Crypto Arbitrage — Cross-Exchange, Triangular & Why Retail Is Usually Late

Bitcoin costs $64,500 on one exchange and $64,800 on another. You buy here, sell there, $300 of difference lands in your pocket — no guessing the direction, no charts, no emotions. That's what arbitrage looks like on paper, and that exact picture is what courses, bots and "passive income systems" are sold with.

The reality is less photogenic: that same $300, after fees, a transfer charge and a quarter-hour waiting for on-chain confirmations, can turn into a loss. Crypto arbitrage is real — prices genuinely differ, because the market is fragmented across hundreds of platforms — but today the money from those differences is mostly collected by infrastructure, not by a human with two browser tabs open. This article shows how the main flavors of arbitrage work and exactly where retail loses the race.

Educational disclaimer: this material is for education only and is not investment advice. Arbitrage requires holding capital on multiple platforms (risk of insolvency and frozen funds), and its variants involving transfers between exchanges carry execution risk: the price can run away before the funds arrive. Some "opportunities" exist purely because withdrawing funds is difficult or impossible. Don't commit funds whose loss you can't accept.

What Crypto Arbitrage Is

Arbitrage is the simultaneous buying and selling of the same asset in different markets to exploit a price difference. In crypto, the differences come from fragmentation: every exchange is its own micro-market with its own supply, demand and liquidity, and there's no central price fixing. The main flavors:

• Cross-exchange arbitrage. The basic variant: buy on exchange A, sell on exchange B. In the naive version it requires transferring coins between exchanges (and that's usually where it dies); in the practical version you hold capital on both exchanges in advance and execute both legs simultaneously, rebalancing balances later.
• Spatial arbitrage. Price differences between regions of the world — the flagship example is the "kimchi premium," the years-long premium on Bitcoin prices on South Korean exchanges. It sounds tempting, but the premium exists precisely because capital controls and withdrawal restrictions make it hard to close.
• Triangular arbitrage. A rate mismatch between three pairs on the same exchange: e.g. BTC→ETH, ETH→LTC, LTC→BTC. If the product of the three rates comes out above 1 after costs, the loop profits. The upside: zero transfers between exchanges. The downside: these mismatches are microscopic and get scooped up by algorithms in milliseconds.
• DEX/CEX arbitrage. Decentralized exchanges price assets via automated market makers (AMMs), so their prices can drift from centralized exchanges. Opportunities can be larger, but network fees, slippage on shallow pools, and competition from MEV bots come with them.
• Spot–derivatives arbitrage. The difference between the spot price and a futures or perpetual contract. This is a separate, more substantial strategy — we've covered it in detail in the funding rate strategy article.

What the Numbers Say

The most honest lesson is walking an "obvious" opportunity through from start to finish. A textbook scenario (based on CoinLedger's educational materials, numbers rounded): BTC at $64,500 on exchange A, $64,800 on exchange B. Paper profit: $300.

Three hundred dollars of "guaranteed" profit turned into an $83 loss — no catastrophe involved, just a regular fee and a quarter-hour of waiting for block confirmations. This is the typical fate of manual arbitrage with a transfer in the middle: transfer time is an open directional position you never planned to hold.

For completeness, here's the same arithmetic for the triangular variant, with no transfers involved: BTC/USDT at $64,000, ETH/BTC at 0.0520, ETH/USDT at $3,350. The loop USDT→BTC→ETH→USDT gives a multiplier of 3,350 / (64,000 × 0.0520) ≈ 1.0066 — theoretically +0.66% in a few seconds. After three taker fees (3 × 0.1%), you're left with ~0.36%, and after slippage on three order books, less still. And that's assuming all three rates held still for the duration of the loop's execution — something the bots that spot the same opportunity faster take care of.

A second number worth remembering: typical price gaps on liquid pairs of major exchanges today are fractions of a percent and live for seconds. Profitability therefore requires either scale (large capital × many micro-opportunities, i.e. automation) or venturing to the fringes of the market, where spreads can be wider — but a wider spread almost always prices in real risk: thin liquidity, questionable platform solvency, withdrawal restrictions. History has spectacular exceptions — the kimchi premium once let people earn tens of percent on turnover — but those exceptions required solving regulatory and logistical problems most participants couldn't solve. The premium was compensation for what wasn't easy to do.

