ELECTRONIC T-NOTES


CHESSBASE USA'S WEEKLY ON-LINE NEWSLETTER


FOR THE WEEK OF SEPTEMBER 6, 1998


OPENING LINES

by Steve Lopez

I'd like to welcome Electronic T-Notes' new international audience. Though the Internet knows no national boundaries and ETN has always been accessible to anyone with an Internet connection, many chessplayers outside the U.S. were unaware of its existence as it appeared only on the ChessBase USA homepage. A few weeks ago, ChessBase International began carrying ETN on its web site, making it accessible to a wider audience.

For those readers interested in catching up on back issues, there is over a year's worth of ETN archived in .ZIP format at ChessBase USA's web site. There is a link to the archives at the end of ETN at ChessBase International's site.

Some caveats for our new readers. First of all, I write these columns with the idea that we're all chessplayers who are trying to improve. I freely admit that I am no expert on chess. I am, however, fairly knowledgeable about chess databases/playing programs and the methods by which players can use them to improve. I was lucky enough to have spent three years as a full-time ChessBase USA employee, sneaky enough to ask chess questions of the masters and grandmasters who were calling me up with their own questions about chess software, and smart enough to shut up and listen when they were telling me something. As a result, I developed a unique perspective on the integration of chess software, books, and good old-fashioned play to help improve one's game. I consider it a great priviledge to have been a friend to many strong players, despite my own abysmal abilities as a player, and I'm grateful to everyone who has spent some of his valuable time discussing chess with me. Electronic T-Notes is my way of passing on much of what I've learned (and much of what I've discovered) and giving back a little something to chess (which has, admittedly, been pretty good to me).

In other words, this publication is designed with the average player in mind. This is not to say that titled players can get nothing from these pages. To the contrary, I'm sure that the software tips will be valuable to players of all levels. I only hope that titled players will not be terribly offended when I use a column for book reviews aimed at the class-level player (as I did a few weeks ago). I do plan, however, to devote an issue or two to explaining how to use ChessBase to prepare for a specific opponent, aimed at the master or grandmaster player. I've taught a couple of GMs how to do this and it always freaks them out. "How can you know so much about tournament preparation and chess software and yet know so little about chess??!!??" (This is what's known as a "backhanded compliment", by the way, something I got used to from a few tact-impaired titled players). "Just one of the fortunate few, I guess," was my reply (delivered with a smirk, in case they missed the point).

Second, Electronic T-Notes is written from an American's perspective. I make occasional reference to United States Chess Federation ratings, U.S. chess publications, books printed in the U.S. (which may not be available internationally), etc. I apologize in advance if this bothers some international readers. I am what I am. I'm not a FIDE-rated player and have not traveled widely to play chess. However, I've discovered that by being a chessplayer one can discover the world. I have friends from around the world that I'd not have met or corresponded with had I not been a chessplayer (albeit a weak one); it's yet another example of how chess has been pretty good to me.

Third, I am not a full-time ChessBase staff member any longer. If you have specific problems regarding the software or need technical support, you can find the appropriate e-mail links elsewhere on the ChessBase USA or ChessBase International web pages. If, however, you have specific comments or suggestions on Electronic T-Notes, please e-mail them to me.

Finally, welcome! It's good to have you aboard! I'm thrilled to be reaching a much broader audience and happy to be in a position to help you discover ChessBase's and Fritz' tools for improving your game. With a little bit of luck (and a lot of work) we'll all become strong chessplayers. With a whole lot of luck, we'll all become strong chessplayers and I'll finally get to meet Anjelina Belakovskaia. In any case, we'll have a great deal of fun playing and studying chess!




CHESSBASE AND CORRESPONDENCE CHESS

CAN'T SEE THE WOODS FOR THE TREES

by Steve Lopez

The first several issues of Electronic T-Notes examined game trees and how to use them. With the inclusion of tree functions in both Fritz5 and ChessBase 7, it's time we went back and took another look at trees.

First of all, just what exactly is a game tree? It's a way to look at a large number of games as though they were just a single game. The tree function takes a batch of games and merges them together into a single easy to follow game. This is better than the old method of searching for a board position, then having to load and play through the games individually.

For this reason alone, game trees are one of the most important innovations for the developing player. You don't have to remember what's played in a few dozen individual games. You can just create a tree from a batch of games and see at a glance what's been played from the position in question.

