Cross words over Sudoku, well then use poker?

So why this cryptic title? Well, if, like me, you have picked up a newspaper one evening over recent weeks and flicked through past the natural disasters, credit crunch and housing crises to try to find some lighter material, you may have ended up on the “puzzles” page. The typical page will have a selection of brain teasers including the traditional crossword and also the more recent Sudoku puzzles. What struck me - as I was trying to decide between which of these two to start, get half way through, get stuck and then switch to the other – was how these two were an analogue for the changing face of financial trading. The crossword is more old school with the need for a good grounding in the vocabulary of the market, being able to spot relationships, language skills, deduction and being able to decode cryptic clues. Many of these skills are suited to solving by people. Sudoku, the new kid on the block, is far more about numbers, algorithms, calculations, processing power – more suited in many ways to being solved by computer.

Also Sudoku has a major difference to modern markets in that it is a game of perfect information i.e. the rules are clear and there is no information required from outside the puzzle to solve it. Those trying to understand why the price of a barrel of oil has shot beyond $135 know that it does not just rely simply on the fundamentals of supply and demand – there is also psychology, fear, greed, bluff and double bluff. These are all things that people know about, but with which computers have struggled in the past. However, there has been the news recently that computer programmes took on some strong poker players. In the first “man versus machine” poker championship in Vancouver, the world's best poker-playing programme Polaris took on human poker champions Ali Eslami and Phil Laak and, after some interesting competition, the computer won. So what is the big deal about poker you may ask? Well poker, given its mix of incomplete information (you don’t see all the cards) and the softer aspects of the game, aggressive strategies and at the same time the need to use bluff and to hide your intentions – it is not called a poker face for nothing – has been viewed as a model for some financial markets. So the ability of a computer programme to handle these aspects in a world of incomplete information is a major step forward for the application of computing to real life environments going beyond pure economic and rational analysis.

It is surely a matter of time – if they are not there already – that similar techniques are put to use in the more secretive world of investment banks algorithmic trading engines.

NB A version of this blog appeared in my monthly Financial Sector bulletin. To receive a copy each month send an email containing the word SUBSCRIBE in the title to contact at mpi-europe.com