As is implied from our header, a significant part of our time has been working
with the card game Spades. This has been marked by some impressive progress
but the goalposts did not stay put and, even though the new work was clearly
stronger than our original program, it still demanded much more work to get
to a product that both played well and was also perceived to play well. This
was built around MCTS (Monte Carlo Tree Search), which also revolutionised
Computer Go a few years ago, but had the same issues. MCTS is undoubtedly
the most important idea in gaming AI for a long time, but getting it to deliver
what you want for a gaming opponent is non-trivial. Early MCTS-based Go easily
beat other methods but played with a style that invited derision. That barrier
has since been passed and now MCTS-based Go programs dominate commercial Go.
For our purposes the goalposts were higher still. We want MCTS that not only delivers for Spades but will also deliver for all other games as well, but using the same generic method. That is a tall order but it seems that we may have cracked that and created something that plays plausibly well with a generic framework.
In among our work with the York group (who authored the article on Spades here and will be publishing at the AIIDE 2013 conference in Boston), we also contributed to the Royal Academy of Engineering Summer Soirée -Engineering: Design for Living at York University.
The team left above are Professor Peter Cowling, Edward Powley, Daniel Whitehouse and Jeff Rollason, who were all involved in the Spades development. On the right, Peter is explaining the research into MCTS to HRH Princess Anne, while our Spades app is shown displayed bottom right. This event again offered good opportunities to brainstorm our Spades AI.
In review, AI Factory is still looking very strong in the Android market, even though that market is now very competitive. Our apps generally float near the top of the charts for each game type in most categories. Our highly successful Chess Free is ranked #25 in Brain and Puzzle Free, 142 places ahead of our nearest rival. As more investment is made in Android we cannot expect to monopolise indefinitely, but we are looking to take our highly successful franchise elsewhere as well. Already we are collaborating with Blit for Windows 8. Our chess is now on Windows 8 as Chess Pro.
Our most recent app, Solitaire Free, adds this Klondike game to our portfolio and is doing well. We expect to add further Solitaire titles, but first we have other card games in mind and should have a new major release scheduled for Q4.
This newsletter is "late" compared to our previous one, which was "very late". In consequence the number of events between this and the last is less than would be delivered by the customary 6 months we intended. By next time though we expect to announce some significant new partnerships that will spread our franchise across more platforms.