, depending on the facts, fall within crypto-asset exchange business because it can enable users to exchange crypto-assets.[3]
A fully immutable autonomous contract, a protocol whose developers retain control, and a company-operated interface that controls listings, routing and fees should therefore not be treated as legally identical.
FATF’s 2026 DeFi work similarly focuses on whether a person has “control or sufficient influence” when identifying a VASP in a DeFi ecosystem.[4]
Self-custody P2P is another layer. FSA material explains that direct transfers between unhosted wallets sit outside the VASP-to-VASP Travel Rule because no VASP participates. But regulated businesses interacting with unhosted wallets can still face AML/CFT risk-management and information-collection obligations.[5]
“Outside the Travel Rule” does not mean “outside every law.”
2. Keep regulation and tax in separate boxes
Japan enacted a major crypto-market reform on July 15, 2026, moving crypto transaction regulation into the Financial Instruments and Exchange Act framework. The main provisions are generally to take effect on a date set by Cabinet Order within one year of promulgation; some provisions, including stronger penalties for unregistered business, took effect on August 12, 2026.[6][7]
Tax reform is linked but separate.
The FY2026 tax reform creates a 20% separate-taxation framework for qualifying transfers of registered “specified crypto-assets” to a crypto-asset transaction business operator. Its starting date is linked to the financial-law reform and eligibility depends on the asset, counterparty and transaction structure.[8]
That does not justify shortcuts such as:
“DEX always means outside the 20% regime.”
“DEX means tax-free.”
“P2P or offshore means regulation no longer matters.”
Business regulation, AML/CFT, income-tax treatment and eligibility for a special tax regime are different tests.
3. Arbitrage does not require an uninformed victim
This is the most interesting market-structure point.
A spread can arise even when nobody is confused about the price.
A large buy order hitting an AMM moves the pool price along its curve. If another venue barely moves, a discrepancy appears.
The trader who caused it may be perfectly rational. They may simply value immediate execution enough to pay market impact.
A 2026 study of atomic arbitrage on Polygon analyzed more than 360,000 arbitrages and found that in 96.7% of cases most of the opportunity could be attributed to essentially one preceding transaction.[9]
The common pattern is:
large trade → AMM price moves → cross-venue discrepancy → arbitrage restores alignment
Arbitrageurs are therefore not necessarily harvesting stupidity. They are often being paid for immediacy, liquidity provision and reconnecting fragmented prices.
4. But the obvious atomic game really is brutal
The next mistake is to conclude that a simple spread detector is enough.
A study of Ethereum CEX–DEX arbitrage from August 2023 through March 2025 identified 7,203,560 arbitrages by 19 major searchers and estimated $233.8 million in extracted value.[10]
The market is real.
It is also concentrated: the top three searchers captured roughly 75% of both volume and extracted value. Profitability was also related to integration with block builders.[10]
This is not a game where you visually see:
Venue A: 100.00
Venue B: 100.05
and click faster.
The real competition includes:
- mempool or private-order-flow access
- builder relationships
- gas bidding
- inclusion probability
- transaction ordering
- revert protection
- CEX hedge latency
- pre-positioned inventory
A five-basis-point displayed spread is not a five-basis-point profit.
5. For individual research, ask why a spread survives
Instead of racing for discrepancies that disappear in 100 milliseconds, investigate discrepancies that survive for 30 seconds, one minute or three minutes.
Cross-chain arbitrage is a useful example.
A 2025 study across nine blockchains identified 260,808 cross-chain arbitrages between September 2023 and August 2024. 32.37% involved bridges, with a lower-bound profit estimate of about $9.50 million on roughly $465.8 million of volume.[11]
Within one atomic domain, an unprofitable sequence may sometimes revert as a unit.
Cross-chain execution cannot rely on that property. It can suffer from:
- one leg filling first
- bridge latency
- finality delays
- price movement during transfer
- insufficient liquidity on one chain
- inventory imbalance
The spread can survive because removing it requires capital and risk.
A persistent spread does not necessarily mean nobody noticed it.
It may mean closing the spread has a cost.
6. Pre-position inventory instead of waiting for a bridge
One approach is to hold inventory on both chains before an opportunity appears.
Suppose Chain A and Chain B both hold stablecoins and the target asset.
If the asset becomes temporarily expensive on B:
buy on A
sell on B
at nearly the same time.
This reduces the risk that the spread disappears while waiting for a bridge.
But the cost has merely changed form. You now carry:
- duplicated idle capital
- inventory imbalance
- rebalancing cost
- chain outage risk
- smart-contract risk
- bridge risk
- stablecoin depeg risk
- RPC and sequencer risk
That is the key.
When fast bots do not eliminate an opportunity, there is usually a reason fast bots dislike it.
Your edge exists only if you can process that reason more cheaply than competitors.
7. Measure executable net edge, not a screen spread
Comparing displayed mid-prices is one of the easiest ways to invent fake arbitrage.
The real object is executable price.
Conceptually:
net edge
= executable spread
− swap fee
− gas
− price impact
− failed-transaction cost
− hedge cost
− bridge/rebalancing cost
− inventory risk premium
− adverse selection
A displayed 40 bp opportunity may become:
5 bp fee
6 bp price impact
2 bp gas
4 bp hedge
8 bp rebalance
10 bp inventory risk
leaving only 5 bp.
