Context: Shadows over London is a turn based strategy game in which the player controls five survivors and tries to help them escape a destroyed London. The five playable survivors and lovecraftian monsters the player encounters all have unique skills.

Hidden Mechanics and Player Psychology

Whether an attack is successful is determined by a percentage chance that is shown to the player. Even though this chance is already rounded down we still got feedback from our testers that this system felt somewhat unrealistic and often frustrating. To figure out what exactly the underlying problem was I did interviews with our players as well as research on risk management. After realising most people don’t have a natural intuition for how probabilities work we could set design specifications for our system:

  • Avoid too many fails in a row
  • Create tense situations
  • Make players feel “lucky” but allow them to strategize

In the end we added a few hidden mechanics that always give players the advantage and the monsters disadvantages. Here are a few examples:

  • Giving player characters a small dodge chance
  • Adding an accumulative bonus chance after missing likely shots which gets reset after hits
  • Capping monster damage output in a turn according to number of survivors left

The intuitive understanding of chance most players will have likely is that a one out of three hit chance means one hit every three attempts. So our main goal was accommodating this understanding of chance while at the same time not allowing the game to become too predictable.

Fine Tuning using data

When designing the level, it was divided into spatially separated encounters to get better control over the pace of the game. It also simplified understanding what possible situations the player could get into. For me personally plugging the numbers into a spreadsheet to calculate things like expected values helped a lot to understand what all our available parameters did. Some of these numbers had to be gathered from playtests.

Italic values: measured from tests
Bold values: calculated from spreadsheet

Other balancing aspects (like how much medkits heal) were tweaked using graphs as visual aids:

An animated version can be found when clicking on the image

Creating engaging Gameplay

Before designing the level a sort of script was devised that would be the base for modelling the level. In other words a dramatic arch was created first and then the level was modelled around it. To help understanding this arch I visualised the events on a timeline like so:

Mission Timeline

After more playtesting we noticed that the nurse was only used very carefully by players which slowed down the pace enormously. It also resulted in two more issues:

  • If the damage dealer characters die then the game gets extremely difficult
  • Using the nurse exclusively as a healer created not engaging gameplay

In order to understand why players were actively choosing a playstyle that was less fun we conducted interviews with playtesters and found out that we could do two things to solve this issue:

  • Change how we communicate the nurse’s abilities to players
  • Move away from strict role archetypes

At the end some abilities were rebranded and moved to other characters to allow for more diversity in each characters skillset. Grouping skills to their characters while colourcoding their original use shows how much more variety was introduced. As seen characters that did not fit in well with the playstyle we wanted to provoke received most changes.

Collection of abilities for each characters

Below is the trailer we made for the project. More information on how the game works can be found on my blog.