Let's create a game tree and take a look at ways to use it. Our sample game we've used throughout this series on correspondence chess has gone like this:

1.e4 g6 2.d4 Bg7 3.Nc3 d6 4.f4 Nf6 5.Nf3

We're considering the game as White, so let's go ahead and add Black's next move:

5...0-0

The first step every correspondence payer should take before consulting his or her database is to think about what they would like to play in the given position. Nothing is worse than blindly following what others have played in the past. Why is this so? Because many positions played by grandmasters require a grandmaster's knowledge and training to effectively finish and bring home the point. For example, Mikhail Tal frequently threw pieces and pawns overboard (unsoundly) to freak out his opponents. He did this (and got away with it) because he was Tal. The vast majority of us bottom-feeders would be killed a few moves after attempting these dazzling sacrifices. Yet a game tree would show only a win in this situation, failing to point out that it was an unsound sacrifice.

This is the second reason you should think about what you would play in a position: the database/game tree frequently doesn't show the "why" behind a particular candidate move. Figure out your own move, determine your own "why", and then use the tree to check your work.

In the position given above, there are several possibilities. 6.Bd3 and 6.Be3 leap to mind immediately. These are very "1960's style" moves. I can picture Fischer, Mednis, and Sherwin playing these moves in their heyday. Another possibility is 6.Bc4, while 6.e5 might also be played.

It really comes down to a matter of personal taste. I enjoy playing Advance Variations of some other openings, so I'm comfortable with the space-grabbing 6.e5 (though many strong players will definitely say this advance is premature). I would expect Black to reply with either 6...dxe5 or 6...Nd7.

Last week, we showed how to create a new database on a particular opening. Following those steps, we get a database of B09 games (the Pirc Austrian Attack). To merge these games into a tree, we first create an empty tree. Click "New" in the "Database" menu. Under "Save as type:" we select "Trees (*.CTG)" and call the file B09.CTG. Click "Save" and a new icon appears in the Database window. If you like, you can click the Information button at the bottom of the Database window to change how the new tree icon is displayed.

Next, using drag-and-drop, we'll click on the B09 database icon, drag the games over to the B09 tree icon, and drop them in. We see this window pop up:

The entries in the "Copy to tree" window have very specific functions. It allows you to specify a range of games. In this case, we want to drop all of the games into the tree. It allows you to specify the length of the lines (in half-moves or plies). By choosing "40", I'm creating a tree that's 20 moves deep. There are two choices here also: use "Length after ECO" if you want the theoretical main lines to be longer than the off-beat "side" lines. I prefer "Absolute length" because I find these side lines interesting and tend to play them frequently. Clicking "Include annotations" allows variation lines to be merged into the tree.

Once you've clicked "OK" ChessBase will merge your games into a tree. If you have less than 10,000 games in the tree, this should take just a few minutes. Trees of close to 20,000 games naturally take longer but should still take less than ten minutes on a Pentium computer. Once the tree is finished, you can double-click on its icon to open it up. Using the keyboard's cursor keys we can step through the tree to the position we're considering after 5...0-0.

The board shows our current position. The box to the lower right shows the move path we took to get here. The meat of the tree is the window to the upper right: the list of candidate moves and their statistical evaluations. This is the most helpful and the most dangerous part of the game tree.

The most helpful part for the class-level player is the list of candidate moves. This is a quick way to show what's been played previously in this position. We see that all of the moves I mentioned earlier are high on the list, with the inclusion of 6.Be2 (which I'd never even considered). So I already have a valuable piece of information: a candidate move I'd not thought about.

The most dangerous part of the tree for the class-level player is the statistical information. Why dangerous? Because, if used improperly, statistics lie like a cheap rug.

Let's digress while I give you an example. I had a friend who was once a great fan of statistical sports games (that is, sports games in which you get a card for each player and he's mathematically expected to perform pretty close to the way he does in real life). My buddy decided to make a card for himself so that he could play ball with the pros. He created a card for himself in which the only thing he hit was a home run. Every time he came to the plate he was guaranteed to whack one over the fence.

I don't need to tell you that this isn't only cheating, it's also just plain stupid. It will completely throw off other stats in the game. Pitching stats will obviously be affected. But there are other problems as well. He had an RBI leader following him in the lineup. After a few games, he complained that the game wasn't as statistically accurate as the advertising had led him to believe, because the guy batting behind him wasn't getting any RBIs.

"No kidding, you dope!" I told him. "How can he rack up any RBIs when you empty the bags every time you step up to the plate?"

