This is where the training wheels come off. Six lessons built the whole machine: the relative-value mindset that hunts the same asset priced two ways, the cross-DEX spread you close when one pool lags another, the triangular loop that profits off a currency cycle, the atomic bundle and flash loan that let you do it with borrowed money and zero downside-if-it-fails, the cost stack and auction that quietly eat the edge, and the economics of edge, capacity and competition that decide whether any of it is worth doing. No pretests now — just the exam, asking whether it stuck. The trap is usually the option that’s 90% right.
How this exam works
This is a graded exam. Questions arrive one at a time. Once you submit an answer it is final — there is no going back, no retry, and a wrong answer simply fails that question. Your score stays hidden until the very end, where you need 70% to pass. Read every option twice before you commit.
A trader insists arbitrage is about 'predicting where ETH is going next.' What is the sharpest correction?
Select an answer to continue.
Course Recap
Big picture
On-chain arbitrage, end to end
- On-chain arbitrage end to end
- Relative-value mindset
- Same asset priced two ways, right now
- Flat on direction, not a forecast
- Buy low and sell high in one instant
- Self-extinguishing as prices converge
- Cross-DEX discrepancies
- Pool price equals USDC over ETH reserves
- Naive spread ignores price impact
- Optimal size near 2.5 ETH nets about 250 USDC
- Profit is hump-shaped, oversize goes negative
- Trading converges both prices near 1999
- Triangular loops
- Swap a cycle and return with more
- Loop product versus 1 is what matters
- Product 1.005 means a 0.5% edge
- Product below 1 means reverse the loop
- Product exactly 1 is no-arbitrage
- Atomic bundles and flash loans
- All legs land together or revert
- Flash loan repaid in one transaction
- Near-zero own capital required
- Abort the bundle if unprofitable
- Cost stack and auction
- Gas comes off the top first
- Post-gas value gets split by beta
- At beta 0.9 builder takes 864 searcher 96
- As beta approaches 1 searcher approaches zero
- Edge capacity competition
- Edge is the mispricing after all costs
- Capacity capped by AMM pool depth
- Competition bids edge to block producers
- Tiny capacity versus TradFi stat-arb
- No shorting in spot AMM arbitrage
- Relative-value mindset
Key Takeaways
What to remember
- Arbitrage is relative-value, not a forecast — you profit from the same asset mispriced in two places at once and stay flat on direction, buying low and selling high in a single atomic action that nudges the prices back together.
- Profit is hump-shaped, never linear — an AMM pool price is its USDC-over-ETH reserve ratio, every unit you trade moves the price against you, and there’s an optimal size (about 2.5 ETH, netting about 250 USDC here) past which extra size erases the edge and turns profit negative. For triangular loops, it’s the loop product versus 1 that matters — above 1 go one way, below 1 reverse it, exactly 1 is no-arbitrage.
- Atomic bundles and flash loans make it near-capital-free and downside-bounded — borrow, trade and repay in one transaction that reverts entirely if the arb wouldn’t pay, so a searcher runs large trades with almost none of their own money. But atomicity is not inclusion, and a reverted attempt can still cost gas.
- Winning the auction is not keeping the profit — gas comes off the top (1000 gross minus 40 gas equals 960 post-gas), then the priority-gas auction splits what’s left by beta. At beta = 0.9 the builder/validator takes 864 USDC and the searcher keeps just 96; as beta approaches 1 the searcher’s net approaches zero.
- Three forces decide if it’s worth it: edge, capacity, competition — edge is the mispricing after slippage, fees and gas; capacity is small because AMM depth caps trade size (tiny versus a TradFi stat-arb desk, with no shorting); and competition bids most of the remaining edge to builders and validators. Arbitrage is benign backrunning that aligns prices — not the extractive sandwich it’s often confused with.