Classic triggers have no random function. What they have is SetSwitch with Randomize, which flips a switch on or off at random. One switch is exactly 50%, so stacking them builds other odds.
Trigger { -- 보스가 죽으면 주사위를 굴린다
players = {Force1},
conditions = {
Kills(CurrentPlayer, AtLeast, 1, "Boss Unit");
},
actions = {
SetKills(CurrentPlayer, SetTo, 0, "Boss Unit");
SetSwitch("Drop A", Randomize);
SetSwitch("Drop B", Randomize);
SetDeaths(CurrentPlayer, SetTo, 1, "RollDrop");
PreserveTrigger();
},
}
Trigger { -- 둘 다 켜졌을 때만 = 25%
players = {Force1},
conditions = {
Deaths(CurrentPlayer, Exactly, 1, "RollDrop");
Switch("Drop A", Set);
Switch("Drop B", Set);
},
actions = {
SetDeaths(CurrentPlayer, SetTo, 0, "RollDrop");
CreateUnit(1, "Rare Item", "Player Start", CurrentPlayer);
DisplayText("\x13\x03희귀 아이템을 획득했습니다!");
PreserveTrigger();
},
}Switches are shared by the whole map, so two players rolling at once overwrite each other. That is why the roll is marked with a per-player counter and resolved inside the same cycle.
Odds like 25% or 12.5% are easy; 10% or 3% cannot be hit exactly this way. When you need a precise chance, that is what EUD's f_rand is for.
Even after moving to EUD, "is this actually random" is a separate question. Complaints that random team assignment leans one way come up often, and the cause is usually the seed rather than the algorithm.
| randomize() is the old function | It is unchanged since 1.16.1 and builds a 32-bit seed by running the random switch 32 times. |
|---|---|
| Drawing more numbers does not make them more random | StarCraft's generator is an LCG, and an LCG's linear structure is not broken by calling it more often - drawing every frame buys no quality at all. |
| A seed derived from a weak seed is weak | Derive your seed from StarCraft's LCG and it is as predictable as the LCG is. |
| % n is biased | Modulo has bias. Where the distribution matters, use rejection sampling instead of % n. |
The cheapest real improvement is to gather the seed from several hard-to-predict sources and run it through a mixer - game time, frame count, the room title, the nickname. And keep your own stream separate from the game's: save the game's seed with getseed, srand to yours, draw, save yours back, then restore the game's. Otherwise your draws push the game's sequence along and its draws push yours.