Selfish mining explained: hiding blocks can pay, and a majority isn't the only danger
Selfish mining is a strategy in which a mining pool keeps the blocks it finds secret and releases them to make honest miners waste their work. A 2013 paper by Ittay Eyal and Emin Gün Sirer showed it can pay with far less than half of the network's computing power.
In a gold rush, a prospector who struck a rich vein had a choice. Ride straight to the claim office, or keep quiet and keep digging while everyone else wasted their days on worse ground.
Bitcoin miners have the same choice. The rules say: publish a block as soon as you find it. Nothing forces you to.
In November 2013, two Cornell researchers, Ittay Eyal and Emin Gün Sirer, showed that keeping quiet can pay. Their paper's title says it in four words: majority is not enough.
What it is
Selfish mining is a strategy for a mining pool. It finds a block and doesn't publish it. It keeps mining on its secret block, trying to get a lead.
When honest miners find a block, the pool releases its hidden one to compete, or its whole secret chain to replace theirs. Honest miners' blocks become stale. Their electricity was spent on a shaft that leads nowhere.
The pool loses some blocks too. But it makes the others lose more, so its share of all rewards goes up.
How it works
The paper sums up its finding plainly: "We show that the Bitcoin protocol is not incentive-compatible" (Eyal and Sirer, 2013). In plain English: a large enough pool can earn more than its fair share by not following the rules exactly.
How large is large enough? That depends on one number the paper calls gamma: "the ratio of honest miners that choose to mine on the pool's block" when two blocks race (full paper). Gamma measures how good the pool is at getting its block seen first.
The paper's threshold is (1 - gamma) / (3 - 2 × gamma). Above that share of the network, selfish mining pays.
- Gamma 0: the pool always loses the race. Threshold: one third.
- Gamma 0.5: half the honest miners pick the pool's block. Threshold: one quarter.
- Gamma 1: the pool always wins the race. Threshold: zero.
Now the paper's own revenue formula, recomputed for a pool with 35% of the network's power:
- Honest, it would earn 35% of the rewards.
- Selfish, with gamma 0, the formula gives about 36.7%.
- Selfish, with gamma 0.5, it gives about 41.6%.
At 25% with gamma 0.5, the formula gives exactly 25%: the threshold, where cheating and honesty pay the same.
That's the uncomfortable bit. The 51% attack article explains why a majority of hash rate is dangerous. This paper says a determined minority can already bend the rewards in its favour.
Why it matters
The incentive. Bitcoin's security rests on the idea that following the rules pays best. Selfish mining is a case where, on paper, it doesn't.
The snowball. If a selfish pool earns more per machine, other miners have a reason to join it. The paper puts it bluntly: "the colluding group will increase in size until it becomes a majority."
The fix the authors proposed. When two blocks of equal length race, miners should pick one at random instead of the first one they saw. The paper says this yields gamma 0.5 and "a threshold of 1/4". That doesn't remove the problem. It raises the bar.
Who pays? Honest miners, through blocks that end up stale. Users barely notice, except that a few more blocks get replaced.
Where people get it wrong
- "Only a 51% attacker can cheat." The paper shows a pool can gain unfair rewards with far less, depending on gamma.
- "Selfish mining steals coins from users." It doesn't touch anyone's balance. It shifts mining rewards from honest miners to the selfish pool.
- "It's been proven to happen on Bitcoin." The paper is a model. We found no primary source confirming a selfish-mining attack on Bitcoin's main chain, so treat claims of one with care.
- "It needs a secret cartel." A single large pool can run it alone. The paper's worry is the opposite: that others would join it.
Watch the stale blocks
A selfish pool can't hide its effect for long. It leaves a trail: more stale blocks than usual, and one pool winning an odd share of block races.
If you hear that someone has "solved" selfish mining, ask which threshold they mean. Majority wasn't enough in 2013, and good intentions aren't a threshold either.
Frequently asked questions
What is selfish mining?
Selfish mining is a strategy in which a mining pool keeps the blocks it finds secret and releases them only to compete with blocks found by honest miners. It makes honest miners waste work and can raise the pool's share of rewards above its share of computing power.
Who discovered selfish mining?
Ittay Eyal and Emin Gün Sirer of Cornell University described it in their paper "Majority is not enough: Bitcoin mining is vulnerable", first posted to arXiv in November 2013.
How much hash rate does selfish mining need?
It depends on how often honest miners build on the pool's block when two blocks race. The paper's threshold runs from one third of the network, if the pool always loses races, down to zero, if it always wins them. With a 50% chance, the threshold is one quarter.
Is selfish mining the same as a 51% attack?
No. A 51% attack needs most of the network's computing power and can rewrite recent transactions. Selfish mining can work with less and aims at higher mining rewards, not at reversing payments.
Has selfish mining happened on Bitcoin?
We found no primary source confirming a selfish-mining attack on Bitcoin's main chain. It would leave traces, such as more stale blocks than usual and one pool winning many block races.

