You've Overpaid for a Transaction. Here's Why the Mechanism Matters.

You set a gas price, hit confirm, and watched the network swallow twice what you expected. Or you set it low and the transaction sat in the mempool for forty minutes while cheaper-looking ones sailed through. Both experiences trace back to the same root: which fee market the chain is actually running.

Two dominant designs exist. The first-price auction, which Bitcoin and pre-2021 Ethereum both used, and the base-fee model introduced by EIP-1559. They behave differently in almost every situation that matters. Not marginally differently. Structurally differently.

How a First-Price Auction Actually Works (and Why It's Punishing)

The mechanic is simple. You name a price. Miners (or validators) pick transactions in descending order of fee until the block is full. Highest bidder wins. Nobody tells you what the clearing price is before you bid.

That last sentence is the whole problem.

Imagine two friends, Priya and Leon, both trying to swap tokens on a busy afternoon. Priya checks a gas estimator and bids 60 gwei. Leon, slightly more cautious, bids 80 gwei. Both get included in the same block. Leon paid 33% more for identical blockspace, and the miner collected every gwei of that premium. There was no mechanism pushing either of them toward the actual market-clearing price, because the market-clearing price only becomes visible after the block is mined.

This creates what you might call a winner's curse dynamic. When the network is congested, users don't know how congested. They guess. Sophisticated traders with real-time mempool data consistently outbid retail users. Wallet estimators lag. During a spike, a user following a wallet's "fast" suggestion from thirty seconds ago can land in the next block or miss it entirely, depending on how fast the spike is moving.

The variance is enormous. On a quiet Sunday the clearing price might be 5 gwei. During a popular NFT mint it might hit 300 gwei. The auction has no built-in signal broadcasting either of those numbers in advance.

What EIP-1559 Actually Changed

EIP-1559 didn't abolish auctions. It changed what you're bidding on.

Under this model, each block has a target size (on Ethereum, nominally 15 million gas) and a hard ceiling at double that (30 million gas). The protocol computes a `baseFee` automatically. If the previous block was more than half full, `baseFee` rises by up to 12.5%. If it was less than half full, it falls by up to 12.5%. The baseFee is then burned, not paid to validators.

You, the user, set two numbers: a `maxFeePerGas` (your absolute ceiling) and a `maxPriorityFeePerGas` (the tip you're willing to pay validators above the baseFee). You pay exactly `baseFee + tip`, as long as that sum stays under your `maxFeePerGas`. Any gap between your `maxFeePerGas` and the actual cost gets refunded.

The concrete version: suppose the current baseFee is 20 gwei. You set a `maxFeePerGas` of 30 gwei and a `maxPriorityFeePerGas` of 2 gwei. If you're included, you pay 22 gwei total. The 20 gwei base is burned. The 2 gwei tip goes to the validator. If baseFee has risen to 28 gwei by the time your transaction is processed, you pay 30 gwei (28 + 2), right at your ceiling. If it's risen to 29 gwei, you don't get included until it drops back.

The result: users mostly compete on the tip, not the total fee. The baseFee is a shared, algorithmically-set floor that everyone can read before they bid.

The Behavioral Difference in a Congestion Spike

This is where the two designs diverge most visibly.

Under a first-price auction, a spike in demand is a scramble. Nobody knows how high to go. Gas prices can jump 10x in two blocks because every user is simultaneously overbidding against an invisible clearing price. The miner captures all of that premium. Users who can't access real-time mempool data (which is most people) either overpay badly or miss their window entirely.

Under EIP-1559, a spike in demand causes the baseFee to ratchet upward at a maximum rate of 12.5% per block, roughly every 12 seconds on Ethereum mainnet. A fee that starts at 20 gwei can't reach 200 gwei in one block. It takes about 20 consecutive completely-full blocks to go from 20 to 200 gwei, which at 12-second block times is roughly four minutes. The path is legible and predictable. Wallets can show users a meaningful forecast.

The catch: the 12.5% cap is a ceiling on how fast the protocol adjusts, not on how high fees can go. If demand sustains long enough, fees do climb very high. EIP-1559 smooths the curve. It doesn't flatten it.

Still, there's a second-order effect that's easy to underestimate. Because the baseFee is burned, validators under EIP-1559 don't benefit from users panic-bidding against each other. Their revenue from any given block depends almost entirely on the priority tip, which users are less likely to inflate wildly because it doesn't change their inclusion odds much once the tip exceeds a small threshold. The perverse incentive for validators to artificially delay blocks to drive up fees is at least partially defused. That matters more than most people acknowledge.

What People Get Wrong About "Predictability"

The most common misconception: EIP-1559 makes fees predictable. It doesn't, quite. What it makes predictable is the direction and rate of change of the baseFee for the next block. Genuinely useful. But fee levels over any span of hours or days remain volatile in response to genuine demand shifts.

Haven't we all, at some point, confidently set a "fast" gas price and watched the network laugh at us anyway?

The other thing people get wrong is assuming first-price auctions are always worse for users. On a chain with reliably low and stable demand, the auction is fine. You bid slightly above the prevailing rate, you're included, nobody's stressed. The auction's pathologies are worst under congestion. On a quiet chain, the lack of a baseFee signal barely matters.

Think of EIP-1559's baseFee like the water level in a harbor lock system. Ships (transactions) still need to float above it, and the level still rises with the tide (demand), but the rise is governed by predictable gate adjustments rather than a sudden wave. First-price auctions are the open ocean: same destination, much rougher crossing.

The Burning Mechanism Is Not a Side Note

When ETH is burned as baseFee, the supply effect is real and non-trivial at scale. During periods of high network activity, the burn rate has historically exceeded new issuance, making ETH net-deflationary over those windows. This is a structural property of EIP-1559 chains that first-price auction chains simply don't have.

It also changes the alignment of incentives. Validators on an EIP-1559 chain want the network to be used (to collect tips) but don't directly profit from fee spikes the way miners did under the old model. Whether that alignment shift is sufficient to remove all manipulation incentives is a legitimate debate. My read: it reduces the incentive, it doesn't eliminate it.

The fee market design, in other words, isn't just a UX question. It touches validator economics, token supply dynamics, and the distribution of information between sophisticated and ordinary users. First-price auctions are honest about being brutal. EIP-1559 is a deliberate attempt to make the brutality more legible. The more interesting question isn't which design is cleaner in theory. It's whether legibility actually protects the users who most need protecting, or whether it just makes the extraction easier to watch.