Say, if I were to make a tag for questions about logic-defying spaces, would non-euclidean suffice? Or should I go with non-euclidean-space or something else?
@Vaillancourt I'm guessing the non-euclidean tag was prompted by the recent Tardis question.
@DMGregory I think that's good, but that just me. If in doubt, you could always make both tags & have on as a synonym for the other, but again, that's just my take.
@DMGregory As long as you're here, I had something I wanted to ask. WRT your teaching exp, do you have any tips for coming up to speed quickly on prototyping new/repurposed mechanics?
Yes. Specifically, I'm toying with using a hexflower or something like it as a combat resolution mechanic.
Basically each cell would be some sort of turn based action indicating damage, range, # of targets.
Player rolls dice & picks which dice go onto multiple hexflowers - one per chracter or unit.
In theory this is well & good & the 2 unit paper mock I did was interesting enough to make me want to dig into it more.
My question is: are there some methods for getting a sense of how it plays out in the large?
Scaling up the paper mock is a bit slow, but I'm not sure that moving to code is the next move.
Granted, this might meaty enough for a full question post & if that's where I should take it I will. Since it's also 'fuzzy' I didn't know if that'd be a good format fit though.
I figured I'd run it past you as your classroom experience might be a good lens for it.
Yeah, this would give you the probability distribution of the location in the grid on the xth move.
So rather than a qualitative "down and left is more likely" you could work out precise expected values, like "After three turns, I'll have dealt an average of Z damage in total"
Right - that makes sense. I wasn't sure that I'd want to include the directional skewing as used in the link. But it seems to me that an MC sim would still be useful for - it'd be a relatively low investment way to look at the aggregate of many play throughs, which might be what I'm really thinking about.
And assuming I wanted to do something with the idea, it'd also be a way to do some preliminary balance & maybe even some adversarial AI.
And on the flip, if I did want to add skewing, it'd also hold up w/out needing to rewrite the whole model, so it's robust wrt that design choice - probably others as well.
Thanks for sharing your insight!
Oh nice - if I want to use transition blocks in places, I could probably use MCs to make a heat map to see what block does to the X-moves decision space.
Looks like after a large number of turns, you're about 60% more likely to be in the bottom cell, and 20-30% less likely to be in the other corners, compared to the average cell.
Ah, and you can also use it ask questions about steering. If you want to get somewhere, & have to choose between move X or Y, changing the starting % at row 1 to model one or the other will give you a probability map.
Again, rad & thanks. This gives me a starting place to explore from.
To verify my understanding of the math, in cell A3, the formula is =(11*B2 + 9*C2 + 5*D2 + 1*E2 + 3*F2 + 7*G2)/36 because there's 11 different rolls that go from B to A, 9 rolls to get from C to A, etc. out of (hence division) 36 possible rolls that could be made?
In practice, when setting these up, I always get at least one transition wrong - that's why I add the total column, so I can check whether I'm leaking or double-counting probability anywhere.
Once I see a leak/double-count happening, I can set all the initial probabilities to 0, then set one at a time to 1, to figure out which cell's outflows I've messed up.
If this was a novel, this is the point where I put the book down, because it's too good to continue reading & the story needs to be savored before proceeding.
I'm off to let this steep in my brain & see what sorts of broth it makes.