How Strategy Games can use Output Randomness

Been thinking a lot about output randomness and what it means in games. My overall opinion on output randomness has not really changed much, but I think from the process of working on Gem Wizards Tactics, which does have some light output randomness, I have come to some new conclusions about when and why it’s a problem—and perhaps more usefully, when it isn’t. To put it another way, people who I had long argued with on the topic of output randomness weren’t 100% wrong, but they were maybe 80% wrong.

Level of impact

In games like XCOM 2012, the level of impact that output randomness has on the state is extremely high. In this game, every time you shoot (the way you interact with anything 95% of the time in the game), you have a % chance to hit. Hitting an enemy often instantly kills them, removing them completely from play. Missing an enemy does absolutely nothing to them. So the RNG that determines this is determining a lot, and this effect compounds on itself. As I’ve talked about before, output randomness has a way of “severing” the ties from state to state, and thus the system meaning that could otherwise be built.

In other words, I bring some troops out, shoot at some aliens, some hit, some miss. I do another turn, and another. At this point, why is the state the way that it is? Well, partially because of choices I made, but also largely because of the sheer noise of the output randomness.

The other thing to be concerned about is that output randomness of this sort is almost always highly variable randomness. Yes, if you shoot enough times, eventually it all starts to move towards the mean, but for smaller collections of shots, which encounters are made up of, it could be that you just get 3 misses in a row and that could be enough to completely screw you over. Or you could get three critical hits in a row and now you’re gonna steamroll.

EDIT: Another thing I forgot to mention about impact in these kinds of games is whether “damage taken” affects things at all levels. In XCOM, your soldiers do full damage even if they have 1 health left. In Gem Wizards Tactics, and in Advance Wars IIRC, the unit’s attack damage scales down with their health. This means that all attacks “mean something”; they don’t have to kill in order to start changing the landscape.

In Gem Wizards Tactics

As I mentioned, there is output randomness in Gem Wizards Tactics. Basically when you attack another unit, it tells you the estimated damage (say, 3 damage, out of 10 total health). The actual resulting damage is likely to be 3, but it could be plus or minus 1 or 2. And then on top of that, there’s another rule related to retreating that caps damage taken from an attack at 4 in most situations (if the unit takes more than 4 damage from an attack, it tries to retreat; if it can, it caps the damage at 4). So it’s overall very light, the range of possible outcomes, as compared with something like XCOM. There is no “missing”, and on top of that, any damage or other effects from spells and abilities is entirely deterministic.

You might ask, why have output randomness at all in GWT? This is where I have to give some credit to the output randomness defenders over the years. They are right that by putting an interaction into the hands of a random dice roll, you minimize its role in the skill. Auro is a pure tactics game, it is all about the tactics themselves, there are not really any longer arcs, and so I want to highlight those tactics as much as possible. In GWT, by comparison, there are longer arcs, there’s a larger map with objectives, there are rushdown and econ strategies and so on. So because of that, I want to (at least, slightly) minimize the tactics.

Furthermore, some RNG in the combat—even a small amount—minimizes calculation. In an article I wrote on that subject awhile back, I was largely thinking about ways to use input randomness to achieve this goal. I was still close to working on Auro, a pure tactics game at that time, and in that context, it’s true: you should avoid output randomness. But in the context of a larger system, tactics can be minimized somewhat with output randomness, while safely preserving the larger arcs. I think this is something that could be very useful for designers building larger strategy games, who don’t want tactics to completely dominate everything (as it would in say a Civilization with 100% deterministic combat), but who also don’t want tactics to become a huge meaning-destroying noise machine (as is the case in the vast majority of games that use output randomness).


Thanks for reading! If you enjoy this article, consider becoming a Patron on! Patrons get access to early prototypes, bi-weekly builds of Gem Wizards Tactics, special Discord roles and more.