Robert Wiblin

If both sides of a conflict view 'credibility preservation' as the #1 goal then they could end up fighting a total war over something they don't intrinsically care much about, because backing down and looking weak are felt to be so unacceptable.

Might still have nothing to do with it and Putin has surely made things worse for R from any perspective with his actions (a bit like the US-Iraq, error must be part of it). But a steelman of the theory isn't as silly as "it's because Ukraine was going to join NATO tomorrow". 3/

"Is it that Putin wouldn't want a large hostile neighbour over which it lacks econ leverage, with a decent military and powerful allies, and that R's relative military strength is declining so it would be even more costly to try to drag U off that track at any point in future" 2/

The "is it because Russia hates NATO" shorthand has probably been unhelpfully narrow. Could instead ask: 1/

The end result is that even if things *might* be hopeless — that is, have a very high annual probability of disaster — so long as they might also be persistently safe, you always have a decent chance of making it through OK even over very long timescales. 5/

We can't average (1%+0.1%+0.01%)/3 = 0.37% and just project that out. Instead we need to project forward the 3 scenarios individually then average them at the end. You can actually use Max's nice graph for that, averaging multiple projections for values you think plausible. 4/

In fact the chance of not getting a tails is actually 50%, no matter the # of flips! That's an extreme artificial case but the same issue bites in nuclear risk. Let's say we think the annual risk is either 1%, 0.1% and 0.01%, we don't know which and all are equally likely. 3/

If a coin is equally like to be weighted such that it comes up tails 100% of the time, or heads 100% of the time, you can't average the two, get 50% for each, and say the chance of not getting a tails after 2 flips is 0.5*0.5 = 25%. 2/