My friend was only looking at the end-of-game stats. He wasn't paying a lick of attention to what was happening during the game.

That's exactly what we're doing when we're looking at statistics in a game tree: we're looking at end-of-game stats instead of what's going on while the wood is actually being pushed. Needless to say, this can be terribly misleading.

There's an interesting passage from Eric Schiller's book Unorthodox Chess Openings that's been quoted a lot on Usenet lately that reads as follows:

Statistics have no place in the study of openings. The simple fact is that there is no strong correlation between the evaluation of an opening and the result of a game. If an opening is convincingly refuted, it doesn't matter what its prior tournament record is. Opening fashions change and popularity is by no means an indication of any objective merit in an opening.

Contrary to what a few vociferous blowhards on the Internet would try to lead you to believe, Schiller's right. An opening line might have been played a few dozen times in tournaments, but if some clever GM finds a means to bust it, it's history. Your statistical tree won't show you that, though (unless you do a bit of digging). You'll just see that a line's been played over 40 times with an 85% success rating and go right for it. Your opponent, meanwhile, who's just read a report on the bust in Inside Chess, will play the killer move and you'll be sunk.

Similarly, a move might have only been played one time with a 100% success rate. It might be a killer move. Then again, it might just mean that the opponent was an idiot.

Schiller, with his line about "no strong correlation" between the opening and its eventual result, may be overstating the case a bit, but his remark is a valid one. Two players might play the first fifteen moves flawlessly but the remaining moves between the start of the middlegame and final resignation can be a virtual "comedy of errors". As Tartakower once said, the player who makes the second-to-last mistake is the winner.

Another potential pitfall for those of us down here in the fishpond is that a particular opening may be fine for titled players but may be certain death for us at class level. Certain openings require positional and theoretical middlegame knowledge far beyond that possessed by the average player. These openings may have great statistical results and are well-suited for use by titled players but leave room for endless errors by the rest of us and should be avoided.

In short, statistics should be used as guideposts but should never be followed blindly. So how does one use them? Exactly the way I did several paragraphs ago: decide on your move before consulting the game tree. Then use the tree to check your idea and possibly spot additional ideas that you hadn't considered.

Let's look at the game tree once again:

You'll recall that I was considering playing 6.e5 in this position. It's been played a lot (324 games), has a reasonable success percentage (58%), and the average rating of players who have used the move is into the Master level (2285). Seems safe enough at first glance, but the best thing for me to do is step though some more of the tree to see if there are any obvious surprises.

Notice that some of the statistical listings in the tree are given in grey. This means that there's not enough of a statistical sample for the numbers to have any real value. Ergo, take the 70% success rating of 6.a3 with a large dollop of salt.

How do I ultimately decide whether or not to play 6.e5? I need to look ahead through various branches of the tree to see what kind of positions arise after 6.e5 is played. For example, I predicted that Black might play 6...dxe5 or 6...Nfd7. However, I failed to consider both 6...Ne8 and 6...Ng4. I need to search the database to find the games in which these moves were made to see if there are any ugly surprises in store for me.

In the database window, right-click on the icon for the database from which the tree was created. Select "Properties" and mark the box that says "Reference database". Then in the game tree, with the position in question on the board (say after 6...Ng4), right-click on the list of candidate moves and select "Search games". ChessBase will search through the database and put all the games in which that position appeared on the clipboard for your perusal. If it's a small number of games (in this case, four), highlight all of them and click the button on the lower left corner of the Clipbaord window. ChessBase merges all the games into a single game, with the current position on the board and highlighted in the move list, so that you can see exactly where all the branches occur.

What if you'd like to look at all the games in which 6.e5 was played (all 324 of them)? Is there a method for doing this? Certainly! Right-click the candidate move list and select "Search games". All 324 games will be sent to the Clipboard.

The first thing you'll see on the Clipboard should be the "Survey" games from the ChessBase Opening Encyclopedia (if you followed the method I suggested in last week's ETN). Play though these first. Follow the main line of the Survey game initially, then take a look at the main (first-level) variations.

After playing through the Survey games, return to the Clipboard. Click the "Sort Clipboard" button (the new button depicts an "A", a "Z", and an arrow pointing downward). A window titled "Sort by" appears; select "Result" and click "OK". All of Black's wins will be grouped together; likewise for draws, White wins, lines ending with an ECO evaluation, and lines ending with no evaluation. Since we're pondering what happens after White plays 6.e5, we'll want to look at White wins first.

