ELECTRONIC T-NOTES


CHESSBASE USA'S WEEKLY ON-LINE NEWSLETTER


FOR THE WEEK OF AUGUST 1, 1999


USING YOUR COMPUTER AS A RESEARCH ASSISTANT (PART ONE)

by Steve Lopez

I've been playing, studying, and writing about chess for a long time and I've learned a (very) few things about the game. Among them are:

Not to sound like your grandpa (who used to walk fifteen miles barefoot through the snow to get help with his chess analysis from the master in the next town), but I still remember the days when PCs weren't available and everyone had to do their own analysis. There was much head-scratching and rending of garments as players tried to figure out where they had gone wrong in games they'd lost. This was not necessarily a bad thing (aside from the torn shirts), as many of us learned to analyze positions fairly well, if not during the game then at least in the post-mortem.

The arrival of affordable computers changed everything. Now we can just click a button and have Fritz or Junior or Hiarcs or Nimzo or Crafty or any of dozens of other playing programs show us where we went wrong. While this is very convenient, it's also very dangerous. It makes us lazy. It can also be extremely misleading, because chessplaying programs don't always know what's best in every position.

But there is a way to get the best of both worlds. The trick is to find ways to use your computer as a research assistant to give you ideas and point you in the right direction while at the same time not allowing the infernal device to do all of your thinking for you. In this issue of ETN (as well as the next), I'm going to offer up some ideas on how to accomplish this. As always, take what I say with some small degree of skepticism (I'm not a grandmaster, after all); I'm just going to talk about ideas that have worked for me. Use them, modify them, ignore them -- but at least consider them. The key concept underlying these two articles is to learn to use the tools but also to always think for yourself.

As my close friends know, I'm a complete hound for chess software. I collect chess programs the way Imelda Marcos collects shoes. I have over a hundred programs in my collection (and there are a couple of hundred others that are available on the Internet that I haven't got around to downloading yet). About a year ago, I was horsing around with a commercial program I'd purchased that had different "personalities" to play against. One of these "guys" was a big fan of the Schliemann Defense in the Ruy Lopez. Despite being an avid Ruy player, I'd never encountered the Schliemann. After having my head handed to me a few times in this opening, I cracked a repertoire book and followed the line it recommended so that I could at least get out of the opening alive. Of course, there came a point at which my "book" knowledge ran out and I was forced back upon my own resources.

Here's how the line went:

1.e4 e5 2.Nf3 Nc6 3.Bb5 f5 4.Nc3 fxe4 5.Nxe4 d5 6.Nxe5 dxe4 7.Nxc6 Qg5 8.Nd4+ c6


At this point, I had no clue as to what was going on. What's the "right" move at this point? Let's take a look at the position for a moment.



I'm faced with two threats. The c6-pawn is not only blocking check, but is also threatening to cop off my b5-Bishop. Meanwhile, over on the Kingside, Black's Queen is bearing down on my g2-pawn.



Since I'm a pragmatic guy, I didn't try looking for some clever finesse that would allow me to give up a pawn but get future considerations for it. I just bit the bullet and played 9.Bf1, killing two birds with one stone: the Bishop moves out of danger while protecting the g-pawn.




My development is suffering a bit for this Bishop move, but I figured that I could make the time up later in the game.

I went on to win the game and, since this particular program I was playing had a small opening book and no learning function, I was successful with this line in future games as well. But I started wondering about 9.Bf1. How good is it really? Is it worth playing in real life against my fellow humanoids or is it just a trick I could use to beat this particular computer program?

My first step was to crack open a few books. I discovered that the move wasn't covered in any of my available literature. In fact, I discovered that 8.Nd4+ wasn't a "book" move, that 8.Qe2 was the preferred move at this point. No offense to the world's GMs, but I wasn't crazy about 8.Qe2 because I think White has too many loose ends dangling. Black still gets pressure on the Kingside, plus a timely ...a6 gives White some grief over on the Queenside. While this stuff is OK for strong players, I'm not skilled enough to work through all the twists and turns over the board. I decided that playing it safe with 8.Nd4+ still gives me the extra pawn with the attendant material advantages. Hey, let's make Black prove that the gambit was sound!

