by Antistone
The way I see it, there's 3 goals you might have in mind for your event cards:A. Reduce the odds of losing before the player deck runs out
B. Increase the odds of completing your objectives
C. Minimize permanent negative modifiers (adding surveillance, scratching cover)
For goal A, there are actually two subgoals: avoid losing by incidents, and avoid losing by running out of agents. I think Bureaucratic Entanglements (remove 2 agents) is the best overall--it stops up to 2 potential incidents until the next escalation, and also returns 2 agents to the pool.
One Quiet Night gives you up to 4 agents in the late game, but it only stops incidents for 1 turn (rather than until the next epidemic), and it often increases the odds of incidents after the following escalation, because you'll be reshuffling fewer cards. But it could be a better pick if you think you're more likely to lose by running out of agents than by incidents.
Counter-Intelligence stops one incident risk permanently, but it only stops one, and it doesn't remove any agents at all.
Forecast seems pretty lackluster to me. Doesn't block any agents, delays incidents by up to 2 turns in the early game but gets worse as the game goes on.
Cover of Darkness sounds like it removes a lot of agents, but if you removed 4 agents when you played this card, you probably could have removed 2 without playing any card, so the additional removal was only 2 (unless you needed to move 4 cities away anyway). And you need very specific circumstances for it to shine.
For goal B, it depends on the objectives, but Encoded Messages (2 players each pass the other a card) is often useful. This can also help you make a team earlier, which indirectly helps with the not-losing part.
Airlift (move 1 pawn anywhere) and A Holdover From The War (add 1 safehouse anywhere) both have the potential to save a significant number of actions.
Weekend Rendezvous seems like it's usually inferior to Airlift.
There are some other cards in this category unlocked later.
For goal C, Diversion (move up to 3 incident tokens to 1 city) is pretty incredible at stopping surveillance from being added to the board. You can just move the incident tokens to a city that already has 3 surveillance, or (better) to a 4th city that already has an incident token. This also significantly reduces the odds of incident chains (where one incident triggers another incident) because incidents can only add agents to cities with incident markers (but they don't care how many incident markers a single city has).
If you are already winning pretty consistently, and you just want to win better, then I think Diversion is the clear winner.
Of course, this becomes less relevant the later you get in the campaign (additional surveillance has more time to be a nuisance when it happens early in the year).