Psychochild: I think the “30% chance to avoid fear effects” randomness is a great example of un-fun randomness. You’re going to notice the times when it doesn’t work a whole lot more than the times when it protects you. At the very least, the game should add stronger notifiers (a special effect, say) when it’s working. At best, refactoring it into something more tangible makes sense to me.

I think it’s inaccurate to say that random isn’t fun. I think it’s more accurate to say that games that feel highly random are generally less fun than games where you feel like your skill has an impact on the game.

Even if the game doesn’t ‘feel’ highly random, there can easily be a bunch of issues. Playing a Storm Summoner in City Of Heroes usually doesn’t feel that random, and does feel very dependent on skill. How you use Hurricane, Lightning Storm, and Tornado is a lot more important than how any individual die roll goes, often to a vast degree. Randomness isn’t fun, though; the difference from a merely average series of rolls and a great series doesn’t feel any better than average, while the difference from a horrible series and an average is very damaging if not quickly fatal.

Game balance is all about statistics. The player hits enemies three times out of four for somewhere between x and 2x damage every N seconds, you can predict what the average time it’ll take to kill a baddy. That all falls apart the second you run into actual players, who care not about what happens on average but what they remember. Randomness means you will be randomly amazing and randomly screwed. You guess which is more memorable.

The appearance of randomness in a game is essential. Even if you can make a highly skill-based MMOs, you’ll need a bit of the appearance of randomness in terms of damage (most MMOs), powers available (card games/Phantom Dust), accuracy (most MMOs/most online FPS), encounters, and/or rewards. Otherwise, players will get bored. S4 League is an example of largely non-random play — only spawn points and other players are remotely unpredictable — and the game suffers from it. The Gauss Rifle, in particular, is oft-exploited, and the game is entertaining primarily if not only because other players add randomness.

Players have absolutely no idea what real randomness looks like, though. A true random number generator, the common seeded PRNGs, Psychochild’s Randomness without Replacement method, or even a few hundred preselected series of numbers, all can be made to ‘look’ random for even the best player’s memory and unpredictable on any relevant scale. You are, after all, dealing with the Magic Number Seven, plus-or-minus two.

The goal, first and foremost, needs to be about what is fun. If killing 40 Yeti isn’t fun, why are you telling people to do it and calling it a game? If killing 40 Yeti is fun, why have half the players kill many more Yeti than that?

I wasn’t a fan of having to wait for the low % drop rate yeti item to drop either, but it sure didn’t hurt me. A better way Blizzard could have resolved that issue was to intertwine other quests by the yeti area, but it didn’t work that way at that time. In WotLK (from the little I’ve played [Note: Do not end up in the hospital and miss two weeks of college, when you're already on an overload schedule]) quests are much more intertwined in the same zone, and frankly the drop rates are probably too high and easy, although the game is now called Casualcraft for a reason. =s

The Wi flag was a real issue and not superstition though ;)

One problem in our case is that I added a feature where spawns will increase as more players are in an area, so they won’t have to fight over spawns. This, of course, lead to people planting extra characters in an area to increase the spawn rate significantly. Of course, they also got used to those increased spawn rates and complain that they are now “forced” to have extra characters around.

In the end, player perceptions dominate. Of course, it’s still important to investigate reports of strange randoms. There have been multiple times where a bit of bad math actually did change things in the opposite way it was intended to.

There are two developer approaches to this. First is to look into “random without replacement”. Think of a deck of random cards instead of a rolling a die. You only re-shuffle the deck once a number of cards are drawn. For a very small deck, this may not be significantly different than a die, but it can eliminate long streaks. I posted a bit about this on my own blog at: http://www.psychochild.org/?p=102

Another idea is to make sure you don’t over-use randomness. In WoW, I play a Feral Druid and took a talent that was supposed to give me a 30% chance to avoid fear effects. Sounds great, but it certainly didn’t feel like I a roughly 1 in 3 chance to avoid a fear effect. It felt like wasted talent points to me, so I ditched that talent. In the recent patch, that ability was changed to reduce the duration by 30%, a non-random effect. What about combining the non-random effect with a random effect as well?

My thoughts.

In Magic, rare cards are not always better than the uncommons and commons. (Not to mention that Magic encourages lots of different ways to play.) So even a poor kid can build a strong deck with uncommons and commons. On the contrary, a deck with all rares are not necessary playable.

Speaking of Magic and randomness, MTGO players also complained about streaks of lands or spells. The programmers replied that there is nothing wrong with the random number generator, that it does indeed spit out totally random numbers. So the issue here is, as Eric pointed out, the players’ impression about the RNG. In this case, it seems that being too random is not desirable.

A secondary market takes that lottery mechanic and gives it player demand driven value. Why are high level items in WoW Bind on Pickup? I see it as a way to encourage the loot crapshoot, which is something that I’d rather get rid of.

These rare loot drops shouldn’t be rare because of an arbitrary drop percentage, they should be rare because it’s difficult to do the task requested. That would give a real sense of accomplishment, rather than being the bloke who pulled the Powerball. The player who comes by it first hand via skill would be rewarded for their skill, while the player who comes by it on the secondary market would be rewarded for their effort spent earning the funding to make the purchase.

From casual observation, any time-consuming random that has a success chance below 10% is likely to trigger the creation of “interesting theories” – and, basically, player frustration and annoyance that they cannot get want they want within a “reasonable” timescale.

I’d love to see more “progressive de-randomness”, building on CoH’s streakiness: Each time you do the same thing, the probability of successfully completing it that time increases from the previous time. That rewards a player’s time commitment, while still allowing an (initially) low-probability outcome to be implemented.

I presume that hasn’t been done historically because it requires vastly more data to be stored – we’ve just got stuck with mechanics that were designed to work on minimal hardware.