Bsecure
Articles · How Bitcoin Works

Why Do Bitcoin Transactions Need Confirmations?

A Bitcoin transaction reaches the network in seconds, yet people are told to wait for six confirmations. What a confirmation is, why one block is not the end, and where the number six really comes from.

On March 11, 2013, Bitcoin split in two. A miner running version 0.8.0 of the software, the newest at the time, produced a large block at height 225,430 that older versions of the software could not accept.1 Nodes running the new version built on top of it. Nodes running older versions rejected it and kept extending a chain of their own. For a while there were two Bitcoin ledgers growing side by side, each with its own blocks and its own confirmations.

The fix was a human decision. Large mining pools agreed to switch back to the older version, so that everyone would end up on one chain that all the software could accept, and the chain built on the new block was abandoned.1 Transactions that had been confirmed only on the abandoned side went back to waiting. The developers' post-mortem records that during the split there was "at least one large double spend," made by someone experimenting to see whether it could be done.2

Ordinary users were told to do nothing, since their software would move to the right chain on its own, and it did.1 But the episode showed, in an unusually stark way, something that is true every day in a quieter form. A confirmation is not a receipt. It is a measure of how hard a transaction would be to undo, and that is why Bitcoin makes people wait.

Sent in Seconds, Settled in Blocks

When you send bitcoin, the transaction itself travels fast. Your wallet signs it, and within seconds thousands of computers around the world have checked it and passed it along. The person you paid will usually see it arrive almost at once, marked as pending. That part of the journey is covered step by step in what happens when you send bitcoin.

What takes longer is getting into the ledger. A pending transaction waits in each node's list of unconfirmed transactions until a miner includes it in a block. Miners usually choose the transactions paying the highest fee for the space they take, so when many people are sending at once, a low-fee transaction can wait longer. Once it is in a block, it has one confirmation. Every block added on top of that one counts as one more.3

Blocks Don't Keep Time

You will often read that a Bitcoin block arrives every ten minutes. That is true only on average. Miners make guess after guess at a puzzle, and a block is found whenever one guess happens to win, much like a lottery drawing that never stops. The network adjusts the puzzle's difficulty to hold the average near ten minutes, but nothing controls any single gap.4

Take one ordinary week, October 22 to 29, 2026. Miners found 1,074 blocks. Dozens of them arrived less than a minute after the block before. The longest wait was more than an hour. The fastest run of six blocks in a row took about twelve minutes, and the slowest took more than two and a half hours; a typical run of six took just under an hour.5 So when someone says six confirmations take an hour, they are describing an average, not a promise.

Why One Block Isn't the End

If a transaction is in a block, why not treat it as done? Because a block can still be replaced, in two ways.

The first is innocent. Now and then two miners find a block at almost the same moment, and for a short time different parts of the network have different versions of the latest block. Satoshi Nakamoto described exactly this in the white paper: the tie is broken when the next block is found on one side, and the nodes working on the other side switch over.4 The losing block is dropped, and any transaction that was only in that block goes back to waiting.6 Most of its transactions are usually in the winning block too, and ties like this have become rare, but they show that the newest block is always the least settled. The 2013 split was the same thing on a much larger scale, caused by a software bug.

The second way is deliberate. Suppose someone pays you and also controls a large amount of mining power. He could secretly build his own version of the chain, one in which his bitcoin goes back to himself instead of to you. He cannot use this to take anyone else's coins, only to undo his own payment, because every node still rejects transactions without valid signatures.4 And his version only wins if it overtakes the real chain. Every block added on top of your transaction is one more block he would have to redo while the rest of the world keeps building.

The Odds, Block by Block

The white paper treats this as a race, and the odds can be worked out. The most careful calculation comes from a 2012 paper by Meni Rosenfeld, which gives a cheater's exact chance of success for each number of confirmations and each share of the world's mining power.7

Blocks on top of your transactionCheater with 10% of all miningCheater with 30% of all mining
120% (1 in 5)60%
25.6% (1 in 18)43%
31.7% (1 in 58)33%
40.55% (1 in 180)25%
50.18% (1 in 560)20%
60.059% (1 in 1,700)16% (1 in 6)

A tenth of all the mining power in the world is an enormous amount. Even so, against a cheater that size, each new block cuts his chances by about three times, until six blocks leave him roughly one chance in seventeen hundred. Against a cheater with three tenths of all mining, six blocks still leave about one chance in six. With more than half, he wins eventually no matter how long you wait, which is a different kind of attack altogether.4 And all the while, every block he mines in secret earns him nothing if he loses the race.7

Where Six Comes From

So why six? Not because the network requires it. Rosenfeld put it bluntly: "There is nothing special about the default, often-cited figure of 6 confirmations." It came from two assumptions: that no cheater controls more than a tenth of the mining power, and that a risk below one in a thousand is acceptable.7 Satoshi's own table in the white paper reaches nearly the same place, showing that against a cheater with a tenth of the mining power, five blocks bring the risk under one in a thousand.4 Bitcoin Core, the most widely used Bitcoin software, shows six as its recommended number of confirmations.8 It is a habit built on that math, displayed by a wallet. It is not a rule of the network.

The network does have one waiting rule of its own, and it shows how seriously Bitcoin takes the risk of a replaced block. The new bitcoin a miner earns for finding a block cannot be spent until another 100 blocks have been added on top of it.9 If the block is replaced, its reward disappears with it, and the rule makes sure that reward has not already been passed on to anyone else.

How Long Is Long Enough?

The honest answer is that it depends on what you are handing over. A coffee shop might decide one confirmation is plenty, or even none; the most it could lose is a coffee, and no one is going to mine blocks in secret to steal one. Someone selling a car has more reason to wait for several blocks. Someone selling a house, more still. The more a person could gain by undoing a transaction, the more blocks you want on top of it before you let go of what you are selling.3

That is the real reason Bitcoin transactions can feel slow. Sending is quick. Settling is a matter of probability, and probability takes blocks, and blocks take time. The wait is the price of a ledger that no company keeps and no one can quietly rewrite.

Lesson 4 of the course walks through mining and the chain of blocks, and why that record is so hard to change. For why a seller has to wait at all, see why you can't just copy a bitcoin. What makes any money trustworthy in the first place is where the free first lesson begins.

Sources

  1. Bitcoin.org, alert on the March 2013 chain fork, March 2013.
  2. Bitcoin Improvement Proposal 50, "March 2013 Chain Fork Post-Mortem."
  3. Bitcoin Wiki, "Confirmation."
  4. Satoshi Nakamoto, "Bitcoin: A Peer-to-Peer Electronic Cash System," 2008, section 4 (difficulty), section 5 (ties between blocks) and section 11 (the attacker's race and its odds).
  5. mempool.space, block times for blocks 972,549 to 973,622, October 22 to 29, 2026, checked October 28, 2026. Miners set block timestamps themselves, and they can be off by minutes, so the shortest gaps are approximate.
  6. Bitcoin Core 31.0 source, validation.cpp, which returns a replaced block's transactions to the waiting list.
  7. Meni Rosenfeld, "Analysis of Hashrate-Based Double-Spending," 2012, revised 2014, Table 1 and section 6. The white paper's own table, a simpler approximation, gives a cheater slightly lower odds.
  8. Bitcoin Core 31.0 source, transactionrecord.h, the wallet's six recommended confirmations.
  9. Bitcoin Core 31.0 source, consensus.h, the 100-block rule for mining rewards.

Keep reading