Arbitrage: the force that keeps prices honest
Markets have no central authority that decrees one price for one asset. What they have instead is arbitrage: the practice of buying something where it is cheap and simultaneously selling it where it is expensive, pocketing the difference at — in theory — zero risk. Arbitrageurs are not trying to keep prices consistent. They are trying to get paid. Consistency is the side effect, and it is one of the most reliable forces in finance: every time prices diverge, someone profits by pushing them back together.
Understanding arbitrage matters even if you never run an arb yourself, because it explains why prices across venues track each other so tightly, why that tracking occasionally breaks, and what those breaks tell you about stress in the system.
Cross-venue arbitrage: the simplest machine
Suppose the same crypto asset trades at 30,000 on venue A and 30,060 on venue B — a 0.2% gap. The arb is mechanical: buy on A, sell on B, capture 60 per unit. Your buying pushes A’s price up; your selling pushes B’s down; the gap closes. Run by enough participants with enough capital, this process keeps the prices of identical assets locked together to within the cost of doing the arb.
That last clause is the whole game. The gap does not close to zero — it closes to the arbitrage band: the zone within which the discrepancy is smaller than fees plus spread plus the risk of executing two legs. Inside the band, divergence persists because correcting it would lose money. The width of that band, instrument by instrument, is effectively the market’s error tolerance.
Triangular FX arbitrage: a worked example
Currencies offer a more elegant version: three prices that must agree with each other. Take three pairs — EUR/USD, GBP/USD, and EUR/GBP. The first two imply a value for the third: if EUR/USD is 1.1000 and GBP/USD is 1.2500, then one euro is worth 1.1000 dollars, and each dollar is worth 1/1.2500 pounds, so the implied EUR/GBP is 1.1000 divided by 1.2500 = 0.8800.
Now suppose the actual EUR/GBP market quotes 0.8770 — euros are cheaper bought directly with pounds than the implied rate says they should be. The loop:
- Start with £877,000. Buy euros directly at 0.8770: you receive €1,000,000.
- Sell the €1,000,000 for dollars at 1.1000: you receive $1,100,000.
- Sell the $1,100,000 for pounds at 1.2500: you receive £880,000.
- You end where you started — in pounds — holding £3,000 more than you began with. Roughly 0.34%, earned in seconds, with no view on any currency.
Each leg of the loop applies pressure: buying EUR/GBP pushes it up toward 0.8800, and the other legs nudge their pairs toward consistency. In practice, a 30-pip triangular gap in liquid major pairs is fantasy — real discrepancies are fractions of a pip, visible for milliseconds, and harvested by machines colocated next to the matching engines. The example is honest about the mechanism and wildly generous about the size.
If you spot what looks like free money across two markets, your first hypothesis should be that one of the prices is stale, one of the markets is untradeable at the displayed size, or a cost you forgot is eating the gap. Genuine free money has the approximate lifespan of a soap bubble.
Why gaps close fast
Arbitrage profits decay with competition in a way directional trading does not. A value investor’s edge can persist for years because acting on it requires conviction and patience. An arb requires neither — only speed and capital — so it attracts exactly the participants best equipped to exhaust it. The result is an arms race in which the gap-closing time has collapsed from minutes (humans on phones, decades ago) to milliseconds and below. The faster the arbitrageurs, the tighter the band, the more honest the prices everyone else sees. High-frequency arbitrage is, in this narrow sense, a public service that charges admission.
The limits to arbitrage
“Riskless” arbitrage is a textbook idealization. Real arbs leak risk and cost from every joint:
- Fees, twice or more. Every leg pays trading fees. A 0.2% gross gap shrinks fast when each side costs 0.10% to 0.25% to execute — fee tiers like Obsidiate’s (taker 0.25% at Bronze down to 0.10% at Diamond) are the difference between an arb existing and not existing for a given trader.
- Legging risk. The two legs do not fill in the same instant. If the market moves between leg one and leg two, your riskless trade becomes a directional position you never wanted.
- Latency. The price you acted on is already old. The faster competitor saw the same gap and closed it while your order was in flight; you arrive to find the opportunity gone — or worse, inverted.
- Transfer time. Cross-venue crypto arb often requires moving assets between venues. During the minutes (or longer) a transfer takes, you hold inventory exposed to the market — which is why persistent cross-venue gaps appear precisely when transfers are congested or withdrawals are restricted. The gap is not free money; it is the market pricing the risk of the pipe.
- Capital and access. Keeping inventory pre-positioned on multiple venues, in multiple currencies, is expensive. The arb pays the firms that have already paid that fixed cost.
These limits explain a pattern worth remembering: when you see a large, persistent price gap between two venues for the same asset, it is rarely an opportunity. It is a symptom — of withdrawal halts, credit fears, settlement friction, or one venue’s prices being effectively fictional. Arbitrageurs with better information and infrastructure than you have already declined the trade. Ask why.
What arbitrage means for everyone else
For the ordinary trader, arbitrage is mostly invisible infrastructure. It is why quoting one global price for an asset is a coherent idea at all, why your charting platform can show “the” price of a currency pair, and why index products track their components. The practical residue: trust tight markets to be honestly priced, treat wide divergences as warning lights rather than gifts, and remember that the fee tier you trade on quietly determines which inefficiencies are even theoretically yours to capture.
Key takeaways
- Arbitrage enforces one price for one asset — not by rule, but because divergence is profitable to correct.
- Prices converge only to within the arbitrage band: fees, spreads, and execution risk define how much divergence persists.
- Triangular FX arb exploits inconsistency among three pairs; the worked loop turns £877,000 into £880,000 — and in real markets, machines harvest such gaps in milliseconds at a fraction of that size.
- Real arbitrage carries legging risk, latency risk, transfer risk, and fee drag — “riskless” is a textbook word.
- Large persistent gaps are symptoms of friction or distress, not free money; the professionals saw them first and passed.
- Lower fee tiers widen the set of capturable inefficiencies — costs decide who gets to do this business.