Last week we talked about the AI in Haxity. Some of its shortcomings and what we are aiming to improve. You can read the article HERE. We are now deploying the new and improved AI(full patch notes at the bottom). This patch is the first, with more to come, to address some of the issues the community have discussed.
We have changed a lot of the algorithms to make the AI “smarter”. The AI will now place higher value on certain modifiers like Starter, Finisher and Adjacent making their general placements stronger. This will lead to the AI being more consistent and therefore more predictable in its behavior.
In addition to this certain AI opponents will put specific preference on certain card color, meaning they will lean more into their primary color if possible with their given hand. This will again make it so if you face a Skill/Yellow heavy opponent(like Sewer Rat) it will not be uncommon to see a full Skill/Yellow placement.
We also did an overhaul on how the AI utilizes hacks. We have many examples where the AI was using a “Momentum” Hack on a temporary card or using “Lock” after you have already passed. The AI is now playing smarter with the Hacks they have, and are now mimicking a human opponent to a larger degree. You can expect the AI to do proper setups, baits, feints and in general not waste hacks.
That said, we have noticed that many people found the campaign to be harder than intended. When we initially made the different AI decks we had certain synergies in mind and made them lean heavily into that. Since this is a deckbuilder you will not always be able to build as strong synergies as the AI had at different points in the run, making it feel like you were “cheated” out of a win. Since we are now making our AI smarter in a lot of ways, we are also taking measures to make it much more predictable and less overpowered. We are tuning down the Threat 0 and Threat 1 decks with less hacks and more skewed towards a certain color. This meaning that you can more easily counter draft and placement predict based on the sheer amount of one color in the AI deck.
We will continue improving every aspect of Haxity, and we would love some more feedback on this patch and how the game now feels. If you want to chat directly with the developers, most of us hangs out on DISCORD, and we also check Steam discussions daily.
Hope you all have a good day, and see you in Haxity.
Here you have the complete list of features included in the AI Uprising update:
- Improved algorithms on how the AI values card slotting, value of card positions.
- AI is now better at playing hacks, and understands the effects of hacks much better and can also try predict some opponent responses.
- AI is now open for specific preferences, as in we can define a specific AI opponent to prioritize some cards above all else.
- Campaign opponent (AI decks) changes:
- Tons of changes to AI opponent decks for Tier 0 and Tier 1: AI has slightly fewer hacks.
- AI has much more skewed color balance in their decks, making them always have much more of their strategy’s main type of moves (easier to counter play).
- Updated AI opponent descriptions.
- AI opponents have a stronger preference to play key moves of their given strategy.
- Updated Banshee starter decks for Versus, half of them got changed to now better reflect Banshee after her changes.
- The Ripper and Cyber Rats now display number of hacks played.
- Eruption scaling nerf: 8 dmg with +1 dmg per infect stack (from 6 base dmg, +2 dmg per stack).
- Hardsuit nerf: 6 shield instead of 8 (we don’t want it to be mostly just strictly better than Combat Shield).
- Corporate Coverage disabled (will be changed) Beefcake now +16 health fixed (circumventing an issue we have with % health).
- The Ripper has expend when it’s a finisher (was always intended, just forgotten to add)
- Tutorial is now an option available from the main menu, not campaign