10,226 Blocks at the Same Gas Price: Cheap Is Not the Same as Fixed

There is a sentence this blog keeps writing as a footnote, and tonight I went and checked whether it is still true.

The sentence is equal price is not equal cost. It turns up whenever we make the claim that a Satoshie ticket price is a constant in deployed bytecode rather than a number on a dashboard, because the claim has an obvious hole in it. The ticket costs the same for everybody. Getting the transaction that buys it into a block does not. Gas is an auction, priority fees are real, and on a raffle that fills up, whoever paid for inclusion got in and somebody else did not. We do not run that auction, we do not receive the money, and we cannot prove it away.

That is the footnote. It has been sitting in posts for months as an admission rather than a measurement. So I measured it: ten consecutive Base blocks, every receipt, plus a scan of the base fee across 10,250 blocks in a row, plus the same measurement taken at thirty-day intervals back through the last year.

The result is not what the footnote implies. Tonight there is no auction worth the name. That turns out to be a more interesting problem than the one we were warning about.

TL;DR

• The median Base transaction in a ten-block sample on 3 October 2026 cost 0.000000426 ETH, about a ninth of a US cent all in, including the L1 data fee.
• Base’s base fee sat at exactly 0.005 gwei in 10,226 of 10,250 consecutive blocks (99.77%), and the highest reading in the whole scan was 0.99% above that number.
• That is a floor, not a price. The chain ran at roughly 5% to 16% of its 400 million gas limit all day, so the fee market never had to open.
• Floors move. In a 1,024-block window on 5 February 2026 the base fee peaked at 2.354886577 gwei, 471 times tonight’s floor, and in January 2026 the floor itself was lower than it is now.
• At February’s peak, the gas on a modelled 120,000-gas entry would have been 76% of a $1 stake and 0.8% of a $100 one. Congestion does not make a game expensive. It removes the smallest players and leaves the published odds untouched.
• Which is why the cost of entry will never appear in a Satoshie fairness claim. It is not ours to commit to.

What I actually measured

Everything below is from public Base RPC endpoints on the evening of 3 October 2026, and every number is reproducible by anyone with a laptop and no account anywhere. ETH is priced at $2,684.67, the CoinGecko spot figure at the time of the measurement.

The transaction sample is ten consecutive blocks ending at block 52,131,803, which is 1,331 receipts. Ten of those are system transactions that pay nothing, leaving 1,321 real user transactions. For each one I took the gas used, the effective gas price actually charged, and the L1 data fee that Base pays Ethereum for posting the batch, and added the two fee components together. That total is the real, all in, what-left-your-wallet number.

The median was 426,151,411,899 wei. That is 0.000000426 ETH, or $0.00114. The ninetieth percentile was $0.0101. The single most expensive transaction in the sample, out of 1,321, cost 25 cents.

The breakdown is the first surprise. The L1 data fee, the component that crypto spent two years worrying about and that blobs were supposed to fix, had a median of 5,949,377,524 wei. That is $0.000016, and it is 1.40% of the median total. The thing we were all watching is now a rounding error on a rounding error. Base’s own execution cost is 98.6% of what you pay, and Base’s own execution cost is almost nothing.

The fee has not moved, and that is the finding

I started by sampling the base fee once an hour for twenty-four hours. All twenty-four samples came back at exactly 0.005 gwei, to six decimal places.

That should have made me suspicious rather than confident, and it did, because twenty-four readings out of 43,200 blocks is not a measurement of anything. So I pulled the base fee for 10,250 consecutive blocks, which on a two-second chain is five hours and forty-two minutes with nothing skipped.

Of those 10,250 blocks, 10,226 sat at exactly 0.005 gwei. Twenty-four did not. The highest reading in the entire scan was 0.005049294 gwei, which is 0.99% above the floor, and every one of those excursions collapsed back within a block or two.

The suspicion was justified for a different reason than I expected. A separate scan, sampling the same chain at five-minute offsets from my hourly marks, caught 2 October at 18:40 UTC sitting at 0.007326018 gwei. My hourly sample, taken at 18:35 on the same day, had recorded 0.005. The fee moved 46% and back inside the gap between two of my own readings. I only know that because two scans disagreed. If I had run one of them and written the post, I would have published a cleaner number and a false one.

