The boards are going to be a LOT smaller than full sized boards. The classic board will probably only have like, 3 spaces on each edge between the corner spaces. I expect them to be tiles that are at most 6" squares.
Harder to prototype then. Though I just approximated 6āā with my hands and ya know what. On a secondās reflection, I think that works just fine. I guess itās back to plan A then, boring brown cardboard salvaged from amazon boxes that I drew on with sharpie and a ruler.
If the board are big, you wonāt be able to fit more than one on the table. Remember, a player can add a board to the game as an action. You could have ALL of them out in play almost immediately.
Also, the game would take forever to end and get really boring and repetitive if it was just one big board, or if a board took to long to finish.
That seems to be the case and maybe it isā¦ However, you donāt know until your first play test and second if the game is taking too long with full sized boards, just make the rules so they fill faster. Like for the classic board 1 move = 5 moves. The space issue canāt be ignored, even my godly sized dining room table cannot comofrtable accommodate 8 boards like 4 max.
Iāve long been working on a combination pen & paper role-playing game setting set in a distopian future where the players play characters that play a MMO. Iāve also been trying to write some semblance of a short story that establishes the kinds of characters and stories itās intended to play out. The fundimental idea is incredibly rich corporations, oligarchs, etc. gained control of essentially all capital and automation. The bread and circuses to keep some people āemployedā temporarily till the powers that be feel completely unthreatened by any sort of revolt is, for the player characters, playing these games for profit. The corporate AIs for the game actively police for cheats, exploits, tool-assisted play, and other AIs so that nobody has an āunfair advantageā. The players, however, are all people using some kind of advantage to climb the rungs of the games virtual wealth. If they get caught they might get a warning, a ban, or the company might send someone to kill the playerā¦ all depending. Advance your hacks far enough and youāll gain access to server resources and an entirely different āgameā where the relatively wealthy and powerful play at their own games manipulating and using those below for their entertainment. Similarly, break that world and gain access to resources in the real worldā¦ actual weapons, drones, subordinate AIs, maybe even WMDs.
Thematically I think of it a bit like Paranoia meets The Matrix.
Iāve been working on many game ideas but one in particular has always stuck out to me and Iāll be starting initial testing this week with some friends which is simultaneously exciting and terrifying. Unfortunately itās not the kind of game you can really do much with alone besides making sure it feels like a game so the first test will be pretty shaky.
1 Like
I recently played around with writing an AI for my Impulse variant. Itās not perfect, but it makes decisions I wouldnāt and adds an element of chance to the game to which I must respond. Iāve found it fairly helpful in solo playtesting.
If the AI wins at about the same rate a player would, it would imply that the game breaks down to a very simple heuristic tree and may well be fundamentally flawed.
That is interesting. Itās very hard to make an AI that is equal to humans. We have methods to make AIs that will inevitably lose to humans based on a set of heuristics plus some RNG. We also have methods to make godlike AI such as AlphaGo that humans canāt compete with.
I guess we can always use machine learning, but just donāt train it to win as much as possible. Instead, train it to get a score as close as possible to the score of its opponent. Sometimes it would get a little more, sometimes a little less. The worry is that over time it would be able to achieve an exact tie in almost every game. You would have to train it to consider that a score as close as possible to the opponentās score as a success, but a tie would be a failure.
My specific interest is in a practical use case.
If an AI that is of the level of complexity a human could follow with training alone (or a limited-scale reference of if-then statements) can be as good as unaided and free-playing humans: the game is boring.
Another way to put it:
If you can fit a set of heuristics for an āAIā player in the reasonably-sized rulebook of your game on a single page, and that AI performs as well as reasonably skilled humans, your game is inherently boring to any reasonably intelligent human after a very short number of plays.
Practical ML applications are almost always a search problem. Itās easy to limit a search: restrict time, restrict memory, restrict beam size, etc.
Yeah, I follow this. Thatās another reason to have one, actually; I can test the game for that major flaw without boring anyone else.
How can you tell if there is a flaw in the game or a flaw in the AI?
Ah, the Blizzard problem. So many metrics, so much data, so much polish that they trust their numbers even up until theyāre betrayed by them.
Iām looking for major flaws. The AI performs legal moves with a very basic strategy, so itād be hard to make it break.
Whether or not itās playing optimally is a different story, and at least in my case, was beyond my ability to write. At that point, Iād be writing out my own decision tree, and then itās preferable for me to get human players so Iām not playing against my own brain.
I really just donāt see the purpose of building a barebones AI that any competent player should crush other than to figure out if the game is basically random or not. You might not be able to play a completely real game against yourself but you should be able to tell if there is some reasonable play and counterplay in the game. Once you know itās functional and has at least some play itās time to get a competitive game to the table with playtesters.
AlphaGo already does try to get a score as close to a tie as possible. You only need to get a score of +.05 in Go to win so thatās what it goes for. If you push it harder it just pushes back until itās got .05 more than you and then it pulls back.
You can really see this in a bunch of the games it played with itās Master variant, thereās a moment where Master starts to play some pretty slack moves, itās at that moment that the AI calculated that itās won no matter what and it just starts to solidify itās territory to prevent shenanigans.
Itās a good teaching/learning tool to introduce people to the game. More board games should provide āAIā like that.
Itās a handy way to handle situations where a player leaves the game. āIf someone quits, their foo does bar every round.ā
It gives a sense of the depth of the heuristic tree (based on how many rules you need to write for the AI to be competent).
It gives an opponent of constant skill against which to test strategies in a controlled fashion.
2 Likes
I should try writing an actual competent AI for this Impulse thing, but stupid Chudyk games are fiddly with like 17,000 options. Itās not that the decision tree is hard, itās that itās got lots of words in it.
Goddammit Carl, why are you this way?
So in poking around at re-redesigning Impulse to be something I actually want, I ran into a neat little Amazon shop:
These guys sell a variety of blank board game components - cards, tiles, discs, boards, etc - for reasonable prices.
The thing that caught my eye are these dry-erasable poker cards:
Theyāre compatible with wet-erase markers, so you can get some and make a deck that you can shuffle without erasing.
The only downside so far is that they can wet erase with the moisture from your fingers. It takes some friction and sliding to do it, so you can still hold the cards in your hand and look through them so long as youāre gentle. Takes a bit of getting used to, but itās workable. Still, blank components for not much money!
3 Likes