I feel like you are underestimating how bad the human brain is at dealing with randomness. Other people have mentioned it, but humans really are utterly, utterly incompetent regarding anything to do with on the fly probabilities. You could simulate the average human’s final hour-long-considered decision of whether or not to attempt something pretty much with a coin flip. We seek out patterns everywhere, which helps with giant decision trees but can cripple us when we start trying to impose patterns on actual randomness.
I think in a way this could make it easier to code an AI for BB than for Chess or Go, because if you added in some standard human favorite strategies and set up the obviously proper order in which to do actions you don’t have to plan ahead for future turns. Attempting moves in order of risk and picking a play on a strictly turn-by-turn basis would probably get you close enough to human play to be considered Deep Blue in comparison to whatever it is Cyanide pooped out and packaged up for AI.
I have made games and I think you are overestimating the difficulty. The AI does not have to consider all of the possible moves or start with no assumptions. You can design a series of decision trees that grow from different starting conditions and then give different teams different choice weighting based on that.
Just because Cyanide can’t do state-based AI design doesn’t mean that it’s impossible.
You can also code a Chess AI based on assumptions and decision trees, but those AIs aren’t nearly as good as the brute force AIs that just consider every possibility (and that won’t work with Bloodbowl).
I’m trying to present an alternative to the very popular argument “Hey computers are smart, they can win chess, and learning to play Bloodbowl isn’t too hard, so if the Bloodbowl AI can’t match a mediocore Human player then the only explanation is lazy neglect on Cyanides part.” (Thats not directed at anyone in this thread, I think its the accepted opinion of most Bloodbowl threads on the internet).
Like I said, I’ll be happy to be proved wrong. In fact I’d love it if they made the Bloodbowl 2 AI easily moddable so that modders can prove me wrong and I can enjoy a single player campaign.
But I think the only explanation is lazy neglect on Cyanide’s part. The current BB AI is extremely simplistic and defaults to one of a very small number of plays. If it has the ball then it cages up even if there is no threat to the ball carrier. Also it tries to put receivers downfield - even for teams with a poor throwing game. If it doesn’t have the ball and the ball is free then it tries to pick it up even if it’s the last turn of the half and there’s no scoring opportunity. It routinely throws unnecessary -2d blocks and almost never uses rerolls.
Chess isn’t a good analogue for BB because every turn is a discrete moment. One player moves and there’s a definite ‘best move’ in response to that. In BB the AI needs to consider a chain of moves that may or may not be dependent on each other. However arriving at that chain isn’t terribly difficult. As I said, weight different plays differently per team and based on starting conditions. There would be quite a lot of those play-chains but it’s not something that is particularly difficult to achieve and would improve the AI to the level where the single player game was at least more interesting. I doubt that Cyanide will do that however as the main focus of BB from the boardgame through the various Cyanide versions has always been the competition against human opponents. Single player is an afterthought.
Possibly of interest to readers of this thread. Cyanide is doing another Bloodbowl Themed game called Blood Bowl: Star Coach. This one will be a Football-Manager style game.
I love Bloodbowl and all. But do we really need so many Bloodbowl games without a single Necromunda or Mordheim game? I guess Cyanide are just squeezing the IP they’ve already purchased. I’m still hopeful for a Necro or Mord game though. Games Workshop has been showing alot more interest in renting out their IPs lately, and hopefully the popularity of XCOM will tempt them to allow a direct conversion of these 2 excellent games.
The new developers of Space Hulk has said that they will unveil another license they got from Games Workshop soon, and that they will make a game about it. People seem to speculate that you may be in luck from that.
The blood bowl Star Coach sounds damn interesting!
Seriously, I wouldn’t be too worried about this with a good developer. With its multiplayer format, BB:SC could work well as a free game with Orcs and Humans available for free, and the rest of the teams available as IAP. It’s Cyanide, though, so natural scepticism to F2P x 2.
Stat boosting items. Star Players. Replacement Players. Entry fees for matches, even? All handled through the simple expedience of letting players buy in-game cash with real cash?
Those are things that spring to mind immediately one starts discussing F2P in a BB context. It would surprise me a lot if Cyanide haven’t thought the same things.
I’m played more than a fair share of Blood Bowl, but it’s just really hard to get that excited at this point by the prospect of anything that’s still in Cyanide’s hands. It’s like when Golden Corral had a big marketing push around how they now were serving steaks in their all-you-can-eat buffet. If you’ve ever eaten at a Golden Corral, it’s hard to get motivated by them announcing anything new on their menu.