This Thursday I was just doing what I do best, wasting time on the internet when, after a few mentions of apparently bugged hit chance on Demolish, this tweet floated across my feed:
The high hit rate on Demolish is part of an intended system to prevent miss streaks. It's just too effective on high miss chance abilities.—
Jonathan LeCraft (@TheCrafticus) February 21, 2014
My ensuing Twitter discussion with Crafticus follows:
Me: How long has this “intended system” been in place? Because this changes every little bit of percent-based decisionmaking.
Me: I thought I was playing in a world governed by the laws of probability, now I find out God cheats at dice?
Crafticus: It’s been in since launch in one form or another.
Me: Is it part of the reason we used to have “high, moderate, low” instead of numbers?
PET BATTLE HIT PERCENTAGE IS A TOTAL LIE
YOU HAVE NEVER BEEN FLIPPING A COIN WHEN YOU USE A 50% HIT CHANCE ABILITY
Ok. Time to stop yelling. I’m going to back up a couple steps here and talk about the nature of randomness.
In everyday life, we use numerical probabilities to refer to single events often. “There’s an 80% chance of rain tomorrow.” “There’s like a 90% chance that those two are secretly dating.” “There’s only a 50% chance I’ll make it to your birthday bar crawl.”
However, in retrospect, an event either happened or it didn’t. Those “probabilities” were completely meaningless. I made it to 0% of my boss’s birthday parties in 2014. It rained here on 100% of this Thursdays. Those adorable kids in your guild are either madly in love or just friends. Etc.
The most meaningful way to talk about the probability of events that ACTUALLY OCCUR is through observation of many similar events. For example, in Dilbert’s office back in 1996, there was a 40% chance that taking a sick day would result in a three-day weekend:
Bringing it back to WoW: Let’s say that every time you used Demolish (when no accuracy buff or debuff was active), you wrote “H” for every hit, and “M” for every miss. (You probably had to do this kind of “experiment” with coin flipping in math class at least once in your childhood.)
The ONLY way for the tooltip hit percentage chance (50%) to be correct and meaningful to pet battlers is if, as the sample size grows larger, the number of hits and misses approach each other. However, we know that, in the game as it stands, this is emphatically NOT the case:
Of course, when I say that probability is only about observation, not prediction, I am lying a little bit. I am only talking about “real life” as humans experience it. When talking about GAMES, this is often a lie. I don’t need to shuffle thousands and thousands of decks of cards to know that there is a 1 in 13 chance the top card (or the bottom card) is a king. Any well-balanced d20 will give me a 13 exactly 5% of the time. And every single time I flip a coin, I have a 50% chance of it coming up heads.
But here’s the thing about the coin: Each flip is independent of the others. Here’s a great example from a Wired article from a couple years ago:
Imagine a professor asks her students to flip a coin 100 times. One student diligently did the work, and wrote down their results. The other student is a bit of a slacker, and decided to make up fake coin tosses instead of doing the experiment. Can you identify which student is the slacker?
The second one “looks” more random–however, the first has actually been experimentally derived, while the second is made up.
Because randomness? Doesn’t give a shit about you. Randomness is crueler than your abusive ex. Randomness is BY ITS VERY NATURE completely inhumane. The lottery is just as likely to come up “4-8-16-23-42″ as it is to be “1-2-3-4-5″ as it is to be whatever dumb numbers you happened to pick.
Of course, we don’t like to live in a world we can’t control. Even the most staunchly atheist gamer (I’m talking about myself here) might blow on their dice before they roll, or whisper a prayer to the little babby RNGsus. We notice patterns where there are none: Random effects in Hearthstone, for example, are always supposed to resolve in pro player Gnimsh’s favor. And of course, we develop superstitions like the Gambler’s Fallacy: When we’re on a streak of tails, the next flip is more likely to be a head, right? After all, we deserve it. The universe owes us just one win.
