This piece uses plenty of market jargon. Words with a dotted underline like this can be hovered over or tapped for a quick explanation, without leaving the flow of reading.

Efficient market theory says prices already reflect all information, so no one can earn sustainably superior returns. But in the real world, an entire trillion-dollar industry lives off exactly one job: finding where markets misprice things and pocketing the difference. These two things can't both be strictly true. Grossman and Stiglitz pointed out this paradox back in 1980: if prices were perfect, no one would have an incentive to do the work of correcting them - but if no one corrects them, prices can't be perfect. Markets are efficient precisely because people believe they aren't.

The answer to "why doesn't arbitrage disappear" lies in capital, liquidity, leverage, regulation, and psychology. Those very factors make arbitrage one of the most stabilizing and most dangerous activities in the financial system. This piece walks through the seven most common arbitrage strategies, then distills them into a single principle.

1. Arbitrage: The Merchant's Trade in a Suit

In textbooks, arbitrage is a risk-free trade: buy an asset where it's cheap, sell it simultaneously where it's expensive, pocket the difference. Its foundation is the law of one price: two assets with the same future cash flows must have the same price today. Gold in London and gold in New York, after shipping costs, must be priced equal; if not, someone will ship gold across the Atlantic until the gap disappears.

This trade is far older than finance itself. Before exchanges existed, it went by a less glamorous name: trading. Merchants bought rice cheap where the harvest was good, shipped it to where the harvest failed, and sold high; bought salt at the coast, sold it in the mountains. History hasn't been kind to them - "trader" is almost a slur, someone who "buys at the source, sells at the tip" without growing a single grain of rice - but it was exactly they who pulled the price of rice in two regions closer together, i.e., did precisely the job that finance theory would later dress up with a fancier name: arbitrage. It's no accident that the world's first futures market, the Dojima Rice Exchange in Osaka (standardizing forward rice vouchers from 1730), was built by merchants. Derivatives were born from merchants' needs, not the other way around.

A version of this trade is alive and well today, one almost everyone has encountered: personal-import reselling. An iPhone in Singapore or Japan is several million VND cheaper than the Vietnam price thanks to taxes and promotions; resellers fly over, buy, carry it back, and sell at a markup. From outside, that looks like free money - "just buy low, sell high." From inside, that margin has to pay for plenty: airfare, capital tied up in inventory, the risk that domestic prices drop once official stock arrives, a defective unit with no warranty to exchange, and occasionally a session with customs. The job that looks like "picking up a spread" is actually the business of selling your tolerance for risk for a fee. Hold onto this image of the gray-market reseller: every strategy in this piece, even the trillion-dollar machines, is a suited-up version of it.

Seen that way, modern arbitrage is really trading money. The merchandise is no longer rice or salt, but risk being priced differently in different places. The distance covered isn't just space (between two exchanges), but also time (between spot and futures) and form (between a convertible bond and its component parts). But the structure of the trade stays the same: it needs good logistics to "ship" fast and cheap, large capital to hold inventory while waiting, and the profit per trip is thin enough that it only survives on scale and turnover.

And the history of the trading business also foreshadows the fate of the arbitrage business: every time shipping technology improved, a generation of old-school merchants lost their trade. In 1866, the transatlantic telegraph cable went live, and the gold-price and cotton-price gap between New York and London shrank from several percent down to near transaction costs - an entire generation of merchant sailors' profits vanished within a few years. Today, the "picking up money off the sidewalk" type of arbitrage has gone nearly as extinct, in exactly the same way, eaten alive by high-frequency trading algorithms in milliseconds. What's left, and what this piece is about, is more accurately called risk arbitrage: two prices for the same thing diverge, and they will almost certainly converge, but the road from here to there can be very bumpy. Arbitrageurs don't pick up free money; they get paid to carry a specific risk while waiting for two prices to meet - exactly as a merchant gets paid for holding the risk of a shipment, not for "knowing where it's cheap and where it's expensive." Where it's cheap and where it's expensive, the whole market already knows.

2. Basis Trade: Profiting From Convergence

The iPhone with its "sale price locked in a month ahead" at the end of section 1 was already a miniature futures contract. Futures are just the standardized version of that promise, traded on an exchange. And because they have an expiry date, futures prices are subject to an irresistible gravitational pull: at expiry, the futures price must equal the spot price, because at that point the contract is the asset. The gap between the two prices before that date is called the basis.

