07 November 2017

Reconsidering an Online Chess Database

Decisions, decisions. A few years back, faced with a technical problem, I scrambled to find a new online chess database and documented the effort in a series of posts on this blog (February/March 2014):-

In that 'Choosing' post, I described the way I use an online database to keep up with opening theory. Basically, I start by looking at what the top players have been playing in a particular variation during the last couple of years, then broaden the parameters as the game goes deeper. I discovered that none of the databases could support this methodology and I decided to continue using the same toolset as before, accepting an increased security risk.

The database I've been using, Chesslab, was conceived around the year 2000 and is based on Java. This was adequate at the time, but the times have changed and the technology is no longer robust enough to protect against malicious activity, which is nowadays the primary disadvantage of doing anything online. Earlier this year, Firefox withdrew support for Java, effectively disabling Chesslab. Explorer still supports Java, but throws up so many warnings that I have to doubt my judgement in continuing to use it. On top of this, the downloaded PGN files are not usable 'as is' and have to be converted into a text format that can be loaded into PGN readers.

These days my main interest in online correspondence chess is chess960, where online databases have no value in the opening. A year or so before those posts on online chess databases, in Cup Play (May 2013), I wrote about continuing to play the traditional start position ('RNBQKBNR'). A new tournament will start in a week. If I play, I'll need a database to guide me through the opening phase, but which database?

Looking again at the work I did in 2014, there was one service that was still under development, Chess-db.com. I decided to take another look at it. Indeed, its 'Opening Explorer' is now an improvement over what I found in 2014, and in the direction of what I need. It has a filter on rating, but no filter on when a game was played. Is the glass half-empty or half-full?

Unfortunately, the filter on rating is not up-to-scratch. It restricts results based on the rating of the player(s), but unnecessarily adds a +/- 100-point band around the search. If I'm looking for the current practice of 2700+ players, I don't want 2600 players included in the search, but I do want 2800 players. In addition, the results look dubious. Here is a side-by-side comparison of the results for the traditional start position filtered on a 2700 rating and on 2800.

The results are identical, indicating that something is amiss. Having said that, the ability to go deeper into any variation and see what top players have tried is an advantage. Is this good enough for a tournament section of six games? The event will be the quarterfinal stage of a multi-stage tournament. Because I always prefer to finish what I've started, I'm reluctant to abandon the tournament now.

In the 'Resurrecting' post mentioned above, I wrote, 'As for giving up traditional chess and switching to chess960, I am certain that day will come. It just won't be today.' Perhaps that day is now. I have a week to decide.

No comments: