The 2013 selfish-mining paper by Ittay Eyal and Emin Gün Sirer [0] introduces a variable γ:
We denote by γ the ratio of honest miners that choose to mine on the pool’s block, and the other (1−γ) of the non-pool miners mine on the other branch.
The basic idea is that the attacker pool withholds their block until they learn about a competing block from an honest miner. Since honest nodes prefer the first block they see, this variable expresses the ability of the attacker to race their block in front of the honest block.
Specifically this refers to the fraction of miner nodes that see the attacking block first. As explained in their simulation section:
We assume block propagation time is negligible compared to mining time, as is the case in reality. In the case of two branches of the same length, we artificially divide the non-pool miners such that a ratio of γ of them mine on the pool’s branch and the rest mine on the other branch.
The simulation produces figure 2:
I’m confused about the most pessimistic scenario γ=0, i.e. any time the attacker pool tries to race in front of the honest block they fail miserably. Intuitively I would expect such a pool to consistently lose money, so why doesn’t the red line stay below the grey line?
Is there some additional assumption in the paper that I’m missing?
[0] https://arxiv.org/abs/1311.0243
