03 September 2009

Itatsi: Part 2

My communication with the author of the game Itatsi.
Itatsi word game
Jim writes:
Thanks for the support. Itatsi is a work in progress. I'm still trying to figure out how to make it into an attractive game. So far only my duty bound family has said it is playable and interesting.

I got the idea while playing with Elsberry's Weasel page. I tried using words as the target instead of sentences. I discovered that words could be made in ten or twenty generations with a small population. From there it was just a matter of learning JavaScript and figuring out how to do the dynamic HTML. I've dabbled in these before, but never to this extent.

I had visions that Itatsi had commercial potential, so I hid the mutation engine and fitness scoring engine on the server side. ( I also needed a database to handle the fitness table.

The fitness table went through several generations. I'm told I'm using a hidden Markov model. I didn't know what that meant until I looked it up. The last math course I took was intro to statistics in 1970. I developed a frequency table for letter pairs and triplets. The Weasel children are scored for fitness base on a database lookup.

The fitness scoring was an afterthought. I originally intended to make a very difficult game in which the player would be completely on his own. This is still and option, but I doubt in many humans could have the skill and patience to go beyond five letters. So I decided to make a hinting engine.

[and my response]

Dan: Thanks for the reply. I did a bit of research (ie: playing) so I could understand it better. I discovered that I would often exhaust the un-scored words on for the current stem, and I by intentionally choosing a parent with a lower fitness I could quickly move to a parent that would allow me to score well again. In other words, I had to go down in fitness before I could move up; an interesting evolutionary strategy!

This certainly qualifies as "a series of interesting decisions", which is Sid Meier's definition of a game, and as such it is playable as-is. My wife played a lot of "Scrabbulous" before Hasbro sued to shut it down, and Itatsi has some similarities to that, though a very different game engine (I might be mistaking Scrabbulous for another game). The scoring system allows for a competitive goal, which is always good in a game. It would be nice if you could compare score with other players too.

It occurs to me that if different players could all start with the same random seed number, or if a single player could replay a game starting with the same seed, this would allow a common starting point for competition. Letting two players play with a common list of score words (one player can't score a word if the other player already got it) would add a different sort of competition.

I congratulate you on a clever idea, and I hope you might have some success in taking it commercial.

Jim Responds:
If you check the demo mode an set a game length of 100 or 500 and a long word length you may notice that the demo engine automatically chooses the second most fit child every fourth generation. You can see this by watching the fitness scores. I call this my extinction event. I find it gratifying that you found that strategy independently.

The yellow (or gold) highlighted child has the highest fitness score; the silver child the second highest. These hints can be turned on or off. As currently written, a player accumulates no score while hints are on. I have been thinking of having the program swap turns with the human player and compete for score. The program has the advantage in early rounds, but the program has the disadvantage of not knowing about scoreable words, and not knowing to avoid words that have already been scored. It does remarkably well anyway in demo mode -- something that I find relevant to the issue of irreducible complexity.

You might also find the variable word length demo interesting, if given 500 generations. The French language appears to have less distance between words, and 10 letter words appear quite often in demo mode. You can switch language in mid-game.

Dan again: And I did. I played the variable word length, and I found it quickly took me into longer strings than I could easily manage. I therefore tried to keep the length from growing too much and so I could consistently get new words, but this may have been poor strategy; Longer words score many more points, and I can afford not to score any words for several generations if there is a big payoff down the road. Darn, Now I want to go try this out - a hallmark of a fun game. :-)
GBR Giant Battling Robots Favicon
Post a Comment