Computer play - request for comments
complete
A
Alfie Kirkpatrick [Developer]
I have added a first version of computer/bot play at beta.backgammonhub.com. I'd like to get some feedback before developing this further.
To start a bot match, click the "Play a Bot" button at the top of the page.
At the moment all matches are 1 point, using the GNUBG preset skill level "casual player", but I've done most of the work to support longer matches and higher skill level including 3 ply.
There is some more housekeeping I need to do, for example, filtering bot matches in your profile, and I don't think bot matches should be included in your moving average PR presented to other players. I'm not sure whether blunders from bot matches should be included in your blunder database.
I'm a bit concerned that when bot play is available, players may be less motivated to challenge other (real) players. Would love to get people's thoughts on this.
I'm planning to make stronger opponents "members only" to encourage subscription and offset the resource costs of running the bot.
Comments very welcome!
B
Backgammon Hub Admin
complete
A
Alfie Kirkpatrick [Developer]
This is live 🎉
D
Denis-George Constantin (playhunter)
I too think that when bot play will be available, people will play more against bots rather than other humans (but I think you will need to actually test it for 3-4 months to be sure whether this is true or not).
Keep separate stats (win rate and PR) for play against bot and play against other humans. As for blunders against bots, I think you should bundle them into the same personal blunder database along with blunders against people.
Allow non paying members to feel what a max power bot is all about ! So maybe limit their bot play accessibility to only one single match per day (or 3 matches per week) any length. But I am not sure why would you allow weaker bots ?
A
Alfie Kirkpatrick [Developer]
Denis-George Constantin (playhunter) thanks for feedback.
For win rate and PR, am not going to record at all for now. Blunders will be included with the ones from human play as you suggest.
I like your idea of limiting bot play by number of matches rather than skill level, but it's quite a bit more work to implement. My theory (yet to be tested!) is that players will not like to play many games against a clearly inferior bot, but that serious players may like to subscribe and play a stronger opponent.
Alongside this I will probably be implementing Alex Z's suggest of board setup / position analysis using the strongest GNUBG settings, possibly even rollouts.
It's important not to impact human play too much, so will definitely be keeping an eye on the stats after launch!