The reason the fee is pinned is visible in the same data. Base’s gas limit is 400,000,000. The blocks I sampled across the day used between 19.5 million and 62.1 million. The chain is running at roughly 5% to 16% of capacity. EIP-1559 raises the base fee when blocks are more than half full and decays it when they are not, and Base has not been more than half full in a very long time, so the fee has decayed as far as it is permitted to decay and stopped there.

A price that never moves is not evidence of a well-functioning market. It is evidence that the market never had to open.

A floor is a number somebody chose

Here is the part that matters, and it is the reason this is a post rather than a tweet.

0.005 gwei is not a law of nature and it is not a property of rollups. It is a configured minimum, and configured minimums have a history. I sampled 1,024-block windows at thirty-day intervals going back a year and recorded the lowest base fee in each window:

• Every window from 7 March 2026 to tonight: minimum exactly 0.005 gwei. Seven months of the same floor.
• 6 January 2026: minimum 0.002837544 gwei.
• 7 December 2025: minimum 0.001516993 gwei.
• 3 October 2025: minimum 0.00400667 gwei, with the window topping out at 0.0117.

So the floor that has not moved in seven months did not exist nine months ago at the value it has now. Somewhere between January and March it was set where it currently sits. I could not find that change documented anywhere I could reach tonight, which is itself worth noticing: the single parameter that determines the minimum cost of touching this chain changed, and the evidence I have for it is a scan I ran myself rather than a changelog I read.

And then there is 5 February 2026. In one 1,024-block window on that day, the base fee ranged from 0.184261727 gwei to 2.354886577 gwei. That peak is 471 times tonight’s floor. Same chain, same contracts, same players, eight months ago.

Nothing in tonight’s data would have let you predict that, and nothing in tonight’s data rules it out happening again next week.

95% of people tip when nobody is asking

One more thing in the receipts, because it complicates the comfortable reading.

Of the 1,321 user transactions, 1,256 paid above the base fee. That is 95.1%. The median tip was 0.001 gwei, which is a 20% premium on top of a base fee that was already at its floor in a chain running at 7% utilisation. Nobody needed to pay it. There was no queue. Effective gas prices in the sample ranged from 0.005 gwei at the bottom to 0.505 gwei at the top, a 101x spread between the cheapest and the most eager, on a night when every single one of them would have been included anyway.

Mostly this is wallet defaults rather than strategy. But the mechanism is live and the wiring is in place. The two-tier inclusion market we warned about in the footnote already exists on Base. It simply costs a fraction of a cent to be in the top tier tonight, so nobody notices they are in an auction. That is a very different situation from the auction not existing, and the difference only shows up when demand arrives.

What congestion actually does to a draw

Now the arithmetic the whole measurement was for.

I do not have Satoshie’s deployed source in front of me as I write, so I am not going to quote our entry cost as a measured figure. Take 120,000 gas as a model for a draw entry, which is in the right neighbourhood for a mint plus a transfer and is the kind of number a reader can replace with their own. Hold the ETH price fixed at tonight’s $2,684.67 so the only thing moving is gas:

• Tonight, at the 0.006 gwei median effective price: $0.0019.
• At 1 gwei, a figure Ethereum mainnet considers a quiet morning: $0.32.
• At February’s 2.354886577 gwei peak: $0.76.

That is a 392x range on the same action, and here is what it does depending on what you staked:

• $1 entry: 0.19% tonight, 75.9% at February’s peak.
• $5 entry: 0.04% tonight, 15.2% at February’s peak.
• $20 entry: 0.01% tonight, 3.8% at February’s peak.
• $100 entry: 0.002% tonight, 0.76% at February’s peak.

The gas cost is flat in absolute terms and therefore regressive in proportional terms. It is the same toll on a $1 ticket and a $100 one, which means it is seventy-six per cent of one of them and under one per cent of the other.

