Simple Game Theory Model of November 2018 BCH Fork

In mid-November, BitcoinABC plans for what they call a “network upgrade” which will include new consensus rules that are not backwards compatible. The big changes are to unlock OP_CHECKDATASIG and to enforce canonical transaction ordering.

Coingeek and nChain (BMG) are not supportive of these changes and have declared that they will run a different node implementation called BitcoinSV, which will not have replay protection and will not be following the same rules as the upgraded BitcoinABC (from 0.18 onward).

This means that come mid November there is going to be a significant amount of disarray if these two groups continue their course. This is because at least 1/3 of the hashrate on BCH has declared that they do not support a proposed hardfork — this is the classical contentious hardfork.

Hashrate on BCH Oct 10, 2018

So right now the parties seem to be playing chicken. But ABC has a choice to call off the upgrade, so technically they still have some option here to either Fork or Cancel Fork. And SV can either Concede to the fork or Fight it. And ABC can either Concede to their fighting or to Fight On.

Payoffs for players

I have assigned these different outcomes numerical payoffs reflecting the perceived “gain” from what has happened. Don’t think of this as money or anything, just “points” like in a videogame for doing well.

Below is a decision tree modeling these options, with payoffs in the format of (ABC, SV):

Extensive form multi-stage two-player (“Team BitcoinABC” (Jihan and ABC devs) and “Team BitcoinSV” (Coingeek and nchain))

We can use backward induction to find the solution (i.e., the intersection of best responses, or nash equilibrium) to this:

  • ABC in stage 3 will choose ”Fight on” if SV ”Fight” in respones to ”Hard Fork” (2>1)
  • SV will in stage 2b always choose ”Fight” in response to ”Hard Fork” (2>0)
  • SV will in stage 2a always choose ”Play Nice” in response to ”No Fork” (3>1)
  • ABC thus will choose ”No Fork” over ”Hard Fork” (3>2)

Notes

Obviously the payoffs are assumptions and can be altered however one sees fit, but these ranks are more or less reasonable as I see them (please share in comments if you have improvements or ideas).

The payoff for Fork/Fight/Fight is particularly dubious, because there is so much uncertainty about how that shakes out.