So looking at the board after 8.Nd4+ c6 9.Bf1, I decided that this was a position I could probably live with. The e4-pawn is isolated and can easily be dealt with later. Black has developed his Queen early (a standard no-no), providing me with a possible target. And there's a lot of space around his King. He'll need to figure out where to place his light-squared Bishop so that he can castle to get the King out of the center. So my opponent has a few things to think about.

It's at this point that I decided to use some software tools to double-check my analysis. There are two ways to do this: have a chess engine look at the position or check a database to see if the line's been played before and how the games turned out.

My personal preference is to check the database first. I'm interested in seeing what the big boys have done when they found themselves in the same spot before I let a chessplaying program chew on the position. After firing up ChessBase 7, I reconstructed the position after 9.Bf1.

I set the Mega Database as my reference database. To select a reference database, just click once on the database's icon to highlight it, then click the "Information" button at the bottom of the Database window (it's the button that looks like a blue dot surrounding a white letter "i"). You'll get a new window in which you'll find a check box marked "Reference database". Click it to check it and then click "OK". You'll now see a new symbol added to the icon for the database you've chosen. The symbol looks like a scrolling piece of paper -- this shows that the database is the reference database. You'll generally want to select your largest database as the reference database, unless you have a smaller database just on the particular opening you're studying. Since I was new to the Schliemann, I'd not created a new database of Schliemann games -- I just went with my largest database.

Going back to our example, I've selected Mega Database 99 in reconstructing the research I performed. At the time this actually occurred (in 1998), I was using Mega Database 98 of course.

Once I had the board position set up and the reference database selected, I clicked "Opening report" in the Game menu. This is (in my opinion) one of the most amazing features of ChessBase 7. This will create a complete report on the opening, containing information that would take literally days to compile by hand. ChessBase 7 acts as your research assistant, just the way a "second" does in the world chess championships. Instead of making some poor slob work all night to put togther this information, you simply have the computer do it in a few minutes.

Once ChessBase has completed its research, it presents us with an opening report -- a complete breakdown of the opening fit for a grandmaster. The first screen of the opening report looks like this:

The screen provides us with a nice title, the moves leading to the position, a graphic thats illustrates the last move, and a note telling us how many games were found in the database. This note is also a hypertext link; clicking it will bring up the Clipboard with a list of the games in which this position appeared. As always, we can double-click on any of these games to be able to play through it.

The next screen tells us something about the history of the variation:

This gives us a nice bit of background on the position. The move's not a new idea -- someone tossed it out there against Ossip Bernstein in 1905. It's still being played, too -- the last time was in 1997. We have direct links to be able to view both of these games. And there's a bar graph showing how often it's been played annually (twice a year, tops).

Next we come to a section about who has played this move:

In this particular case, the line "Strong grandmasters who have used this line as White" is a bit of an overstatement. Only two of the mentioned players have a performance rating of over 2400 while using this line. However, the program has provided me with a list of the strongest players (for whom ratings are available) who have tried this move. I again get direct links to the games in question, as well as links to the player's profiles in the Player Encyclopedia (in case I'm burning with curiosity as to who Alexandre Dgebuadze is). I can see their scores in this opening as well as the dates in which they played it.

All of this stuff has been pretty interesting so far, but next we'll come to the real meat of the opening report:


Now I can see some numerical data for the move I'm researching. This is potentially one of the most dangerous areas of the opening report, because it contains statistical data.

I've ranted about statistics in earlier issues of ETN (in fact, I did it in the very first issue of ETN), but I'll run through an abbreviated version of the rant for those who arrived late. Statistics will lie like a used Yugo salesman wearing an ugly cheap polyester plaid suit and smoking on a nickle cigar, who is trying to sell you a model with over 100,000 miles on it for three times the fair market value (i.e. twice the price of his cigar). In other words, you can't take statistics at face value!