Example with the S&P index: spot is at 600, a futures contract expiring in 3 months trades at 605. Basis = +5. Trade: buy spot, short futures. At expiry, say spot is at 602, then futures must also be at 602: the spot position gains 2, the short futures position gains 3, for a total of exactly 5 points, regardless of whether the market rose or fell. No one pays anyone periodically; the profit comes from the gap closing itself.

The largest version of this trade sits in the U.S. Treasury market: the Treasury cash-futures basis trade. Cash Treasuries usually trade a touch cheaper than the futures-implied price, because pension funds and mutual funds prefer the convenience of futures. Hedge funds take the other side: buy the actual bonds, short futures, and fund the position with overnight repo borrowing.

The spread per trade is only a few basis points, so to make it worthwhile they run 10-20x leverage. Per Fed estimates, by September 2025 this trade reached about 830 billion USD, double its prior peak in early 2020, while hedge funds' total long Treasury position reached 2.4 trillion USD and net repo borrowing hit about 1.8 trillion USD by the end of 2025.

3. Funding Trade: Profiting From Cash Flow

The second costume is also born from futures, but in a much younger corner of the market. Crypto created a type of futures that doesn't exist in traditional markets: perpetual futures, contracts that never expire. No expiry date means no gravitational pull toward spot. Instead, exchanges use an artificial mechanism: the funding rate. Every 8 hours, if the perpetual price is above spot, longs pay shorts; if below, the other way around. Whoever is on the "wrong side" of the crowd gets paid by the crowd.

Example with BTC: spot at 100,000 USD, perpetual at 100,000 USD, funding +0.05% every 8 hours. Trade: buy BTC spot, short BTC perpetual. BTC rises to 120,000 USD? Spot gains 20,000, futures loses 20,000, roughly a wash. But every 8 hours, longs pay funding to the short position. At 0.05% three times a day, that's about 0.15%/day on notional. The profit doesn't come from price, it comes from cash flow. Like buying an apartment and collecting rent every month: today's home price doesn't matter.

This is why yield-bearing stablecoins like Ethena can pay double-digit yields in a bull market: their entire model is a giant funding trade packaged into a token. And it's also why that yield isn't "risk-free."

4. Same Position, Two Sources of Profit

Basis trade and funding trade are easily confused because they look identical from the outside: both long spot, short futures, both market neutral. But the source of profit is completely different.

The same long BTC spot + short BTC perpetual structure can be either a funding trade or a basis trade, depending on the expectation. If the expectation is "funding will stay high," it's a funding trade. If the expectation is "the perpetual is at a 1% premium to spot and the premium will go to zero," it's a basis trade. And most interesting of all: you can capture both at once. Spot 100,000, perpetual 101,000, funding +0.05%: entering the position means capturing both a 1,000 USD basis that will gradually disappear as prices converge, and funding paid steadily every 8 hours while you wait. Total profit = basis + funding. That's why crypto funds love this trade so much: Bitcoin's CME basis exceeded 20% annualized in November 2024 and was still around 10% annualized in mid-2025.

5. ETF Arbitrage: The Invisible Price-Keeper

Leaving the derivatives world, the next costume sits right inside the most ordinary investor's portfolio - the person accumulating ETF shares every month. Every ETF has two prices: the fund's share price on the exchange, and the net asset value (NAV) of the basket of securities inside. These two prices don't automatically equal each other; they equal each other because someone gets paid to force them to. Authorized participants (APs) - large market-makers - have the special privilege of exchanging ETF shares for the underlying basket and vice versa. Is the ETF more expensive than the basket? APs buy the basket, exchange it for newly created ETF shares, sell them on the market, and pocket the difference. Is the ETF cheaper than the basket? Do the reverse. Millions of people who buy VOO or VWRA every month get prices close to NAV not by magic, but because this arbitrage crew is constantly at work.

But March 2020 exposed the mechanism's limits: corporate-bond ETFs like LQD at times traded about 5% below NAV. Not because APs fell asleep, but because NAV at that moment was a fictional number: the underlying bond market had frozen, the price baked into NAV was a stale one, while the ETF price was the real price sellers were accepting. To capture that 5% gap, an AP would have to take on a basket of unsellable bonds in a market in freefall. That "discount" was precisely the price of that liquidity risk.