But if 15–20 bp remains after realistic costs and survives for 30 seconds or more, it becomes research-worthy.
The useful statistic is therefore not “largest observed spread.”
Build a spread survival curve:
for opportunities with at least 15 bp of net edge, how much remains after:
- 1 second
- 5 seconds
- 30 seconds
- 1 minute
- 3 minutes
A personal strategy should not depend on a market where every opportunity is gone by second one.
8. Measure who wins
Searcher concentration is another filter.
If the same address or builder path captures almost every opportunity, a visible spread may not represent accessible alpha.
More promising markets are those where:
- winners are distributed
- opportunities survive multiple blocks
- dedicated builder relationships are unnecessary
- larger size does not instantly destroy net edge
- failed-transaction rates remain manageable
“There is profit in this market” and “I can capture that profit” are different statements.
9. Choose the friction, not merely the competitor
Different forms of arbitrage are different sports.
Major Ethereum DEX-to-DEX atomic arbitrage Extreme latency, builder and order-flow competition.
CEX–DEX arbitrage
Large market, but highly concentrated and dependent on both CEX hedging and block inclusion.
Smaller DEXs within an L2 Potential opportunities, but RPC, sequencer and liquidity behavior must be tested.
L2-to-L2 or cross-chain inventory arbitrage Non-atomic risk and capital lock-up become barriers to entry, leaving room for optimization beyond pure speed.
Stablecoin arbitrage
Many small discrepancies, but depeg, redemption, liquidity and inventory risks must be separated.
New or thin pools
Large-looking spreads can hide honeypots, transfer taxes, malicious token logic, disappearing liquidity and bad price data.
Weak competition is rarely free.
Competition is often weak because the job is annoying.
If that annoyance can be automated, it may become the actual opportunity.
10. A better research sequence
There is no need to start with real-money execution.
A disciplined process is:
- Reconstruct executable quotes for the same asset across DEXs and chains.
- Reproduce gas, fees and price impact.
- Attribute each spread to the transaction or event that created it.
- Build 1-second to 3-minute spread survival curves.
- Measure win rates and concentration by searcher address.
- Model pre-positioned inventory and realistic rebalancing.
- Stress-test partial fills, failed transactions, RPC latency and sequencer outages.
- Separate fit and untouched holdout periods.
- Prefer broad parameter plateaus over one magical setting.
- Reject the strategy if nothing remains after all costs.
Evaluate more than win rate:
- net P&L
- net edge per trade
- capital efficiency
- inventory turnover
- maximum inventory imbalance
- failed execution rate
- spread half-life
- searcher concentration
- contribution by chain and pool
- performance excluding the single best anomaly
Conclusion: search for friction, not fools
DEX arbitrage is not merely a game that pays when beginners are present.
Large orders, demand for immediacy, AMM curves, fragmented liquidity, non-atomic cross-chain execution and inventory constraints can create spreads even when participants are sophisticated.
The problem is that major instantaneous opportunities are often captured by highly specialized searchers.
So the useful research question is not:
“Where is there a price difference?”
It is:
“Which price differences remain profitable after all costs for at least 30 seconds and are still accessible without privileged infrastructure?”
If competing over 0.1 seconds is hopeless, stop making 0.1 seconds the game.
Do not search for suckers.
Search for the price the market is still paying someone to absorb its friction.
References (11)
- Uniswap Labs, “Why is a pool out of sync?
- 金融庁, 金融審議会「暗号資産制度に関するワーキング・グループ」（第6回）議事録, 2025-11-26. DEXについて技術的性質に合わせた規制を継続検討。
- 金融庁, 「FinTech実証実験ハブ」支援決定案件の実験結果について, 2026-03-13. DEXプロトコルの開発・設置は一定の場合に暗号資産交換業に該当する余地があるとの整理。
- 金融庁, FATFによる「DeFiに係る規制上の課題に関する報告書」の公表について, 2026-07-29
- 金融庁, 「暗号資産・電子決済手段の移転に係る通知義務（トラベルルール）」およびAML/CFT関連資料。アンホステッド・ウォレット間P2PとVASP関与取引を区別。
- 金融庁, 国会提出法案等：第221回国会「金融商品取引法及び資金決済に関する法律の一部を改正する法律」, 2026-07-15成立。
- 金融庁, 「令和8年金融商品取引法等改正（20日後施行）に係る政令の公布について」, 2026-07-29. 一部規定は2026-08-12施行。
- 財務省, 「令和8年度税制改正の大綱」および所得税法等改正法案要綱。特定暗号資産の20%分離課税枠と適用開始条件。
- Flashbots Collective, “The Origins of MEV: Attribution of Arbitrage Opportunity Creation,” 2026. Polygonの1,050,000 blocks、360,026 atomic arbitrageを分析し、96.7%でsingle-source hypothesisを確認。
- Fei Wu, Danning Sui, Thomas Thiery, Mallesh Pai, “Measuring CEX-DEX Extracted Value and Searcher Profitability: The Darkest of the MEV Dark Forest,” 2025
- Burak Öz et al., “Pandora's Box: Cross-Chain Arbitrages in the Realm of Blockchain Interoperability,” 2025