And this is the bit worth sitting with. A congestion event does not make a provably fair draw unfair. The VRF proof verifies exactly as it did. The odds, published as a function of entry count before anyone bought a ticket, are the same honest numbers. What changes is who can afford to be in the room, and the odds have nothing to say about that, because the eighth instalment of this series already established that a provably fair draw says nothing about who else is in it. Congestion is that argument with a price attached. The draw stays fair and the field quietly stops containing people who bet a dollar.

There is no proof that catches this. There is no admin key to point at. Nobody did anything.

Why this is not going in a fairness claim

The temptation here is obvious and we are not going to take it. “Entering a Satoshie draw costs a ninth of a cent” is true tonight, it is a great line, and it would be the single most dishonest sentence on this site.

A fairness claim has to be the kind of statement that survives a change in conditions, because the entire point of putting a guarantee in bytecode is that it does not depend on how the week is going. Our ticket price is a constant in a deployed contract: it is the same number in a bull market, in a congestion spike, and at four in the morning, and you can read it before you commit. The cost of reaching that contract is a configured floor on somebody else’s chain plus the aggregate demand of every other user of that chain. Those are different categories of thing, and the fact that one of them is currently cheap is not a reason to quietly file it under the other.

So the discipline is: the cost of entry gets published as a measurement with a date on it, never as a property. Tonight’s measurement is above. It expires.

What the design does commit to is narrower and it is the part that is actually ours. Escrow, VRF resolution and payout happen in the same transaction, which means a winner does not pay a second gas bill to collect. That matters more than it sounds. A claim-based payout turns your prize into a second transaction at a gas price nobody could quote you when you entered, and in a congestion event it is the small winners whose prizes get eaten by the cost of claiming them. We made a version of this argument this morning, about balances that expire behind a claim step. Here it is again as an economic point rather than a notification one: every extra transaction a design requires is an extra unpriced toll, paid disproportionately by whoever staked least.

The rest stays on the checkable list rather than the promised one. Resolution logic verified and immutable on BaseScan. The VRF coordinator address fixed in deployed code and readable before you stake. Ticket count readable before you buy. The request and fulfilment transaction ids published with every result. None of those depend on what blockspace costs this evening.

Honest limits

One: the 120,000 gas figure is a model, not our number. I did not have the deployed contract source to hand while writing, so I have deliberately not asserted what a Satoshie entry costs in gas. The ratios above hold for whatever the real figure is, because the structure of the argument does not depend on the constant, but a reader who wants our specific number is right to want it and I have not supplied it here.

Two: ten blocks on a Saturday evening is a snapshot. It is the quietest kind of moment on the quietest kind of day. The 10,250-block scan is better, but five hours and forty-two minutes is still not a month, and the thirty-day interval samples are 1,024-block windows, which is thirty-four minutes each. I caught my own hourly sampling missing a 46% move by five minutes, so I am not going to pretend the denser scans are immune to the same failure in a smaller way.

Three: the February comparison holds ETH’s price fixed, and nobody paid these dollars. Applying tonight’s $2,684.67 to February’s gas price isolates the gas variable, which is the honest way to show the multiple and a misleading way to show a cost. The defensible claim is the 392x and the 471x. The dollar figures attached to February are arithmetic, not history.

Three questions worth asking any on-chain game

1. What does it cost to enter, and is that number a property or a measurement? If a platform tells you playing is nearly free, ask what it would cost at 2 gwei. If they have not worked it out, they are quoting you the weather.
2. How many transactions does one full round require? Enter, resolve, claim, withdraw. Every one of them is a separate unpriced toll, and a design that bundles them is handing you something a disclosure never could.
3. Who sets the minimum fee on the chain you are playing on, and when did they last change it? On Base that number moved sometime between January and March this year. If you cannot find out when it changed, you are relying on a parameter you cannot name the owner of.

The footnote was wrong in the way footnotes usually are, which is that it was right about the mechanism and silent about the magnitude. Gas is an auction. Tonight the auction clears at a ninth of a cent because nobody else wants the blockspace, and 95% of people tip anyway out of habit.

Cheap is not the same as fixed. A price that has not moved in 10,226 blocks is still a price, and somebody set the floor it is resting on.

📷 Photo by Vladislav Klapin on Unsplash