Crypto Arbitrage Explained: The 5 Types — and Which Actually Clear Fees

TL;DR: Crypto arbitrage means profiting from the same asset being priced differently in two places — and in 2026 almost none of the "obvious" gaps survive fees, slippage, and transfer delays for a normal retail trader. The one edge that's still realistically accessible isn't spot arbitrage at all; it's low-turnover funding-rate carry, and even that pays a modest, non-guaranteed yield.

If you've searched "crypto arbitrage" you've seen the pitch: buy Bitcoin cheap on one exchange, sell it dear on another, pocket the difference, repeat forever. Risk-free money. This article is the version those pages won't write — the one that does the arithmetic after costs. We'll walk through the five real types of arbitrage, give each an honest edge estimate and its catch, then tell you which one a regular person can actually use.

What crypto arbitrage really is (and why "risk-free" is a myth)

Arbitrage is buying an asset in one market and simultaneously selling it in another where it's priced higher, capturing the spread. In a textbook it's risk-free because both legs happen at once. In crypto, they almost never do.

Every real trade leaks money at four points:

• Commissions. A typical taker fee is ~0.1% per side. A round trip costs you ~0.2% before you've earned a cent. Your price gap has to beat that just to break even.
• Slippage. The quoted price isn't the filled price. Size moves the book against you, especially on thin pairs.
• Transfer time. Moving coins between exchanges takes minutes and costs a network fee. The gap you saw usually closes before your deposit clears.
• Withdrawal limits and lockups. Unverified or new accounts hit caps; some withdrawals face holds. Your capital isn't as liquid as the spreadsheet assumes.

Stack those up and "risk-free profit" becomes "a small, uncertain edge that you're racing bots to capture." That framing matters for everything below.

The 5 types of crypto arbitrage

1. Cross-exchange (spatial) arbitrage

How it works: The same coin trades at $X on Exchange A and $X + gap on Exchange B. You buy on A, sell on B, keep the difference.

Realistic edge: For liquid pairs like BTC/USDT between major venues, gaps are typically 0.02–0.10% — below the ~0.2% round-trip fee floor. To trade it at all you must pre-fund balances on both exchanges so you never physically transfer coins mid-trade; even then the minimum profitable spread runs ~0.15–0.25%, which rarely appears on liquid pairs and, when it does, vanishes in seconds.

The honest catch: By the time a human sees the gap, a co-located bot with fee rebates has already closed it. The wide gaps you do see are usually on illiquid coins or sketchy exchanges you can't safely withdraw from — the spread is the market pricing in that risk, not free money. Kraken and Coinbase are among the more reputable venues if you insist on testing this, but temper your expectations.

2. Triangular arbitrage (within one exchange)

How it works: Exploit pricing inconsistencies between three pairs on a single exchange — e.g. USDT → BTC → ETH → USDT — ending with more than you started. No transfers, so it sidesteps the biggest cross-exchange headache.

Realistic edge: Cycles that look profitable yield maybe 0.1–0.5% gross. A widely cited 2024 study scanned 4,879 candidate triangular opportunities on Binance and found none remained profitable after transaction fees and liquidity constraints.

The honest catch: This is the most bot-dominated corner of the market. Opportunities last 2–15 seconds, and sub-100ms execution reportedly captures the large majority of the profit. You are not beating those systems by hand, and by the time a fee is applied to each of the three legs the gross edge is usually gone.

3. Funding-rate / cash-and-carry (delta-neutral) arbitrage

How it works: You buy the spot asset and simultaneously short the same amount via a perpetual futures contract. Your net exposure to price is roughly zero (delta-neutral). You earn the funding rate — a periodic payment perp shorts collect from longs when the market is bullish.

Realistic edge: This is the one that survives. Over a full market cycle, realistic net returns are roughly 10–15% APY in favorable conditions and 0–5% in neutral markets, after fees. In stable markets a ~0.01% per 8-hour funding rate compounds to roughly 11% gross annualized. Bybit institutional data cited delta-neutral books staying positive all 12 months of 2025 with under 1% max drawdown — modest, but real.

The honest catch: It's a yield, not a jackpot. Funding can turn negative (you'd then pay), forcing you to unwind or flip sides. You carry exchange counterparty risk, liquidation risk on the short leg if you under-collateralize, and the return scales with capital — small accounts earn small dollars. It's the least sexy type and the only one built for a patient retail participant. Perp venues like Phemex support both legs; keep leverage low. This is closest to the boring "carry" trade professionals actually run.

4. Cross-chain / DEX–CEX arbitrage

How it works: Prices differ between a decentralized exchange (on-chain) and a centralized one, or across blockchains. You buy where it's cheap and sell where it's dear, sometimes using flash loans for capital.

Realistic edge: DEX spreads run 0.05–0.5% with opportunities lasting 1–2 seconds. On paper it's the highest-yield category; in practice the yield accrues to a tiny number of specialists. Research found ~$233.8M extracted across 7.2M CEX–DEX arbitrages by just 19 major searchers, with the top 3 capturing ~75% of the value.

