The Cost Stack & the Auction

You spotted a clean $1000 mispricing across two pools. You build the bundle, it lands, it commits. Congratulations — you keep maybe a hundred of it. The rest evaporates into a stack of costs you half-forgot and an auction you didn’t realize you were bidding in. This lesson is the part nobody puts on the highlight reel: the slow, undignified subtraction that turns “I found $1000” into “I netted $96.”

The cost stack — gross spread is not profit

Think of the gross spread as a restaurant’s menu price and your net as what’s in your pocket after tax, tip, and the parking ticket you got walking in. The number on the menu was never the number you keep.

Here’s the full subtraction, top to bottom:

1. Slippage / price impact — your own trade moves the pool price against you. The bigger your size, the worse the fill. This is already baked into the realized spread: when we say “gross of $1000 after slippage,” we mean the spread you can actually execute, not the dreamy mid-price you saw on a chart.
2. Gas — the EVM charges you for computation whether you win or lose. Land the bundle: you pay gas. Get front-run and revert: you still pay gas. It’s the cover charge.
3. Priority fee / builder payment — the bribe. To get included and ordered correctly (your backrun must land right after the trade it targets), you pay the block builder. This is the big, variable one, and the next two sections are entirely about how brutal it gets.

The formula is just honest bookkeeping:

$\text{net} = \text{gross} - \text{slippage} - \text{gas} - \text{builder payment}$

A worked stack on a $1000 gross opportunity:

You found a thousand. You keep ninety-six. Welcome to arbitrage.

When it matters

Always — but most of all when your gross edge is thin. A $1000 spread can absorb $40 of gas and shrug. A $50 spread cannot: gas alone might revert it into a loss. The cost stack is the difference between an opportunity and a trap, and the thinner the spread, the more the stack decides your fate.

Backrunning vs sandwiching

Both backrunning and sandwiching are about where you stand in line relative to someone else’s trade. The difference is whether you’re a scavenger or a mugger.

Backrunning = you land immediately after the trade or block that created a misalignment. Cross-DEX arbitrage is the textbook benign backrun: a whale dumps ETH on pool A, knocking it below pool B’s price; you follow with a buy-low-on-A, sell-high-on-B loop that drags the two pools back into alignment. You harmed no specific user. The whale already got their fill; you just cleaned up the price gap they left behind, the way a janitor follows the parade. This is why arbitrage searchers are mostly backrunners — and why arbitrage is generally considered the good MEV.

Sandwiching = you wrap a victim’s pending swap with two of your own. You buy in front (pushing the price up), let the victim execute at the worse price you just created, then sell behind them for profit. Recall from the MEV & ordering lessons: this is the classic harmful MEV — the victim’s slippage is the attacker’s revenue. You didn’t find a misalignment; you manufactured one at someone’s expense.

When it matters

It matters for both your strategy and your conscience. Benign backrunning is broadly tolerated and even encouraged — it makes markets more efficient — while sandwiching is increasingly detected, filtered, and punished by private orderflow and anti-MEV relays. Building a backrun-only arbitrage business keeps you on the side of the market that gets welcomed rather than blocked.

The auction bids your profit away

When several searchers chase the same opportunity, they don’t politely take turns. They bid against each other for the one block slot that captures it — a priority-gas auction (PGA), or in proposer-builder separation (PBS) terms, a block-space auction run by the builder. And auctions do exactly what auctions are designed to do: they transfer the prize from the bidders to the seller. Here the seller is whoever orders the block.

Picture an auction for a $960 banknote where the only cost of bidding is the bid itself. A rational bidder will go to $959. The next will go to $959.50. The note sells for essentially $960 — to the auctioneer’s benefit — and the “winner” walks away with pennies. That auctioneer is the builder/validator, and the banknote is your post-gas spread.

We capture this with a single dial, the competition intensity $\beta$ (between 0 and 1): the fraction of the post-gas profit that gets bid away to the builder/validator.

$\text{searcher net} = (\text{gross} - \text{gas})(1 - \beta) \qquad \text{builder/validator} = (\text{gross} - \text{gas})\,\beta$

• $\beta = 0$ — monopoly. You’re the only one who saw it. You keep the entire post-gas profit.
• $\beta = 0.5$ — you split the post-gas profit down the middle with the block.
• $\beta \to 1$ — perfect competition. The fee is bid up to the full opportunity; your net collapses toward zero and the builder/validator pockets nearly all of it.

Drag the dials below. Watch your green slice shrink as the auction heats up:

When it matters

β is highest exactly where the money looks easiest: large, obvious, well-known spreads on liquid pairs that every bot already monitors. Those are the most competitive, so β sits near 1 and nets you almost nothing. The durable edge lives where β is low — opportunities others can’t see (private orderflow, novel venues, faster detection) or can’t reach (latency, exclusive relationships). Lowering your effective β beats chasing a bigger gross.

Worked: who keeps the 1000

Let’s run the canonical numbers end to end, so the punchline lands with arithmetic behind it.

Step 1 — gross after slippage. The realized opportunity is $1000 (price impact already folded in).

Step 2 — subtract gas. Gas is a fixed $40, paid win or lose:

$\text{post-gas} = 1000 - 40 = 960$

Step 3 — apply the auction at β = 0.9. The builder/validator takes β of the post-gas profit:

$\text{builder/validator} = 960 \times 0.9 = 864$

Step 4 — what’s left for the searcher. The remaining $(1 - \beta)$ slice:

$\text{searcher net} = 960 \times (1 - 0.9) = 960 \times 0.1 = 96$

So of the $1000 spread you found, you keep $96 — under a tenth. The block builder and validator — who did none of the searching and took none of the discovery risk — pocket $864 just for controlling the order of transactions.

When it matters

This is the single most important reframe in on-chain arbitrage economics: gross opportunity is not your business; net-after-auction is. When you size a strategy, model β honestly for the venues you target. A $1000 spread at β = 0.95 ($48 net) is a worse business than a $200 spread at β = 0.3 ($112 net) — even though the first headline number is five times bigger. Chase low β, not big gross.

Recap