Comment by PlunderBunny

9 hours ago

I figured out that, in the early versions of Realmz for the Mac, there was an enchanted item (a helm?) that added a few points to the wearer’s stats when equipped, but when removed it took away more than it added (i.e. left you worse off). It only took a few seconds for my programmer brain to start me furiously equipping and then un-equipping it, and sure enough, when that stat went to -127, one more equip-and-unequip was sufficient to give me a +128 bonus for that stat.

I never understood how that particular bit of logic could have been coded that way - it seemed so unnatural.

I would figure the reason was because calculating it based on your inventory every time it's needed would've been expensive, so the developers decided to cache it, but instead of recalculating it from scratch when there's a stat change, they decided to try updating it incrementally, but I guess also imperatively. A string of mistakes has to happen to lead to such a situation but it doesn't seem super unnatural to me.

  • I wonder if it was percentage based. That'd be one of the easiest mistakes to make.

    - Player has a stat of 100

    - Wears "+10%" helm

    - Player has a stat of 110

    - Removes helm, game subtracts 10%

    - Player has a stat of 99

    Actually, now that I think about it more, that wouldn't work as described above because once your stat got close to zero you'd enter a kind of Zeno's Paradox situation.

    • it's more likely that the stat for equipping is calculated differently to when unequipping (ala, "same" code in two different places, and one got modified at some point and forgot about the other).

    • > I wonder if it was percentage based. That'd be one of the easiest mistakes to make.

      How would that be one of the easiest mistakes to make? There are no operations for working with percentages, nor is it a thing you'd want to do.

      To make that helm work as you describe, you'd need to multiply the stat by 1.1 when equipping the helm and then by 0.9 when deequipping it. But at that point you've lost any justification for making the mistake; those are just two random numbers that don't cancel each other out. It's not like you're adding a number in one place and then erroneously subtracting that same number somewhere else.

      (Also, these are very strange multipliers to use this way, because they can never produce exact results. You're bound to see rounding errors if you insist on update-when-equipping and update-when-deequipping instead of recalculate-stats-when-equipment-changes.)