6. FX Carry Trade: Interest Rate Arbitrage

The fourth costume leaves the stock exchange entirely for the largest market on the planet: foreign exchange. The formula sounds almost too crude to believe: borrow a low-interest-rate currency, park it in a high-interest-rate currency, pocket the spread. The classic version: borrow yen at near-0%, buy USD assets or Mexican peso assets yielding 5-11% (2022-2024 period). Theory says this shouldn't be profitable: the high-interest currency should depreciate by exactly the interest rate differential (uncovered interest parity). In practice, for decades, that hasn't happened fast enough, and the carry trade became one of the biggest strategies on the planet, with estimates ranging from 160 billion USD (the hedge-fund share via FX swaps, per BIS) to several trillion dollars depending on how you measure it.

The price of that steady stream of interest income showed up on August 5, 2024: the BOJ raised rates, the yen jumped, and every yen-borrowing position had to buy yen to repay debt all at once. The more the yen rose, the more positions lost money, the more people had to buy yen. The Nikkei lost 12% in a single session - the worst since 1987 - the VIX hit ~65, a level seen only during Covid and 2008, even though nothing that big happened in the world economy that day to explain it. The carry trade had made money like clockwork for years, then gave back a substantial chunk of it in three days.

7. The Professionals' Three Insurance-Selling Trades

The next three variants exist almost exclusively in the institutional world: they require a large balance sheet, cost-effective access to derivatives, and the ability to sit on losing positions for months without facing redemptions. Retail barely touches them, so it's enough to just understand the mechanics - but they can't be skipped, because these are the three most naked versions of the insurance-selling business.

Merger arbitrage - selling insurance on deal breakage. Company A offers to buy Company B at 95 USD/share, but B's stock only rises to 88. The 7 USD gap is the odds on the deal closing: buy at 88, wait for the deal to close, receive 95. Premiums collect steadily from dozens of successful deals, occasionally paying out one big claim when a deal falls apart - as when the FTC sued to block Microsoft's acquisition of Activision in late 2022, and the stock dropped to ~76 USD (the spread widened to 25%) before the deal ultimately closed at exactly 95 USD in October 2023.

Convertible arbitrage - buying cheap options. A convertible bond = a plain bond + a call option on the stock, and because this hybrid product is illiquid, it typically sells for less than the sum of its two components. A fund buys the convertible bond, shorts the stock to neutralize price risk, and pockets the "cheaply sold option." In 2008, this "neutral" strategy lost about 34% when prime brokers pulled leverage and it turned out the entire convertible bond market was held by essentially one type of holder: the very arbitrage funds now facing margin calls together.

Volatility arbitrage - selling insurance on volatility. Implied volatility in option prices is almost always higher than the realized volatility that actually follows, because people are willing to overpay for insurance. Selling options or shorting VIX futures is how you harvest that overpayment, called the variance risk premium. On February 5, 2018, the VIX more than doubled in a single session, and XIV - the most popular short-volatility product - lost ~96% of its value and was delisted: one night gave back years of collected insurance premiums.

8. The Theorem: Same Risk, Two Prices

By now, the seven strategies look like seven different jobs: a crypto trader collecting funding, a bond hedge fund running repo, an event fund reading litigation filings, an options desk computing Greeks. But line them up side by side, and a common structure emerges. Every arbitrage strategy reduces to one proposition:

Prove it by walking back through each strategy. Each row below answers two questions: what has two prices? and what risk is the gap a fee for?

Read top to bottom, this order isn't random: it's a sequence of increasing "hiddenness" of risk. Funding trade and basis trade have risk you can see on a screen every day. ETF and carry trade bury risk inside liquidity - something that only disappears exactly when you need it most. Merger, convertible, and volatility arbitrage sell insurance on rare events, where the risk stays invisible for years and then materializes in a single week. But it's all the same job: find where one risk is being sold at two prices, buy the cheap one, sell the expensive one, and pray you have enough capital to survive until the two prices meet.

From this frame, the phrase "the market is mispricing things" also needs restating more precisely: most so-called mispricing isn't the market being stupid. It's different people, with different constraints, paying different prices for the same risk - a pension fund banned from using leverage pays a premium to buy futures, a panicked retail investor sells an ETF at any price, crypto longs starving for leverage are willing to pay 40%/year. Arbitrageurs don't fix the market's mistakes; they sell a service to people who are constrained, and the spread is the invoice for that service.

