simon made an excellent chart of the spells, which I am shamelessly stealing rather than drawing my own:
Nice!
Do you think the underlying data model was too complicated to decipher?
The exact mechanism would have been pretty tricky to figure out, but that’s not necessarily a problem?
The general payoff landscape seems fairly discoverable, but with high effort (which I, in the event, didn’t end up exerting) because of having to disentangle the interactions of the combinations of different spells and different mana strengths. The outcome being a probabilistic binary one also reduces the effective amount of data points as compared to a numerical outcome, which becomes more of an issue the more different effects you need to disentangle.
I think there is likely plenty of data for this once you take into account the symmetries of the problem, but maybe not quite so much without the symmetries (which are obscured by the different pick rates and the initial missing dark mana). I actually considered the possibility that the pick rates were the only asymmetry source, but didn’t look into it, whoops. Even so, I had been planning on proceeding under the assumptions that asymmetries between the spells were sufficiently unimportant that looking at symmetry equivalence classes of spell combinations would be useful, and that only offensive spells of one mage vs. defensive spells of the other and vice versa matter (i.e. that there are no important interactions between offensive and defensive spells of the same mage, and that there are no important interactions between offensive spells of different mages or defensive spells of the different mages), and also that only the mana types associated with the spells involved matter. In retrospect, this should have succeeded if I put in the effort, because in actual fact all of these assumptions happened to be exactly correct, but I didn’t actually know any of these assumptions to be true, which in combination with the effort that would have been required reduced my enthusiasm to proceed. I could also have been probing each of these assumptions, but that also would have taken time.
Or too simple to feel realistic?
Of course, but that’s not necessarily a bad thing.
given that multiple people commented on it very quickly I may have made it too obvious.
I had fun finding it (independently though reporting after abstractapplic), even though I literally had not looked at the actual duel results for anything further than the mana results themselves to click into place.
I also had fun in general, and even though I didn’t end up proceeding beyond a certain point, I expect I would have had further fun if I had done so.
The exact mechanism would have been pretty tricky to figure out
I was getting close. :) Had I spent 3x as long, I probably would have gotten it. Where I left off:
convinced it was very likely to be a repeated exchange from HP=X to HP=0, and thought the exact HP number would be static and “nice”
leaning heavily toward turn-based rather than simultaneous
had thought of a small handful of possible ways “damage” could be happening, one of which was this exact mechanism (convinced of x0 and x2 because Pokemon/etc)
plotted calibration curves for potential mechanisms
The things I was missing were:
automation to easily try out new damage mechanisms and then try all the (relatively small) combinations
notice that “choose randomly” and “choose best” didn’t work quite right, and figure out that it was actually “choose weighted” (this is the hardest part for sure, and it’s possible the calibration curves even look fine without it, which would scrap my chances)
Nice!
The exact mechanism would have been pretty tricky to figure out, but that’s not necessarily a problem?
The general payoff landscape seems fairly discoverable, but with high effort (which I, in the event, didn’t end up exerting) because of having to disentangle the interactions of the combinations of different spells and different mana strengths. The outcome being a probabilistic binary one also reduces the effective amount of data points as compared to a numerical outcome, which becomes more of an issue the more different effects you need to disentangle.
I think there is likely plenty of data for this once you take into account the symmetries of the problem, but maybe not quite so much without the symmetries (which are obscured by the different pick rates and the initial missing dark mana). I actually considered the possibility that the pick rates were the only asymmetry source, but didn’t look into it, whoops. Even so, I had been planning on proceeding under the assumptions that asymmetries between the spells were sufficiently unimportant that looking at symmetry equivalence classes of spell combinations would be useful, and that only offensive spells of one mage vs. defensive spells of the other and vice versa matter (i.e. that there are no important interactions between offensive and defensive spells of the same mage, and that there are no important interactions between offensive spells of different mages or defensive spells of the different mages), and also that only the mana types associated with the spells involved matter. In retrospect, this should have succeeded if I put in the effort, because in actual fact all of these assumptions happened to be exactly correct, but I didn’t actually know any of these assumptions to be true, which in combination with the effort that would have been required reduced my enthusiasm to proceed. I could also have been probing each of these assumptions, but that also would have taken time.
Of course, but that’s not necessarily a bad thing.
I had fun finding it (independently though reporting after abstractapplic), even though I literally had not looked at the actual duel results for anything further than the mana results themselves to click into place.
I also had fun in general, and even though I didn’t end up proceeding beyond a certain point, I expect I would have had further fun if I had done so.
I was getting close. :) Had I spent 3x as long, I probably would have gotten it. Where I left off:
convinced it was very likely to be a repeated exchange from HP=X to HP=0, and thought the exact HP number would be static and “nice”
leaning heavily toward turn-based rather than simultaneous
had thought of a small handful of possible ways “damage” could be happening, one of which was this exact mechanism (convinced of x0 and x2 because Pokemon/etc)
plotted calibration curves for potential mechanisms
The things I was missing were:
automation to easily try out new damage mechanisms and then try all the (relatively small) combinations
notice that “choose randomly” and “choose best” didn’t work quite right, and figure out that it was actually “choose weighted” (this is the hardest part for sure, and it’s possible the calibration curves even look fine without it, which would scrap my chances)
pin down HP=100 rather than HP=20, 50, 200, etc.