Improving the balance of the RPG

[MaksiRose]

Okay, so I'm doing this as a discussion because I want to be able to revisit my thoughts later on in case I want to add to them or want to see what else I thought that I might have forgotten until then.

So the RPG has a huge balancing issue. And the way I see it this is down to two factors. The first is about skill, the second is about randomness.

First, let's address the skill issue. The games you can play don't have difficulty settings like easy or hard since it's a community game, and one person making it too easy or too hard for themselves might ruin the fun for others. But not every player is the same, and having one single difficulty setting means that some might have no challenge at all, while others are struggling to get anything.
This is also accompanied by disabilities. There are three kinds of games, the ones that test speed (i.e exploring), the ones that test logic (i.e scavenging), and the ones that test memory (i.e recovering). In favor of accessibility, people should be able to toggle up to two of these kinds of games off in their settings, which would then be replaced by games of similar difficulty but asking for a different skill. Additionally, the games could also be made "smarter" and kind of track how well a player is doing as opposed to how well a player should be doing, and adjusting itself to be harder or easier accordingly. But this itself is difficult, since it requires balancing to determine how well a player should be doing. This would need thorough testing, for example by making a version of the bot that does not feature these games, but simply gives you the win, neutral or fail result based on the percentages given. If the result is not as wanted, the percentages could be tweaked until a good result is given, and then of course the bot has to be smart enough to know how to adjust the difficulty in such a way where that percentage is actually reached as well. This is where this method would get actually hard to implement.
Another important aspect is though to make sure that the player really understands what they are doing. Things like the play tutorial would solve this, clearer explanations should be given to the player in general.

Second, let's address the randomness. The problem is that randomness still makes up a huge percentage of gameplay, to a point where it is not fun anymore. While randomness should not be removed completely as it does make the game more interesting and introduces a form of excitement as you are not sure what is going to happen, it should be tweaked in such a way where it does not allow players to be screwed over completely, nor to be completely overpowered. #534 for example would make the game much less random, as your input directly correlates to your chance of winning. To counteract finding more items than you need to use, the enemies level would be determined based on how well as a pack you are doing. Meaning, if you have many items but little use for them, the enemies would become harder, but if you have little items but much use for them, the enemies would become easier. None of this is random, and part of the problem would become that players might feel cheated if they always get enemies of the same high level. Things like the bot picking randomly from a range of levels around the ideal level the enemy would have could solve this. Another issue is that the player might feel like they are being cheated if the enemy levels get increasingly higher and higher. This is where things like adjusting the gameplay based on the players skill would come in, as mentioned in the previous section.
Another important part is what kind of items the player gets. This could be done by giving a weight to each item, where a higher weight equals a higher chance of getting it. The weight of each possible item should be determined based on how much they are needed and how many of them are available. For example, when scavenging, you can either find meat or material. It could find every carnivorous or omnivorous player that doesn't have full hunger (or is a certain number of points away from full hunger) and could simulate each of them eating, and then compare that with the amount of meat present in the packs inventory. The more meat the pack has over the closer the weight would go towards 0, and the more meat is needed the closer it would go towards 2. Having the exact right amount would give a weight of 1. The same could be done materials. Problems with this approach is determining exactly how to simulate that, as the amount needed for some things is based on the player, i.e healing or repairing. I don't know how to fix this yet.
Which brings me to the next issue, healing and repairing. The likelihood for these should also be less random. How exactly they would be made less random, I don't know yet.

[MaksiRose]

Hi, I'm following this up with some more thoughts :)
To exactly figure out which level the enemy / plant should be, the following logic could be applied:
Each time you win a round of the game, you get +4 points, and each time you lose, you get -4 points. For three rounds, that means you can end up with 12 points at best or -12 points at worst. The difference between your level and the level of the enemy would also be counted in points. The maximum difference in levels can be 12. The calculation for what outcome you get is calculated as: getWeightedTable({0: -1 * points, 1: 12 - Math.abs(points), 2: points}), where 0 is negative, 1 is neutral, and 2 is positive outcome. If a negative number is assigned, it is treated as 0 (so if {0: -1 * -3, 1: 12 - Math.abs(-3), 2: -3}, then its treated as {0: 3, 1: 9, 2: 0}).

Now to find out what level the player should play against. First, a simulation is done that involves the servers inventory and all the servers players. This simulation would simulate everyone using the items in question (in case of enemies, meat, in case of plants, plants) to fill up their stats. This would involve things such as the dens breaking down further (and the food giving less hunger as a result, and the chances involved when healing. This simulation might need tweaking in the future, as some things must be assumed, such as: would someone repair the dens? Who would heal who? etc. Information about users might be stored in the database in the future that answers the question.
The simulation would now figure out how much items would be needed to fulfill everyone's needs, and can compare this number with the amount existing in the inventory. If the numbers are exactly the same, the level would be the same as the players level. If 12 or more items per person are available than needed, the enemy has 12 levels more. If 12 or more items per person are needed than available, the enemy has 12 levels less. Everything in between is handled accordingly.
This number might need to be tweaked in the future.

We repeat this simulation, except this time it only applies to profiles who have been active since a certain amount of time. This could be for example two weeks. This amount of time might need to be tweaked, and which commands override the activity to be newer might be tweaked as well.

Now we have two simulations, allPlayersSimulation and activePlayersSimulation, both returning a number between -12 and 12. Finally, the level of the opponent would be calculated as: playerLevel + Math.round((2 * activePlayersSimulation + allPlayersSimulation) / 3), which is to give activePlayerSimulation double the weight when calculating the middle amount needed. Also, the actual level the player is going against might be a bit more randomized but still close to this originally calculated number, maybe between 1 more/less (except if its 12/-12 in which case it should not go to 13/-13. Again both the active/all Player Simulation weight plus the randomness to it could be tweaked.