Over the weekend, I had some friends of mine playtest Eon Altar. One player was Marcus, our tank, and she expressed frustration that she couldn't hit anything. I believe the direct quote was, "Blocked? Fuck you! Half my shit ain't landing!" From what I could tell watching, she wasn't really wrong. We checked Marcus' stats after, and he had the correct values, so therefore should have approximately 75% to hit enemies in the first level.
|Our desktop wallpaper for Marcus. He's our tanky character.|
The Effects of Missing Over Time
Of course, with any randomness comes the opportunity for it to fall outside the curve over short time periods. The more attacks you make (samples), the closer to the actual curve you'll probably get. When you only make 3 or 4 actions per combat, missing 3 out of the 4 attacks sucks. It's not super likely, but it's also not unlikely (~5% chance to miss at least 3 attacks of 4 at 75% accuracy). It's quite plausible that our Marcus player had hit upon that 5% over and over again.
Your memory is also tainted by the fact that humans tend to remember negative results far more often than positive if they're all of relative equal intensity (known as negative bias), so even if Marcus was only hitting half the time or less (~25% chance per set of 4 trials), it would likely be perceived as worse than it actually is.
Alternatively, when you look at a game like World of Warcraft, if you have a 5% chance to miss, but you're making anywhere from 30 to 40 attacks per minute, that 5% gets forgotten pretty quick. It's not a high frequency event, nor is it devastating because you have so many actions. Compare that to D&D, where often you're looking at 25% to 50% chance to miss on each roll of a d20, and you only have 3 - 8 rounds of combat.
Such high variability may cause balance issues, as well. A group that is lucky might blast through a level because they just crit everything. Also see: top parses for WoW. Here's an Enhancement Shaman who's one of the top DPS parses for Mythic Gruul. Note their abilities have crit percentages of 33% to 45%, despite the character only having 20% critical strike. Take enough samples, and eventually you'll have outliers.
On the flip side, a group might struggle immensely because they got unlucky. Everybody missing a lot, taking excessive amounts of damage because they're getting hit more often as well.
With sufficient variability, it becomes difficult to predict the outcome of a combat or a level. This is both good (because having to readjust tactics on the fly is where some of the fun of combat comes), but it's also bad (because as a designer, the experience becomes more difficult to control).
But is that a bad thing overall? Having a lucky (or unlucky) streak makes for interesting stories. In D&D the time where a group roflstomped the villain because they had a streak of criticals. Or the other time they barely escaped with their lives because they couldn't hit the broad side of a barn. Players love both of these stories. Games where the experience goes off the rails tend to be quite memorable.
Mitigating the Extreme Negatives of Accuracy
There are also other techniques one can use to dampen the effect of accuracy. Every time you consecutively miss, get a cumulative bonus to hit. This helps reduce the effect of unlucky streaks (as eventually you'll hit for sure). Granted if players know this is in effect, they can game it--"I missed twice, my next hit will hit for sure, time to use my big attack!"--but on the other hand, how many folks "game" luck that way anyways? "I've missed so many times, I HAVE to hit next time, it's so unlikely to miss again." Encoding it into the game might not change behaviour at all.
Another method is to still give some sort of effect when you miss. D&D 4th Edition does this with Daily powers. When they miss, they generally still do half damage and have weaker forms of the effect it would normally cause. Missing still sucks, but at least you haven't completely wasted a very limited resource.
Finally, you could just boost the accuracy of players. If you only have a few events per combat, rather than having a 70% chance to hit, maybe a 90% chance? You'll still have unlucky streaks, but they'll be less likely.
Missing is frustrating. You as a player feel impotent, and when you only have a few actions a combat, each miss has a massive impact on your play. As designers, accuracy is a very easy tool to introduce variability in game play, but we also need to be careful. Balancing fun in this case is very subjective.
Overall, game designers needs remember this: the fewer events, the more variability. The more events, the more likely your event results will mimic the probability curve. When building a game with few events, variability drives the game experience. Whether that's something you want is a question only you as the designer can answer.