To finally be awake to chime in...
1 beat 2 and 3, 2 beat 3, in the first two generations.
That's the problem I'm trying to solve.
Also a silly speculation, feel free to dismiss. Should the goal of this be "generate a new set of units each time", or "generate units instead of having them pre-made"?
The latter. Figure the player plays a whole campaign with the same set of units (or the same initial set, if "research" is allowed). But they could start a new campaign that would then create new units again.
I do remember him mentioning earlier that he may try to apply archetypes to his units; that is, the high armored ones being named something Tank-related and the high damage ones being bombers. As long as we get something within each archetype, or most archetypes fulfilled, I'm happy.
Archetypes are/will be named based on their stats. It's not like the program "generates a tank" and performs a different set of generation rules than it would for "generating a fighter." What happens is that a unit is generated and then based on the result assigned a name (not that I've been doing that yet).
The game could figure out which is the most efficient at just straight up dealing damage over time, ignoring other factors, and make that the DPS related unit. It's a jumble between damage, bullet multiplier, ROF and possibly even reloading if you can shoot fast enough.
Thing is, a low DPS unit that can out-maneuver the other unit can win, even though its DPS is lower. So its not all about the DPS.
As long as the enemy is not oneshoting you and you have a healer your could tank all day.
No healing units exist yet.
But yes, your statement is true.
Of course, I as a support unit in League of Legends (Heimerdinger) HP tanked a melee unit (Warwick) successfully three times. The only time he killed me, my turret killed him a second later. It helped that I was three levels higher than he was, but 1.5 seconds of stun and an additional 1.5 seconds of not being able to attack* meant that I was half-way to having my abilities recharge to do it again.
*The concussion grenade on his head stuns for 1.5 seconds and blinds for 3. Blind means "auto attack misses." Warwick being a melee unit, all his skills are based around making his auto attack better.
so, make the unit, figure out the cost, and then find how many can be fielded.
eg, ship-type one might be a failure, with a point cost of 100, while ship-type 2 might cost 500,
then just have them battle at a 5 to 1
I did poke at that once, swapping between a "16 point buy" and a "32 point buy" (where there were twice as many as the former) and while it
worked in that sometimes one side won and sometimes the other side won, it didn't have a good end result (the triangle would always contain two of one and one of the other, obviously). I only did it to see that there was not inherit power of the units being twice as strong, provided they were half in number.
OR:
How about a rising cost, so the first 10 range cost 1, and the +10 cost 2, +12 is 3 etc...?
That kind of math is
really tricky to work out in an algorithm, as its a recursive action. The other issue is that it disincentives specialization too much and with units tending towards an average center, the triangle units wouldn't appear very unique.
It works for things like D&D because players are
trying to hyperspecialize. My algorithm isn't.
I think it uses point buy still but it may not.
It kind of does, but more in the sense that "all the stats add to 12" kind of point buy. Range is the last stat defined and is based on how many points went unspent, so really it's a stat that's based off all other stats (the higher everything else, the lower the range). Range is also semi-quadratic so that large ranges are distinguishable from each other (78 range and 76 range aren't distinct enough to matter (3%), where as 18 range and 20 range probably are: 10% difference).
Author
Topic: Proceedurally Generated RTS (Read 57150 times)