I meter risk.
Generally, against foes of equal ability, a 50/50 split throws all the results onto the dice.
Pulling it off average to 75/25 or 25/75 you bias the probabilities. . . .These are usually my declarations, call them standard offensive or defensive. The reasons why vary greatly, I'll try to explain in a bit.
Past that the all out 0/100 or 100/0 of full offense or full defense are only worth it under the right circumstances.
It's easier to explain the full declarations, and the standards follow the same logics, but less so, and are more complex and are more a feel for the situation than anything.
Full defense is something I do when my attack matters far less than staying alive. . .if an opponent is attacking you, and your friends are attacking it, it's best to full defend, survive the round, and let them take the enemy down. . ..same if you're just waiting for reinforcements to arrive, like if you're a guard and screaming for help from other guards while keeping an intruder from killing you.
Full offense is something I do when I am convinced I won't be attacked, like from ambush, or when a foe is blatantly after someone else and ignoring me. . .or when time is a premium, like when you're the intruder and the guard is screaming for help and you need to deal with him and get going.
The more common normal variations tend to be similar, but less urgent. . .but you need to be cautious. . .playing for too much time, fighting defensively, is generally prudent, but your enemy has a 5% of rolling open ended high on you, so if you allow them to survive long enough to make 5 more attacks, that's 25% chance they'll open end you, or that you'll open end down. . .so you want to take them down, but survive, but the more attack rolls you allow, the more chances you'll get killed. . .
The more I read the above, the more I wonder how much that helps, since so many factors combine to make the choice, it's hard to express it.