ChessBase 10 trees - part 1

Basically, an opening tree or book is a batch of games merged together so that the user can view statistics for each move in the tree. Here's an example, White's opening moves from the 2008 version of the ChessBase Opening Encyclopedia:

What's happened here is that all of the games from the Opening Encyclopedia have been merged together into a single tree so that we can get an overview of many moves at once. All of the White opening moves from the database are displayed here (in the leftmost column, obviously). There are several columns in the tree display, each containing a different type of information:

• N -- the number of games in which a particular move was played;
• % -- the "success rate" of the move, a measure of how well it did based on the results of the games in the database;
• Av. -- the average rating of the players who played a particular move;
• Perf. -- the performance rating of the players who played a particular move, a function of the player's rating and the game's final result.

Note that the top of the tree displays overall totals for each of the four columns (although, for some reason, the first digit "2" is omitted from the two ratings columns in the above illustration). In the picture, we can see that there are more than three million games in the database. As we move down the column, we see totals for each individual move; for example, the move 1.e4 was played in more than one and a half million games in the Opening Encyclopedia.

The percentages indicate the "success" of a move. White wins are scored as "100.0" (so if every appearance of a move in a database game results in a White win, you'll see "100.0" in the percentage column). Black wins are scored as "0.0", while draws are scored as "50.0". Sometimes incomplete games or analysis lines ending in evaluation symbols will appear in a database; these are also assigned scores which are valued between 50.0 and 100.0 for evaluations favoring White or between 0.0 and 50.0 for evaluations favoring Black.

Let's see this in action -- we'll click on 1.e4 to make that move and see Black's replies:

And we see all of Black's replies from the database. Note that in Black's case low numbers in the percentage column reflect moves which scored well (since Black wins are scored as zeroes). In this illustration the moves are sorted according to the number of games in which they appeared. But what if we'd like to sort the moves by some other criterion?

For example, we might wish to see the moves sorted according to percentage instead of frequency. All we need to do is single-click directly on the "percent" sign (%) at the top of the tree display. We'll then see the move display change to sort the moves by their "success rate":

Since we're looking at Black moves, low numbers are preferred. So we see that in a strict statistical sense the move 1...Na6 is Black's "best" reply. But we also notice that the percentage is displayed in half-tone ("greyed out"). This means that a move didn't appear in enough games to provide a reliable statistical result; i.e. such results should be taken with a generous dollop of salt. As far as "real" moves go (those which were played often enough to generate some meaningful numbers) we see that the Sicilian Defense (1...c5) has done the best for Black.

Let's single-click on 1...c5 and see White's candidate replies:

We now see all of White's replies to 1...c5 in games from the Opening Encyclopedia, sorted by game results. We can immediately blow off 2.g4 (played in a whopping two games). The move 2.Ne2 is interesting (and, if I recall correctly, somewhat in vogue briefly in the early 1990's as an "opponent baffler") but it usually transposes right back into the familiar 2.Nf3 lines. Note that since we're looking at candidate moves for White here, the "high numbers" once again are favored.

This single-clicking directly on a column header to re-sort the moves will work for any of the four columns at the top of the tree display.

What we've just reviewed is the "traditional" functioning of the tree feature in ChessBase, which dates back to the old DOS days when CBTree was a separate program. Over the years, however, I've talked to users who find the "always from White's perspective" percentages to be confusing. That's where CB10 comes in, and we'll talk about that next week. Until then, have fun!

You can e-mail me with your comments on ChessBase Workshop. No tech support questions, please.

© 2008, Steven A. Lopez. All rights reserved.