Turn-based games, the database, and the AI
Posted:
Nov 24, 2016, 2:10 PM
I recently implemented a new feature, allowing you to do database lookups from any position in any game in your game history... at the click of a button. On mobile devices, I went a step further, in addition to querying the database there's now also an option to ask what the (Mark Mammel's) AI would do.
The only thing separating this feature from ongoing games and historical games is a single line of code determining if the game is over or not. I am not keen on providing this feature to ongoing games, but I find myself wondering "what if".
Would this make pente.org a less popular place to play? Is it a nifty feature that would give pente.org an edge? How do you feel about playing someone who can consult the database and AI at the push of a button in their game? ... ?
Re: Turn-based games, the database, and the AI
Posted:
Nov 25, 2016, 3:31 PM
I don't think it would make pente.org any more popular, probably less Nifty feature, but for rated games I would NOT want to play against that opponent. Otherwise I'd just play the higher level AI.
Re: Turn-based games, the database, and the AI
Posted:
Nov 28, 2016, 2:55 PM
You trust correctly
I'm leaving it out, unless a player wants to pay a few hundred USD / year to use such a feature
Newbies get their butts kicked hard enough already, and such a feature would only help the subscribers as it is only available to them.
I was, and still am, wondering how I would react if my opponents had such a feature available. I honestly don't know if I'd be turned off to play or welcome the challenge to improve myself.
Re: Turn-based games, the database, and the AI
Posted:
Mar 2, 2018, 4:49 PM
I get the reasoning behind not having this feature (ability to go directly from an in progress TB game to the DB) for Pente - much shorter length of game, much more populated DB and a much more biased game than Go. Perhaps to a lesser degree it also makes sense for games like gomoku, keryo and C6 (fewer games in the DB, so it is less helpful for finding already played games). However, at least two of the reasons above also make it more likely for the database to be useful for games like gomoku and Pente than for Go. I'd be surprised, unless or until we get Go games imported into the DB and/or an influx of players who have studied the game if one could frequently find positions in the DB beyond 10 stones, much less for games further along, which is where I'd find it helpful for Go. The thing which would be helpful for me and other players who use the database as I do currently is for examining the consequences of a move by playing a line out from both sides of the board to see where it might end up. It's not uncommon for analysts to extrapolate 10 or moves ahead in Pente or Gomoku and I'd do the same in some Go positions, given an easy way to do it, but it's too much of a pain right now to enter 70 stones into the DB in order to see how my next move might play out. As things stand now, not having the ability to use the database to explore the consequences of a move in TB play probably favors those without pets and small children (for setting up an actual physical board) or those who have invested money in a Go program (for storing their ongoing game positions).
Retired from TB Pente, but still playing live games & exploring variants like D, poof and boat
