Learning from the Machine – Part 2

Introduction

This is the second article where we are going to look at how you can use modern technology to improve your game and also get more enjoyment from it on a day-to-day basis. In the first part we covered some of the history of the game. In this article we will complete our review of backgammon software and start to look at actually playing the game. 

Computers (Continued)

Not long after TD-Gammon came the first commercial neural net backgammon program, Jellyfish, created by Frederik Dahl. Jellyfish was so named because in comparing its brainpower to that of a man it was roughly equal to a jellyfish. That is not to denigrate Jellyfish - the program was a huge step forward for backgammon and it had a good user interface which made it easy for thousands of players to use.

Suddenly the overall strength of backgammon players took a big leap forward as more and more began to use and learn from Jellyfish. Now, at last, rollouts could be relied upon for many types of position and humans were quick to learn from their silicon friend. Jellyfish was by no means perfect and its handling of back games in particular was poor but as Neil Armstrong would say “it was a giant leap for mankind”.

Next out of the blocks was Snowie, the brainchild of Olivier Egger in Switzerland. It started its life as “Snow White” but that would have caused problems with Disney and hence the change of name. Why Snow White or Snowie? Unlike Jellyfish which has a single neural net, Snowie has seven (the seven dwarves!) and is consequentially more powerful. Snowie is now at Release 4 and it has improved with each release both from a playing strength and user interface point of view. Snowie 5 is already under development but is probably eighteen months away from release. The one drawback to Snowie, and it is a big one for many players is the price. At $380 for the Professional Version (the one that enables you to do rollouts) it is out of the reach of many of the very people who would love to have a copy.
 

Luckily, shortly after Snowie appeared, the GNU Project, which is supported by the Free Software Foundation, decided to produce a neural net program for backgammon. This, not surprisingly, is called GNU Backgammon or GNUBG for short. It can be found at www.gnubg.org. (For those who wish to know the abbreviation GNU is a recursive acronym standing for “GNU’s not Unix”)

Like Snowie it has gone through many iterations and the current release is a very powerful beast indeed. Both Snowie and GNUBG can hold their own with the very best players in the world. There are still technical areas of the game where they need to improve but they never suffer from the distraction of emotional influence which gives them a huge edge over the vast majority of human beings.

Which is the stronger program? The jury is out but the two are very, very close in playing strength and I wouldn’t like to say one is better than the other. Because GNUBG is free and therefore available to everyone we will use it in our examples of how to use technology to improve our game.
 

Playing Backgammon

The way most people improve at backgammon is by playing. Typically, you will be taught by a family member or friend, and your early games are with that other person. If you quickly become competent you will seek to improve by playing against other people and you might join a local club or even enter a small tournament.

At some point you will realise that the chouette, the multi-player version of the game, will give you your best chance of improving quickly. Playing in a chouette with stronger players can be expensive but it undoubtedly accelerates the learning experience.

In parallel with playing you should be reading backgammon books as you can’t hope to really understand the game without studying what the best players have said about the game over the years. Reading and studying also significantly accelerates the learning curve. The bibliography of the game is outside the scope of this set of articles but if you are serious about the game you must study.

But back to playing: you will quickly learn that playing with the same small group of people all the time doesn’t improve your game that much. You will reinforce your strengths but also your weaknesses. To improve you must seek out and play with as many different people as possible in order to hone your skills.

That brings us to a major problem. At the height of the backgammon boom of the 1970’s finding players was not a problem – you could find them at the drop of a hat in any bar or pub, backgammon clubs were plentiful. Finding a pigeon to fleece for a couple of hundred pounds was easy!

Sadly that is no longer the case. Although the game is once again on the rise it has yet to reach the heights of the 1920’s or 1970’s. We also live in a much more time-conscious society than we did thirty years ago. It seems that every second is precious and we are always in a hurry. We also live in an instant gratification society, when we want something we want it now.

How then to satisfy the modern backgammon player’s demand for a game at any time of day or night against other players with a wide cross-section of skills? The answer is, of course, on-line play and it is to that area of the game that we will turn our attention in Part 3 and look at sites like Gammon Empire (www.gammonempire.com)

Summary 

We have completed our analysis of the computer world and selected our engine of choice, GNUBG. In the next article we will do the same for on-line play sites and then look at how we can use the two in harness to develop our skills.

Do you want to use what you've learned? Try our recommended Backgammon Server of this month – Play65. Join 24/7 backgammon games and tournaments with 12,000+ players from all around the world. Click here to play.