How It Works Step by Step (a Numeric Example)

The variant that makes the most practical sense: cross-exchange with no transfer mid-trade, on capital pre-positioned in advance. An illustrative example:

1. Capital prep. You pre-position funds: 10,000 USDT on exchange A and 0.15 BTC (~$9,700) on exchange B. The capital sits there and waits for an opportunity — that's the cost of readiness (and 2x counterparty risk).
2. Scanner. You monitor the A/B spread on BTC/USDT (off-the-shelf arbitrage scanners or your own API script). You set an entry threshold from the cost arithmetic: combined fees on both legs 0.2% + a 0.1% slippage buffer → you're interested in a spread of ~0.4–0.5% and up.
3. Execute both legs at once. The spread widens to 0.55% (e.g. $64,500 vs $64,855). Simultaneously: you buy 0.15 BTC on A for ~9,675 USDT and sell 0.15 BTC on B for ~9,728 USDT. No waiting for a transfer — both legs close within seconds.
4. The math. Gross difference: ~$53. Fees (0.1% × both legs): ~$19. Net profit: ~$34 on ~$19,400 of turnover, i.e. 0.17%. Yes, that's all of it — and that's assuming orders filled at the prices shown on screen.
5. Rebalancing. After a series of these trades, capital tilts (USDT depletes on A, BTC depletes on B). Every so often you move funds back — and only then do you pay transfer fees and take on transfer-time risk, but by that point without an open opportunity hanging over your head.

Notice what this process resembles: not an "investment opportunity" but a low-margin operating business — infrastructure, monitoring, cost accounting down to fractions of a percent. That's exactly why firms with automation, colocation and volume-negotiated zero fees win at it.

Risks — When the Strategy Does NOT Work

• You're slower than the bots. On liquid markets, HFT algorithms close opportunities in milliseconds. Cheap, publicly available bots don't beat institutional infrastructure — and what's left for a manual trader is usually a spread smaller than their own costs. Retail doesn't lag out of laziness; it's structural.
• Costs eat the margin. Taker fees ×2, withdrawal fees, network fees, slippage on a shallow book, the bid-ask spread. With margins of 0.1–0.5%, every cost item matters; work out the full break-even threshold before you click, not after.
• Execution and transfer risk. The variant involving sending coins mid-trade is an open position for the duration of confirmations: network congestion, paused withdrawals on the exchange, a sudden price move. An opportunity that requires a transfer stops being arbitrage and becomes a directional bet with a delayed fuse.
• Counterparty risk ×N. Working capital sits on several platforms at once, often including smaller, "opportunistic" ones. The insolvency, hack or frozen withdrawals of any one of them can wipe out months of fractional gains. Crypto's history is full of accounts that learned this the hard way.
• Regional traps. A large premium on an exotic exchange usually means withdrawing from it is difficult (KYC for residents, limits, capital controls). Easy to deposit, impossible to withdraw — the funds are stuck, and the premium just stays on the screen. Check withdrawal terms before you deposit your first dollar.
• Taxes and bookkeeping. Hundreds of transactions a month means hundreds of events to reconcile. Under Polish tax rules, crypto-to-crypto is neutral, but every pass through fiat and stablecoins needs recording — without accounting automation, active arbitrage can cost more in admin than it earns.

The no-hype verdict: crypto arbitrage is a real phenomenon and a legitimate strategy — for those who treat it like an infrastructure business: automation, capital pre-positioned in advance, cost accounting down to hundredths of a percent, and humility about platform risk. For everyone else, it's a race where the finish line keeps moving thanks to faster players, and the entry fee (fees, transfers, stuck funds) gets paid regardless of the outcome. If someone tries to sell you a "guaranteed-profit arbitrage bot," you've just found the one sure arbitrage in this whole story: theirs, on you.

FAQ

Rafał — Krypto Bez Ściemy

Trader and founder of Strefa Tradingu. He’s been breaking crypto down on YouTube for years — no hype, no signals, with a focus on market structure and risk management.

kryptobezsciemy.plInstagramYouTubeKlub Zero