9. Why Do Funds Still Make Money?

If the principle is this simple, why doesn't competition drive the spread to zero? Because doing arbitrage requires things most market participants don't have. This is exactly the "limits of arbitrage" that Shleifer and Vishny described in 1997: arbitrage in practice is done with other people's money, by a small number of specialized institutions, and that fact itself is what limits it.

Put the pieces together: the spread exists because someone is constrained, and only a small group with enough capital, enough leverage, and enough infrastructure can harvest it. That group makes steady money not because they're smarter than the market, but because they're renting out their balance sheet and selling insurance - two services that always have demand. But the insurance business model has a dark side: sometimes all the customers file a claim at once.

10. Arbitrage Becomes a System-Level Game

There's been a quiet shift in how arbitrage funds talk about themselves. Twenty years ago, the pitch to investors revolved around "we have a magical model." Today, very few funds talk about algorithms. They talk about data pipelines, colocation, FPGAs, risk engines, inventory optimization, capital allocation, execution quality. All words that sound like they're describing a factory, not a scientific discovery. And that's exactly the point.

The best analogy is the logistics industry. Everyone knows how to ship goods from A to B; the map isn't a secret. But only a handful of companies do it faster, cheaper, more reliably, at greater scale - and all the industry's profit flows to that group. No one wins the logistics business by knowing one secret route; they win through systems. Modern arbitrage is the same - the "trading money" business from section 1 has never changed in nature, only the difficulty of the logistics has changed.

This explains a paradox: every strategy in this piece is described publicly, has textbooks, even has tutorial videos, and yet the profit hasn't disappeared. Because the profit no longer lives in the idea. Recall the theorem in section 8: the spread is a fee paid to whoever carries the risk - and in a competitive market, that fee falls to whoever can carry the risk at the lowest cost. Whoever has the cheapest capital, the fastest infrastructure, the best risk management can accept a thinner spread than everyone else; the spread gets compressed down to exactly the level where only they can still survive.

For everyone else, the opportunity is still "visible" on the screen, but no longer profitable. What determines who makes money isn't who spotted the arbitrage first, but who exploits it more efficiently and survives longer.

In real life, this process has already run its course right in front of us: the neighborhood merchant was replaced by giant ecommerce and logistics systems. "Knowing where it's cheap and where it's expensive" is now done by price-comparison algorithms and warehouse networks at industrial scale; the price of a given item in the city and in the provinces is closer than at any point in history. But notice: room for the small trader shrinks without ever going to zero. There's still personal-import reselling, secondhand goods, collectibles - niches the big systems haven't reached, or aren't worth reaching. Retail financial arbitrage follows the exact same shape: any spread large enough and durable enough to feed a big system already has a big system sitting on it; what's left for individuals are niches too small, too manual, or too risky for machines to bother with. Which leads straight to the most important rule of the next section: the spread left over for retail always comes with the reason it's still left over.

11. When Retail Gets Involved: Where Does the Risk Hide?

Yet individual investors try these strategies every day, and the failures share a notably consistent pattern. It's not the "the whole crowd FOMOs into one stock and blows up" pattern like GameStop. The recurring pattern is far more subtle: retail thinks they're doing risk-free arbitrage; in reality they're selling a type of risk they don't realize they're selling.

Box spread: "Literally can't go tits up"

The classic legend of WallStreetBets, 2019. A box spread is a four-leg options position that locks in a fixed payoff at expiry - in theory, a synthetic loan, textbook-perfect arbitrage. The account 1R0NYMAN built this position and made the now-legendary claim: "literally can't go tits up" - no way to lose.

The problem: that math holds for European-style options, but the position was built on American-style options - the kind where the buyer can exercise early, at any time. Early assignment happened, the "locked-tight" structure unraveled leg by leg, the broker force-closed the position, and roughly 5,000 USD in capital turned into a loss of more than 50,000 USD. The formula was read correctly; the assumption behind the formula wasn't.

Funding trade, the "200% APR" version

People who actually run funding trades do it in a very boring way: low leverage, both legs fully hedged, disciplined position sizing, single-digit to low-double-digit annual returns. It's boring, so no one brags about it online. FOMO traders do the opposite: they see 200%/year funding on some altcoin, conclude it's "free money," and pile in with 20-50x leverage. Funding is high precisely because everyone is already long that same altcoin - meaning it's highest exactly where reversal risk is greatest. Funding flips sign, the basis widens, margin call, and the account is gone before the "steady cash flow" can catch up.