The honest catch: You're competing in the MEV (maximal extractable value) arena against bots that see opportunities first and outbid you on gas to execute first. After gas, slippage, and MEV competition the edge is usually too thin to justify the risk. Worse, the CEX leg can fail independently of the on-chain leg, leaving you holding unhedged inventory. This is an infrastructure game — private RPCs, colocation, custom code — not a retail one.

5. Statistical / cointegration pair arbitrage

How it works: Instead of an identical asset, you trade two correlated assets (say two similar L1 tokens) that historically move together. When their price ratio diverges beyond its normal range, you short the expensive one and long the cheap one, betting the spread reverts.

Realistic edge: Not a fixed number — it depends on your model, and it's not risk-free. It's a probabilistic bet that a historical relationship holds. Done well by quant desks it produces a modest Sharpe; done casually it produces losses when the correlation breaks.

The honest catch: This isn't true arbitrage — there's no guaranteed convergence, only a statistical tendency. Correlations break exactly when you're leaning on them (one token gets delisted, hacked, or forked). It needs real backtesting infrastructure, ongoing model maintenance, and cost modeling. Most retail "stat arb" is just directional trading with extra steps.

Bonus: stablecoin depeg & event-market arbitrage

Stablecoin depeg — buying a stablecoin below $1 betting it returns — is occasionally real but rare, and the times it's most below peg are exactly when it might not recover (see past collapses). Event-market arbitrage (prediction markets like Kalshi/Polymarket) looks tidy but the fee-and-spread structure typically eats the theoretical edge; real-money tests of short-horizon crypto binaries have lost money clearing costs. Treat both as niche, not income.

Comparison table

The honest verdict

Here's what most crypto-arbitrage articles bury: for a retail person, the vast majority of "obvious" arbitrage does not clear costs. The persistent, easy-looking edges — spatial gaps, triangular loops, DEX–CEX spreads — are systematically captured by bots and market makers with fee rebates, colocation, and custom infrastructure. When you see a big gap by hand, it's almost always either already gone or attached to a real risk (an exchange you can't withdraw from, an illiquid coin) that is the reason for the gap.

This isn't cynicism — it mirrors what happens when you actually test it. Cross-exchange divergence, "risk-free" gold-token triangles, and short-horizon event-market arbitrage all reliably fail to clear fees in live testing. The spread that looked like profit was the market paying you to take on a cost or risk you hadn't priced.

The one genuinely accessible edge is low-turnover funding-rate carry: delta-neutral, modest, and honest about being a single-digit-to-low-teens yield rather than a money printer. It works precisely because it's not a race — you're collecting a structural payment for holding a hedged position, not sprinting a bot to a vanishing gap.

Who this is realistically for — and safer alternatives

Realistically for: People with meaningful capital (thousands+), patience for a modest yield, and comfort managing two positions across a spot and a perp venue. If that's not you, the funding-rate trade will feel like a lot of setup for small dollars.

Tools/exchanges involved: A reputable spot venue and a perp venue (Coinbase, Kraken, Phemex), an arbitrage/funding scanner if you want to monitor rates, and disciplined position sizing with low leverage.

Safer adjacent alternatives for most readers: If your real goal is "put crypto to work without gambling," plain staking and reputable yield products often deliver comparable single-digit returns with far less operational complexity and no perp liquidation risk. For many people that's the honest better answer than chasing arbitrage.

FAQ

Is crypto arbitrage still profitable in 2026? For high-frequency bots and market makers with fee rebates and colocation, yes. For a retail trader doing it manually, the classic spatial and triangular forms are largely picked clean — gross edges sit below fee floors and vanish in seconds. The realistically profitable retail form is low-turnover funding-rate carry, at a modest net yield.

Is crypto arbitrage risk-free? No. The "risk-free" label assumes both legs execute simultaneously at quoted prices with no transfer delay — none of which holds in crypto. Real risks include slippage, transfer timing, funding flipping negative, exchange counterparty failure, and liquidation on leveraged legs.

What's the most realistic type for a beginner? Funding-rate / cash-and-carry, done small and with low leverage — because it doesn't require winning a speed race. Understand that a neutral market can pay near zero, and that you're managing counterparty and liquidation risk, before committing capital.

Do I need a bot? For spatial, triangular, and DEX–CEX arbitrage, effectively yes — and even then you're likely underpowered versus professionals, so we don't recommend it. Funding-rate carry can be managed manually because it's a low-turnover position you hold, not a gap you chase.

Affiliate disclosure: some exchange links on HashWatch are affiliate links; if you sign up we may earn a commission at no extra cost to you. It never changes our honest assessment — including telling you when something doesn't clear fees.

This article is for information only and is not financial advice. Crypto trading carries real risk of loss. Do your own research.