But that’s just superstition talking. The universe owes us nothing.
Here is what random numbers are for in games: They are kingslayers. You may well be better at Scrabble than your nephew, but if you only draw vowels, it’s pretty hard to make a high-scoring play. MMO loot rolls are completely impartial, ignoring the deserving and rewarding only the Greedy. Dumb kids’ games like Chutes and Ladders, where “You roll the dice, and the game plays itself,” take this to the extreme–they give parents and older siblings no advantage over a three-year-old who just learned to count.
Bringing it back to pet battles: “miss streak prevention” is the opposite of randomness. Where RNGsus is capricious beyond measure, Blizzard is attempting benevolence. When we use low-hit-chance abilities when we pet battle, the gambler’s fallacy is no longer fallacious. The more times Demolish fails to land, the more likely it is to land next time. Azeroth is a world we pay to participate in; it is also a world that is, I just learned, bent around gamers’ superstitions.
It’s not hard to guess that “miss streak prevention” is in the game because losing streaks are just no fun. But shouldn’t the higher damage of low-hit-chance abilities be doing some psychological (as well as mechanical) balancing work?
I don’t pretend to be a programmer or a statistics whiz; I’m just an art-school dropout who got a lot of A’s in math class. But I can see two different ways for “loss streak prevention” to work–either instead of flipping a coin, the game could use a “Student 2″ type model:
Or, it could start out flipping a coin, “Student 1″ style:
If we just take this sample, and shorten the “miss streaks,” we get something like this:
I didn’t use any fancy algorithms like Blizzard is probably using to do this–I just used my intuition about what looks really unfun and the delete key. In my “miss streak prevented” results, 24/44, or 55%, or my attempts to Demolish my opponent hit–as opposed to 24/48 (50% on the nose) in the original sample.
Here is the ONLY WAY to make hit chance 50% while also preventing miss streaks: You HAVE to also implement “hit streak prevention.” But that’s kind of a problem–because while losing streaks aren’t fun? Winning streaks are AMAZING. They are the definition of fun. Demolish–or any other high risk, high reward gaming strategy–isn’t supposed to work out for you all the time. It’s supposed to sometimes fail, often be sort of OK, and every once in a while work out amazingly well and win you a game.
“Miss streak prevention” makes some sense for (say) abilities with a 95% hit chance. After all, in a truly random situation, you could still technically miss 20 times in a row with it (though the chances are vanishingly small).
But the lower the alleged hit chance–the lower the hit chance the first time the ability is used–the harder it is to avoid miss streaks while still allowing hit streaks. (I imagine that there is some kind of breakpoint, where abilities that have been debuffed to a below-50% hit chance can no longer avoid miss streaks, but “guessing how this works” is not really my focus here.)
As you might have surmised from one of my earlier tweets, the reason I reacted so viscerally to this information is that I feel like I’ve been punished for acting “rationally” all this time. In avoiding thinking I believed to be “fallacious”–by avoiding low-hit-chance abilities, by behaving as if I was rolling dice, by avoiding low-hit-chance abilities in most of my builds, I’ve apparently been punishing myself.
How I see it, there are two ways to make the tooltip of Demolish match its behavior. They could give us either the completely unpredictable behavior of the coin flip–or they can raise the lowest base chance to be greater than 50%. Loss-streak prevention is incompatible with a 50% base chance to hit.
I for one would like the former–I would rather live (and, yes, game) in a predictable world, governed by the laws of probability I learned when I was 11: A world where, yes, the “random number” has a tiny chance of being nine forever. I would prefer a game system that can be understood, theorycrafted, even emulated & duplicated outside the actual engine, even if it occasionally just totally screws me over, to a game system where some “benevolent” force is so concerned about my “fun” that the gambler’s fallacy is actually enforced. Knowing that somebody out there is on my side just lessens my sense of accomplishment and joy. Luck is a drug; being given presents can get kind of boring.