The internet has a long-standing tradition: people who make 8%/year rarely write victory-lap posts, while people who lose 90% of their account usually aren't in the mood to post a follow-up either. That's why there's no loud "funding trade disaster" the way there was for GameStop - just people quietly leaving the forum.

Triangular arbitrage: "The spread was never yours"

A familiar pattern on programming forums: code your own bot to scan for price gaps across three currency pairs on the same exchange, the backtest shows opportunities everywhere, run it live and make nothing. It's not that the bot is wrong. It's not that Python is slow - well, actually, it is slow, but in a deeper sense: the spread you see on screen is already a dead spread. HFT systems with colocation and FPGAs (the same machine from section 10) already ate it several milliseconds earlier; the data reaching a retail machine is just a snapshot of the past. The best reply ever posted to this kind of thread summed up the whole problem in one line: "the spread was never yours."

Cross-exchange arbitrage: killed by the distance

The plan on paper: buy coin cheap on Exchange A, transfer it to Exchange B where it's expensive, sell, repeat. Reality: buy, withdraw, wait for network confirmation, and during those minutes the spread vanishes - or worse, reverses, turning the arbitrageur into someone holding the bag right at the top. The on-chain transfer window is naked, unhedged price risk, and it shows up exactly when the market is most volatile - which is exactly when the spread looks most attractive. The gap between two exchanges isn't free money; it's the market wage paid for carrying inventory risk during transit, and that wage is only livable for someone who already has inventory sitting on both ends.

A great many blowups don't come from a flawed model, but from a seemingly obvious assumption suddenly no longer holding. And that's not just true for a 5,000 USD account on Reddit. The next section is this exact pattern, blown up to trillion-dollar scale.

12. Why Does Arbitrage Sometimes Cause a Crisis?

The final paradox: an activity born to make prices more correct is the lead character in many collapses. The mechanism is always the same, similar enough to draw as a loop:

History repeats this pattern with eerie precision:

• LTCM, 1998: a team with two Nobel laureates, specializing in bond convergence trades, ~25x leverage on 4.7 billion USD of capital. Russia defaulted, every "certain to converge" spread widened at once, and the fund lost 4.6 billion USD in under four months. The New York Fed had to convene 14 banks to inject 3.6 billion USD to unwind the position in an orderly way, because letting the fund collapse freely could have taken down its counterparties with it.
• Treasury basis, 3/2020: Covid drove volatility up, futures margin rose, and basis-trade funds net-sold about 173 billion USD of U.S. Treasuries in a single month. The world's safest asset lost liquidity right when the whole world needed it most. The Fed had to buy about 1.6 trillion USD of Treasuries within weeks to restore the market. Notably, this trade's size in 2025 has already doubled its 2020 level.
• Convertible arbitrage, 2008: the "risk-neutral" strategy lost about 34% when prime brokers pulled leverage and it turned out the entire convertible bond market was held by essentially one type of holder.
• Volmageddon, 2/2018, and the carry trade, 8/2024: two times a derivatives market amplified itself - short-vol had to buy VIX futures as VIX rose, carry traders had to buy yen as the yen rose. In both, the forced buying was exactly what pushed the price further away.

13. Conclusion

Back to the opening question: if markets are efficient, why does arbitrage still exist? Because "the market" isn't a single brain searching for the correct price. It's millions of players with different constraints - a pension fund that can't borrow, a bank charged a capital fee, a small investor with fear, a speculator starving for leverage. Each constraint makes one risk get priced differently in two places, and each gap is an invitation: whoever has enough capital, enough nerve, and enough infrastructure can step up to carry that risk, and the market will pay the fee.

So arbitrage never disappears. It just changes costumes: today it's the funding rate on a crypto exchange, yesterday it was the bond basis, last week it was the deal spread on an M&A transaction. Everyone has the idea; the reward belongs to whoever exploits it more efficiently and survives longer. And this piece's theorem also hands you the tool to evaluate it: whenever you encounter a strategy advertised as "market neutral," "delta neutral," "steady returns regardless of the market," don't ask how it makes money. Ask two questions instead: what risk is it selling insurance on, and what is it silently assuming? If you can't find that risk, it doesn't mean it doesn't exist. It just means it's hiding in the hardest place to see - usually inside the seller's own leverage and liquidity.