If a game offers character creation choices one can optimize character creation.
You could certainly make choices designed to improve your chances of having a starship by being a scout or merchant and staying in as long as you could. You could improve your chances of being combat capable by joining the army or marines. None of these were as foolproof as having full control over your character development choices, true, so it was a "softer" but still present form of optimization.
If you joined the army back in classic Traveller (which was also VERY easy to enlist in), you got the automatic service skill Rifle-1. So any and all army vets had at least that skill as a combat skill (and yeah, marines got Cutlass-1 if they never got commissioned). And if you wanted to be a pilot, you joined the scouts because they all got Pilot-1.
You definitely had things you could angle for even if you didn't have full control.
returning to Traveller, its still doesn't change the fact its exceedingly difficult to optimize toward, well, anything in it because there's so much randomness in it. If you go in trying to build a technician, you could end up with a medic instead--might even be a pretty good medic, but that wasn't that much help in trying to optimize toward an electronics-baseds repair and security specialist.
These posts seem to count it as
optimising when a person builds a PC that is the sort of PC they want - a pilot, a soldier, a technician, a starship owner, etc.
That seems a very non-standard notion of
optimisation, given that it is all about choosing
the character's field of endeavour, whereas
optimisation normally is understood to take a field of endeavour as given, and to be about the means to that.
By the definition in use in these posts, the best RPG for optimising is Cthulhu Dark, because the most important step in PC building is
writing down your character's occupation - so if I want to play (say) a telegraph operator, I write down as my occupation
telegraph operator. Bam! Optimisation done.
Or, in other words, this:
By that standard, being able to choose a class in OD&D was optimization, which I think sets the bar so low as to destroy any useful meaning to the term.
Returning to the standard meaning of
optimisation, it relies upon there being competing and (typically) not-fully-transparent-to-superficial-inspection means of achieving the same goal, which for a PC in a RPG is a given field of endeavour in the game. In D&D the obvious site of optimisation is combat, because combat effectiveness is an intricate mathematical output of complexly interacting inputs - number of attacks per round, timing of attacks in a round, chances to hit, damage dice, etc; which are themselves shaped by various interacting factors like class, stats, proficiencies, etc. It is these interactions that create optimisation pathways like (in AD&D 2e) a specialised dart thrower, or (in Skills and Powers) building a cleric with fighter specs, or (in 3E) getting the right balance of feats and deploying them (eg Power Attack spreadsheets).
Champions/HERO and GURPS have (by reputation at least) a lot of complexity in this optimisation space.
The reason there is no
optimisation in this sense in Traveller or Prince Valiant is because the way to be good at (say) shooting is simply to have a high shooting skill (be that Rifle in Traveller, or Archer in Prince Valiant). The way you get that in Prince Valiant is via build choice; the way you get that in Classic Traveller is via sensible choice of table plus lucky rolling.
A residual question is what non-contingent relationships not obtaining looks like? A broad class of examples might be freeform character capabilities, such as HeroQuest abilities or Torchbearer wises and instincts. Being freeform, one might predict that, that which was great for RP will be necessarily optimal (if optimal is taken to mean something like - relevance and strength of leverage over the narrative.) One could argue (and I have seen in actual play) that for a given genre, setting, mode or premise some choices tend to offer stronger or more relevant leverage over the narrative than others. But then isn't that be the same as saying they are less effective for RP!?
HeroQuest revised has a whole section devoted to advice to the GM on how to balance (say) a Strong descriptor against a Breaks Rocks With Her Bare Hands descriptor, the precise point of which is to ensure there is no optimisation.
Torchbearer or Burning Wheel takes a different technical approach - as an illustration, when one of the PCs in my Torchbearer game took Explosives-wise, together with the Belief that
An explosive solution is a good solution, that was my cue as GM to include explosives and opportunities to blow things up. But the upshot is similar.
But in any event, as per the preceding parts of this post, I don't think
building towards what I want to play, given the anticipated parameters of this campaign is the same thing as
optimisation. The former is predominantly about ends. Whereas the latter -
optimisation - is predominantly about means, and the availability of complex interacting means (which underlie the non-contingent relationships that I mentioned upthread) is a feature of only some RPGs.