No coin transfers, no second exchange, no withdrawal delays. Everything happens in one loop on a single exchange — if the math works out in your favour.
Most crypto arbitrage exploits price differences between two exchanges for the same asset. Triangular arbitrage is different — it exploits price inconsistencies between three trading pairs that exist on a single exchange at the same time.
The idea is simple: if you trade USDT → BTC → ETH → USDT in a complete loop, you should end up with exactly as much USDT as you started with, since all three exchange rates should be perfectly consistent with each other. When they're not — even briefly — the loop produces a profit.
A strategy that executes three sequential trades on a single exchange — converting Asset A → B → C → A — to exploit temporary price misalignments between the three pairs. If the cross-rates are inconsistent, the trader ends the loop with more of Asset A than they started with.
Every triangular trade follows the same triangular structure. Here's the classic USDT → BTC → ETH → USDT loop:
Each edge of the triangle represents one trade. The direction of the arrows shows which way the conversion flows. The profit comes from the fact that the implied exchange rate from going around the loop doesn't quite equal 1.0 — it equals 1.003, or 1.005, or whatever the current misalignment happens to be.
Let's walk through a concrete trade on a single exchange. Assume you start with $1,000 USDT and spot the following live rates:
The ETH/USDT price of $3,570 is slightly above where it "should" be given the BTC/USDT and ETH/BTC rates. This misalignment is the opportunity. Here's the full trade:
Spend $1,000 USDT on the BTC/USDT market at $68,000
After 0.1% fee: 0.014691 BTC
Sell 0.014691 BTC on the ETH/BTC market at 0.0522 BTC per ETH
After 0.1% fee: 0.28115 ETH
Sell 0.28115 ETH on the ETH/USDT market at $3,570
After 0.1% fee: $1,002.70 USDT
$2.70 on $1,000 in a single loop. That might not sound like much — but the loop takes under a second, and the same trade can theoretically run many times per minute if automated.
The reason triangular arbitrage is considered an advanced strategy comes down to one fundamental constraint: speed. The misalignment that creates the opportunity is typically corrected within milliseconds on active exchanges, because automated bots from market makers and other arbitrageurs are constantly scanning for exactly these situations.
By the time a human identifies the opportunity, calculates the expected profit, and places three separate orders — the gap is almost certainly already gone. What's worse, if Leg 1 fills but Leg 2 no longer shows a profit, you're now holding BTC with no exit plan that preserves your original USDT balance.
If Leg 1 fills but Leg 2 or 3 becomes unprofitable or doesn't fill at your expected price, you're left with an unhedged position. In a volatile market, BTC or ETH could move against you while you're mid-loop. This is why most serious triangular arbitrage is fully automated — all three orders are pre-computed and submitted simultaneously or sequentially at machine speed.
Opportunities are more likely to appear — and survive long enough to act on — under specific market conditions.
These are the two most common forms of crypto arbitrage, and they're often confused. Here's the key structural difference:
Cross-exchange arbitrage (especially funding rate arb) is generally more accessible. Triangular arb requires either very fast manual execution or coding a bot — and even then, you're competing against institutional-grade algorithms that have been doing this for years. It's worth understanding triangular arb deeply, but it's rarely the first strategy to deploy real capital on.
The mathematical test is straightforward. For any three pairs A/B, B/C, and A/C, calculate the implied cross-rate and compare it to the actual market rate.
For the USDT → BTC → ETH → USDT loop:
Implied ETH/USDT = BTC/USDT × ETH/BTC
= $68,000 × 0.0522 = $3,549.60
If the actual ETH/USDT market price is $3,570 — that's $20.40 above implied. The loop will be profitable in the direction that exploits this gap (USDT → BTC → ETH → USDT).
If actual ETH/USDT is $3,530 — below implied — the loop runs in the opposite direction: USDT → ETH → BTC → USDT.
Most traders who do this seriously build a script that monitors 20–50 coin triplets simultaneously, recalculates the cross-rate every 100ms, and fires orders automatically when the spread exceeds a set threshold (e.g. 0.4% after fees).