Number of players for any game?

It depends what you want to do. If you want to sell as many copies as possible, you put something like "2 - 6 players" on the box. If you want to be honest, you put "OK with 3, great with 4, sort of works with 2 or 5, doesn't work with 6 but we've included enough bits."* ;-)

No-one ever really knows when they start out how many players a game will accommodate, unless it was specifically designed for 2 and only for 2. Usually a multi-player game will handle 3 to 5 players relatively painlessly, although it often works best at one or other end of that range. Beyond that point it starts getting difficult as more players leads generally to a lack of control that causes problems when trying to get the game to balance. If that's what you want, then fine, but it's rare for a game to scale properly beyond 5 players unless it doesn't work with fewer (cf The Great Dalmuti, or Bang!)

(*for instance, Seth keeps insisting that All for One is a 3 to 5 player game. I refuse to play it with 5 as I consider it too much of a random exercise. But I understand that he might disagree with me. He's wrong, of course, but that's his prerogative.)

Also, the structure of the game will influence this. If there's a strong turn order, the situation may change so radically between your turns that you can't effectively plan ahead (which is the effect that Scurra may be referring to in "All for One", where player interaction is low, but player influence is high).

More players work well in games of high interaction, where players can jump in more fluidly to shape events. If this is too rampant, however, the effect is the same, except the play is more chaotic.

Obvious considerations such as ganging up and kingmaking are also very tough to navigate when designing a game that scales from two upward. Games designed for two rarely scale smoothly to three. Other combinations are equally complex. Generally, the smaller the player count range, the more stable the game is at its various loads.

Also, more than two players opens the possability of beign eliminated before the game ends. This is a drag, unless the game either ends when the first player kicks it, or goes into a death-spiral and ends soon after the first body hits the floor. Players hate being weaned out early almost as much as they hate dragging a lifeless tactical situation through an extended game they have no hope of winning.

I design mostly abstracts of late, which has really changed my thinking about multiple players, and the impact of the player count on the design. With all of the complexities heaped into most "themed" boardgames, I think that this is a super-important consideration for designers, and one that most people just shrug off, thinking that just because the game can support more than two or less than five, it can do so gracefully.