check your dice. Bet they’re loaded. if you’re just using acrylic and they’re not from the 90’s… they’re probably fine. a salt water test is going to give you gross errors caused by a bad CG. (loaded dice.) otherwise you’re going to be doing some type of statistical analysis.

(I did this to the murderhobo in my party once. it was… hilarious)

this only works if you can fix the digital dice to be better, heh.

Side note… anybody else here hate karmic dice in bg3?

so a very important question to ask is… do your players want you to thumb the scales, or do they rather want to see where the shitty rolls go? Sometimes, the best solution is to let it play out. Sometimes, the chance of failure makes it more fun, more dramatic. Sometimes, they’re emotionally invested in the characters. Maybe instead of goosing the dice rolls or something, give them options for when the dice rolls fail- maybe they become prisoners and can then escape and find some way of poisoning enough people to complete the objective. maybe they can talk their way out of it; or they can fix up some sort of trap or set up some kind of envirmomental thingamdoo.

this is where you need to know what kind of things your players look for, and what you can sneak into the campaign for them to find.

ways to help out a bit might include tossing some healing potions or useful elixirs into the world, either as loot to be found or as drops. that’s a bit more subtle. or weapon oils for their weapons. Spell scrolls for the mages, too.

alternatively you could toss them advantage occasionally. They’ll know you’re doing it, of course, but then, nobody really calls that out. alternatively you can bring in a DMPC that buffs them, and sets them right so they don’t have to blow so many spell slots while still letting them guide the action.

as a side note, if you’re using a digital dice generator, there may be an issue with it’s script. Older rngs or modern ones that focus on speed or ‘efficient’ coding generally aren’t entirely random that sometimes leads to faulty spreads. If you can, have it give you an absurdly large number of rolls and report the counts of each number. GPt-4 gave this result for a 500d20:

As you can see they’re all fairly even, but not perfectly so. (it averages out to 10.5, but that’s actually somewhat meaningless since true random is going to be… well. random.) in a large pile of rolls, you shouldn’t see them be all perfect.