The Gollop Chamber – the Good, the Bad, and the Ugly of RNG

In the latest installment of his PC Gamer column The Gollop Chamber, Snapshot Games’ Julian Gollop talks about the use of RNG in video games and how people’s perception of how random numbers should work affects their enjoyment of games that utilize them, forcing the developers to occasionally “fudge the numbers” to please their fickle audience. A snippet:

It was based on a ZX Spectrum game I made in 1985 called Chaos and published by Games Workshop (a company not shy of randomness itself—usually in the form of buckets of dice rolled in their army games). In Chaos Reborn you play a wizard, and the casting of spells and combat were subject to binary outcome randomness. The combat is especially brutal, with a success killing a target outright and a failure doing nothing at all. Even a lowly giant rat had a 10% chance to kill a dragon.

Many players loved it, and to them the extreme use of RNG is what made the game different and unique. The tension of each battle was enhanced by the rapid turnarounds, plans thwarted, and opportunities grasped. However, the game is certainly not without a high skill factor, since the best players frequently topped our monthly leagues. The fact that games were short, and very many games could be played in a month, either asynchronously or in live matches, meant that any element of luck would even out.

Unfortunately, a significant group of players loathed the RNG aspect of Chaos Reborn with a vengeance, and they wrote about it copiously. Our Steam reviews were suffering, mainly from this single issue. We did our best to warn potential players that the game used a lot of RNG and required a lot of risk mitigation, but to no avail. So eventually I decided to make a reduced RNG mode for the game which did not use any randomness in the spell casting or the combat. It created quite a different experience which played quite well.

The negative Steam reviews faded away, but it is debatable whether it made a more interesting game. Long-term Chaos Reborn players still preferred the more random ‘Chaos’ mode. On reflection I think there was probably a better way to manage the randomness in the game, and other games have dealt with the RNG problem in different ways.

Firaxis’ XCOM: Enemy Unknown also suffered from an anti-RNG backlash, due to its explicit hit percentage mechanics. “That’s XCOM, baby,” was the answer creative director Jake Solomon is reputed to have given. But in XCOM 2 the RNG is subtly manipulated to make it feel less random, at least at lower difficulty levels.

I now understand that human beings are not very good at evaluating probabilities. In particular when an RNG generates repeated sequences a human will cry foul. For a human, randomness usually means ‘evenly distributed without any detectable pattern or repetition’. This is basically how random numbers are manipulated in many games to meet player’s expectations. One poor result immediately results in a bias towards a better result.

Now some of you might want to point out that computer random number generators are actually only pseudo-random. They generate a repeatable sequence of numbers with an even distribution, but they are very, very long sequences. In practice a human would not be able to tell the difference between a decent computer generated pseudo-random number and a more genuinely random number. This doesn’t reduce the suspicion of computer RNGs though, which is another problem that game developers have to deal with.