In the past, our quest was aided by magazines and periodicals such as Chess Informant that brought us the latest news on our favourite openings. You all know how it works: You spot a crushing novelty played by the world champion on move 20. You rush to try it in your next game only to find your opponent deviates on move 5, and you are mated on move 18.

Today, technology has dramatically changed our access to information. It is easy to keep up with the latest developments on the internet on such sites as Mark Crowther's excellent The Week in Chess. We can couple this information with a strong chess program, and we have a good foundation for serious opening research.

This brings us to the new product from the makers of Rebel, namely Rebel EOC. There are three components to this product, namely

(a) Encyclopaedia of Chess (EOC),

(b) Million Base, and

(c) Rebel 10 Light.

All three components are compatible with each other to form a stand-alone product. In addition EOC and Million Base can be used with Rebel 10. I will discuss each of them in reverse order.

Rebel 10 Light

Rebel 10 Light is a version of Rebel 10 with some features deleted and a reduced strength. However, it is still a perfectly respectable chess playing program with many features, and it's probably strong enough for most players. The database search features are the same as in Rebel 10, making it a good tool to use with Million Base. More details on Rebel 10 can be found on the Rebel web site, which details the program's many features.

Million Base

As the name implies, Million Base is a database of just over one million games, which appears to be current to the end of 1998. About 64% of the games are from 1990 onwards, showing the explosion of chess knowledge in recent times. The database comes in both the Rebel format and the old (cbf) Chessbase format. It can be converted to other formats if required using the tools at the Pitt archive. Three issues are important for a large database, namely duplication, completeness and quality. However, it is difficult to verify these attributes given the number of games involved. Out of interest, I did a few random checks as follows.

Duplication

I used Rob Weir's Cbdedupe utility on the cbf version of the database, and it reported 163 double games based on its "conservative" setting and 307 double games based on its "radical" setting, which is a remarkably low number in both cases. A spot check of some of these doubles showed some zero-move draws and default wins. These were picked up as doubles since Cbdedupe only considers moves rather than players.

Completeness

Million Base has 1004 games by Capablanca. Golombek's book of his games reports that Capablanca played 582 serious games, suggesting that there are also plenty of non-tournament games by Capablanca included. Similarly, there are 906 games by Fischer. Wade and O'Connell give 789 games up to 1972 in their book. Adding the 30 games from the second Fischer-Spassky match gives a total of 819 games. So again, the coverage is good. Curiously, the database also contains the game Karpov-Fischer, Manila world championship match, 1975, won by Karpov with no moves! It's interesting to speculate what Fischer would say about this "game".

Quality

Million Base includes various non-master games, such as around 9000 games from the internet email correspondence clubs. There are different views on whether this is desirable or not. One proponent of non-master games is US correspondence champion Jon Edwards, who argues that any game has some educational value. My own view is that you can become overly concerned about such matters. If these games are not of interest, it is easy to ignore or delete them. You will still be left with more games than you could possibly want for study.

In summary, Million Base is an excellent source of games that should satisfy most players.

EOC

The EOC concept was introduced in Rebel 10, and it is basically a chess tree of opening positions. EOC has 50 million positions and was compiled from the games in Million Base. In comparison, the version that comes with Rebel 10 has 16 million positions based on 300,000 games. Being position-based, the tree copes with transpositions, so that the (absurd) moves 1. e4 a6 2. Bb5 Nc6 3. Nf3 e5 still lead to information on the Ruy Lopez.

EOC can be used as a straight database of opening moves. For any position, EOC provides statistics on the number of games played from the position, ELO ratings, and the results of the games. It also lists the number of times various top players have reached the position. An example of the output produced is shown on the EOC page. This information is invaluable for opening research, and can be used to spot weaknesses in your openings and for learning new openings.

It is interesting to compare the data from the two versions of EOC. The statistics for various opening moves from the two versions are shown in the table below.

Move	Winning percentage (16 million positions)	Winning percentage (50 million positions)
1. e4	54.0	53.8
1. d4	55.7	55.4
1. c4	55.8	55.1
1. Nf3	55.3	55.4

The table shows that there is not much difference in the statistics, and suggests that the data is fairly reliable. Of course, it should be noted that games aren't won solely based on the opening. Moreover, the reliability of the data depends on the number of games. It turns out the 1. Na3 has the best winning percentage (61.5%)! However, this is based on only 13 games, so try it at your own risk.

EOC can also be used in both Rebel 10 and Rebel 10 Light to assist them in the opening. In this way, Rebel's choice of openings is influenced by the results of the games from a given position. This clearly makes Rebel formidable in the opening, and the prospect may be just a little too depressing for most players.

In summary, EOC is an extremely useful tool for opening research, and for strengthening the opening play of Rebel.

The Ultimate Opening

Finally, just what is that ultimate opening we've all been searching for? Ignoring the spurious results based on too few games, EOC suggests that the best opening move is either 1. d4 or 1. Nf3. After this, you just follow the best moves on the tree depending on black's reply, and that's it. Maybe Portisch was right after all - it is easy! Just don't blame EOC if you get mated on move 18 because you had no idea what to do when you opponent deviated on move 5.