There are new versions today! Version 0.2.4 is available for free, and the Supporter versions of 0.2.4 and 0.2.5 are available for their respective Patreon tiers. Check their changelogs here and here.

Long term tests and NPCs/AI balance

On the last devlog entry, I explained the basics of the Character Drives mechanic. There's a fundamental difficulty with implementing such a mechanic: since its effects may only be observed over long periods of time, it cannot simply be tested as soon as it's coded.

To put an example: I recently introduced changes in drives as the result of generic sex scenes, that will make characters gain pleasure drive the more satisfaction they get from sex, but will I be satisfied with how this will translate into a different character behavior? It may be the case that, over a few in-game months, all NPCs tend to grow into hating sex, or getting obsessed with it, but playtesting several in-game months each time I'm going to release a new version is impractical. This is why quite a few weeks ago, I coded functionality to let the game run by itself with the 5 NPC Candidates over long periods of time. These tests are turning out to be fundamental for me to understand the flaws in the game's balance. These are some of my insights:

- What's probably the most noticeable issue is that, over time, Maaterasu *always* tends to be winning. The question isn't so much who will have the merit lead, but how much merit lead will Maaterasu have. This provoked the first nerf to Maaterasu in an earlier version, and it apparently wasn't enough!

- On the other side of the spectrum, Claw used to get her ass beaten - a lot. This was reflected by her gaining a lot of self-improvement drive, which happens upon losing fights. After a few tweaks to her skills, she is now fairly on par with the other Candidates.

- Self-improvement usually grows a lot for the weaker characters, but this doesn't usually reflect on those characters becoming stronger. There are a few possible causes here: NPCs already prioritize training during training period enough to make the potential gains during socialization period marginal, stats' gain becomes far too slow upon entering mid-game, or there simply aren't enough alternatives to spend a significant amount of time training during socialization period.

- In the last few tests, there were instances of fairly disparate love and pleasure growths. Specifically, Valtan was gaining a lot of pleasure while other characters were losing it. This could be the result of Valtan focusing on herself during sex so much that her partners don't usually reach orgasm - and this is an interesting outcome, it's cool that there are emergent mechanics that provoke disparate results for characters that start in different situations - but wouldn't Valtan scheme to get her partners addicted to her?

- Cooperation, domination and ambition drives usually grew in different portions for each character in each run. This looks generally ok to me.

I'm not going to touch the numbers just yet, but being aware of these results will definitely be useful when I get to work on the long-term social goals for the NPCs' AIs.

I'm going to attach text files with the infodumps of a few of these tests in case that anyone's interested, even though they're not easy to read. The order in which drives appear are "Self-improvement", "Pleasure", "Love", "Cooperation", "Domination" and "Ambition". The first number is the drive's value, the second number is the resulting level. The formula to calculate the drives' levels got changed as I made more tests, which explains why different values give different levels later.