There are an awful lot of White wins to plow through! The easiest thing to do is play through annotated games first (that is, any game that displays letters in the rightmost column of the Clipboard listing). After playing through these games, pick out some other games to play through (I like to look for the names of familiar players; in this particular case the names "Unzicker" and "Stein" jumped off the screen at me).

Next look at some annotated draws and losses. Then select a few unannotated games to play through. View these games at your own speed. If you use the VCR controls to watch the games, pick a comfortable speed, one that allows you time to figure out what's going on. It's not unheard of to be able to play through fifty or more games in an evening. Try to spot common strategic or tactical themes in the games. Perhaps the c-file frequently opens and White controls it by playing Rc1. Maybe it's common for Black to establish a Knight outpost on d4. Both sides may be trying to break the center by advancing a flank pawn (either the c- or f-pawn). Look for themes that lead to losses. Perhaps White is sunk if his light-squared Bishop leaves the a2-g8 diagonal.

This is the point at which printed materials also come in handy. Correspondence chess is a rare case in which the piles of "Winning with..." type of opening books actually come in handy for the club-level player. You can look up opening lines and (hopefully) find some explanation of what's going on. In the absence of one of these specialist books, general opening books that explain ideas are a good bet (I've reviewed these books at length in past issues of ETN).

After viewing some games, you should have a pretty good idea of whether or not you like the positions that arise after your candidate move. You should also have some grasp of the major ideas for both sides in those positions. Armed with this knowledge, you're ready to return to the game tree.

Right-click on the candidate move box in the game tree. Select "Display main lines". After a few seconds, ChessBase will display a list of the main variations that have been played from the position on the board. This list can be scrolled through both vertically and horizontally, to allow you to see at a glance the most popular variations. You can click on a move in this box and the tree will jump right to that move. ChessBase will then perform a recalculation of the main lines branching off from the new position.

A truly powerful tool can also be reached by right-clicking in the candidate move list: "Critical variation". This is a single variation that shows what will happen assuming that both players make the best possible moves. ChessBase decides which moves are the best by looking not just at the statistical success of the move but by player ratings as well. This is a great feature not only for correspondence chessplayers but also over-the-board players who are looking for strong variations to add to their repertoires. In correspondence chess, this feature can help you map out how to react to your opponent's possible responses. Note, also, that since this is strictly a statistical feature (as opposed to a chess engine thinking up a move) its use is legal in correspondence play.

However, please take notice of the fact that "Critical variation" is still a statistical function; therefore it should be used judiciously. Your mind should still be your primary resource in considering a move in correspondence play.

We now have a methodology for using a game tree:

1) Look at the position and identify the candidate moves;

2) Select one of these moves as your preliminary first choice, also looking ahead at your opponent's possible responses;

3) Open the game tree and step through it to the position after your opponent's last move;

4) With your side to move, examine the list of candidate moves, looking for moves you hadn't considered or an abysmally low success rate for the move you were considering;

5) Step through the game tree and look at some of the positions that arise after your candidate move;

6) Use "Search games" to send games containing the current position to the Clipboard. Sort the games, then play through games in which the color you're playing won. Then study games in which your side lost as well as draws. Try to find common themes in these games. Feel free to consult printed sources that explain the ideas behind the opening.

7) Return to the game tree. Step through the tree again; this time through you should understand the key ideas better. Use "Display main lines" and "Critical variation" to show you at a glance what has been played before and what the best line of play should be according to past history.

After you've followed these steps, you should be able to decide whether your candidate move is a good one. Make notes on your candidate move and your opponent's possible responses. You could write these down or else keep an analysis database with notes on your postal games. These notes will save you time later when your opponent's next card comes in the mail.

In the final tally, gamne trees are an astounding tool, but they do not replace good old-fashioned skull sweat. Use your head to decide on a move, not what a game tree says is "good". But please do use the tree to check your ideas and to discover new ones.

A note to Fritz5 users: some of these functions do not appear in Fritz' tree (notably "Display main lines" and "Critical variations"). However, Fritz does contain the cool little bar graph at the bottom of the candidate move window that displays wins and draws for the highlighted candidate move. ChessBase lacks this feature.

You may have noted that I've left out the simple [SHIFT-F7] function for position searches, as well as the Opening Report feature. These are also useful tools for correspondence players and we'll look at them in the next Electronic T-Notes. Until then, have fun!

You can e-mail me with your comments and suggestions for Electronic T-Notes.