Article Listings

Current articles

Adding New Muscle to Our AI Testing
We have beefed up our test framework to accelerate our development work.

Unwelcome Testing Realities
There are few free lunches in testing and here we expose some unwelcome truths.

Previous articles

Designing apps that drive user engagement and retention
Andreas Vourkos takes us into the issues involved in user retention.

Human Randomness
This takes the input of a user trial to look at how humans perceive randomness, and its impact on games.

The Rating Sidekick
This little program has proved central to making our Spades program strong.

Mixing Development Practices
It is easy to get sucked into just 2 main development procedures.

Mixing the Immiscible - MCTS and evaluation
Not all tech is exactly plug and play.

Under-sung heroes - Gothic 3
A might-have-been project for us. We take another look.

Keeping on the Tightrope
Finding more ways to avoid calamities during development.

Developing in Fog
Debugging imperfect information games adds a new level of opaqueness.

When Completely Random is Just not Random Enough
Being right is not good enough.

Bidding Systems with Euchre
Harder than expected! We created 4 different systems.

Interest Search Another way to do Minimax
This method is key to our minimax AI.

A Rating System for Developers:
This has been an essential cog in our Shogi development.

A Folding Revolution
All the AI development within AI Factory is built around folding editors.

Searching the Unknown with MCTS
We step back and take a broader look at this technology.

Reducing the burden of knowledge: Simulation-based methods in imperfect information games
The York group digs deep.

Lessons from the Road Championship level AI development
The long road to a winning formula.

Pushing compression limits using Plausibility analysis
Our work on Solitaire threw up some efficient ideas.

Life in the Fast Lane
Hang on tight as Cameron and Stephen take you into the depths of bit-optimisation!

Tuning Spades
Our work on Spades has needed experimentation with new ideas for us. We report here.

Evaluation options - Overview of methods
An area over-burdened with bear traps. We point out some avoidance tips.

Painting by Numbers
Move it! and Sticky Blocks use a dynamic algorithmic system for assigning puzzle colours.

Mixing MCTS with conventional static evaluation
Experiments and shortcuts en-route to full UCT.

Creating Tools for Creating Puzzles
This addresses what tools we used to create puzzles for Move it!.

Digging further into the Statistical Abyss
Another look at the pitfalls involved in assessing test results.

Backgammon Programs Cheat: Urban Myth??
Addressing an unfortunate disparity between perception and reality.

Modular Graphics for Move it!™
This is not AI, but the issues involved in representing endless shapes on limited format devices.

Solving Deep Puzzles
This is our newest technology and is being widely applied. Here we discuss Move-it™.

The Computer Plays Pool
Creating a Pool/Snooker program throws up a few surprises and offers its own special challenges.

The Computer Learns to Play Poker
Our development of a Texas Hold'em poker engine is build around automated learning.

19th CSA World Computer Shogi Championship 2009
Reijer's report on this main event in the Shogi calendar, including Spear and Shotest progress.

TEACH: A Generic Script-based Teaching Language
This flexible script language turns a product into a teaching system.

Extending the Generic API for Render Control
This allows the generic architecture in use to encompass full GUI control.

A Tutor System for Pool
This introduces our new dynamic teaching system for Pool and Snooker. The AI tutor personalises teaching of this game.

The 18th CSA World Computer Shogi Championships
Reijer's report on this main event in the Shogi calendar, including Spear and Shotest progress.

The Puzzle Workbench
This talks about our generic framework for creating puzzle games and plug-in components for use in 3rd party games.

Statistical Minefields with Version Testing
Returning to issues that need to be addressed when developing competition-level AI.

Emulating Biological Systems – 2 of 2
This is the final section in this series, looking at how we approached the creation of Aquatic AI.

Creating Book Knowledge
Opening knowledge is essential for many board games. This looks at the basic needs for constructing such a system.

Developing Biological Emulation using a Generic Turn-based Testbed
This takes the AIF generic testbed architecture into the difficult domain of bio-emulation.

Predicting Game States in Imperfect Information Games
Looking at the simple game of Gin Rummy and effective heuristics to play it.

Negative Plausibility
Taking another different look at this core technique used to re-order moves in search trees.

The 17th World Computer Shogi Championship 2007
Reijer's report on this main event in the Shogi calendar, including progress of Spear and Shotest.

Emulating Biological Systems – 1 of 2
This article looks at the problems of modelling an aquarium, requiring emulation of a real biological system.

Advanced Diagnostic Tracing
This is a follow-on to a previous article on our in-house development, this time looking at the testbed console.

Looking for Alternatives to Quiescence Search
This provides another look into our flagship product, looking at the way it deals with tactical exchanges.

Intelligent Diagnostic Tracing
There are many way of putting code together. This looks at one key aspect of our AI engine development.

Interpolating 3D for Faster Framerates
The demands of a complex animated world can impose a heavy processor cost.

Optimised Navigation of Complex 3D Worlds
The AI needed to support 3D can be expensive. This looks at ways of optimising the maths required.

Fuzzy books: Approximate opening knowledge
Classically, game playing programs use precise openings. Shotest takes another approach.

The World Computer Shogi Championship 2006
Reijer covers this major event in the Shogi calendar, including progress of SPEAR and Shotest.

Plausibility analysis
This is the essential part of minimax search programs faced with vast numbers of moves.

Playing stronger by learning
This looks at simple ways to improve play strength by using extended auto-play learning.

Evaluation by Hill-climbing: Getting the right move by solving micro-problems
Progressing to an AI win by small steps.

Off-the-shelf AI : Plug-in Minimax
A generic resource to create quicker, easier and better projects.

Treebeard - A new way to do chess
Computer Chess has a long history. Treebeard experiments with new ways of solving this classic game.

The World Computer Shogi Championship 2005
Reijer Grimbergen plots the progress of this annual event and AI Factory's program Shotest.

Developing competition level games intelligence
AI Factory game engines have been entering tournaments since 1992. The preparation is its own science.

Writing cpu intensive AI without multi-threading
Handling processor sharing between UI and engine is a big issue, which AI Factory has addressed.

Future articles

Mixing inference and MCTS
Flat Monte Carlo and Gin Rummy - Another twist
Solving Hearts: Analysis of this imperfect information game