Let's dissect the data here. White scores well (65%); on the surface, this looks nice, but it may well be that every White player who won a game here was playing against a rhesus monkey. The only way to be sure that the wins are deserved is to play through the games and make sure that Black didn't make completely idiotic moves and virtually hand the win to White. (Unfortunately, there's no way to be sure that none of the Black players were rhesus monkeys anyway. It might be possible that Ossip Bernstein was later found tied up and gagged in a broom closet in his underwear while a rhesus monkey was sitting at the board, disguised in Bernstein's clothes. Notice that I said "possible", not "likely". See? If words can be misleading, just think of how tricky numbers can be -- which is precisely the point I'm trying to make).

Next I see some performance ratings for the participants in the games. The more successful a move is, the higher the average performance rating will be. However, this requires knowledge of math and that's too much like work, so I skip this part (I simply note that White's performance rating is a positive number, which is enough information for a mathematically-challenged dolt like myself to realize that this is a good thing).

White has won 62% of the games. Nice total -- however, refer to my rant above about playing through the games. There are very few draws. This is useful; the move evidently leads to dynamic positions as opposed to static ones. Time to bone up on my tactics. The length of game information is interesting; it looks like it might require a big early mistake by White to give up the game. However this also indicates that big mistakes are definitely possible -- perhaps there's a common pitfall here that I need to avoid? Once again, the way to find out is to play through the games.

Keep in mind this fact about statistics: the larger the sampling, the more accurate the statistics. In other words, the more games you have to work with in a particular variation, the more accurate the information will be. In this case, thirteen games is not much of a data pool. But since this is a really oddball variation, it'll have to do.

Next we come to the most useful part of the opening report:

This is a huge timesaver when studying an opening. ChessBase 7 took a look at what was played by Black in this position and offered up the most likely replies I'll see and gave me recommendations as to what I should play in reply. Keep in mind that ChessBase did not call on a chess engine to provide this information; it simply looked at the games from the database and derived its recommendations from that information.

Rather than provide three graphics of the three replies its showing me, I'll just use a graphic for the first possibility and describe the other two.

The three replies that ChessBase tells me to expect are:

9...Qe5
9...Bg4
9...Bc5

It provides these moves in the order of frequency in which they appeared in the database. The move 9...Qe5 was played five times, 9...Bg4 was played four times, and 9...Bc5 was played twice. This doesn't mean that 9...Qe5 is necessarily better -- it just means that it was played more often than the other moves.

Looking at 9...Qe5, I can see the number of games (five), the span of years in which it's been played (1954-1996), and the median year in this span (1980). Again, the statistical information would be much more valuable if the opening in question was more popular and there was a greater pool of games from which to draw information. I see that Black scores below average (40%), there were three games available with Elo ratings provided for Black, in which the average rating was 2348, and the average performance rating for these players was 2246 (it's lower than the average rating, meaning that Black didn't do so well with this move).

I can also see the names of players of the Black pieces who have tried the move, their ratings, how well they did, and by clicking on their names I can play through their games.

The next section is the most important. ChessBase has recommended a reply for me to play and provided a link to a game in which that move was played. The recommended move is 10.Ne2. This recommendation is derived from a formula combining the statistical success of the move, the strength of the players who made this move, and a tree look-ahead function that explores the statistical probabilities in the branches that lie ahead in the tree of games containing this move.

So what's my next step? First, I need to understand why Black played this move and why White replied in this manner. The move 9...Qe5 is pretty easy to understand. It centralizes the Queen, protects the e4-pawn, and attacks the White d4-Knight. So the obvious reply is to move the Knight. There are really only two choices: 10.Ne2 and 10.Nb3 (I reject 10.c3 out of hand because I think it weakens d3 too much). The move to the b-file puts the Knight way off toward the side of the board. While 10.Ne2 buries the light-squared Bishop temporarily, it puts the Knight near the center (where all the action is). So 10.Ne2 is obviously the better move in my opinion. But to get a firm grasp of what's involved here, I need to play through the games in which White chose 10.Ne2.

The second choice (and the one which I expected in my own game in this line) is 9...Bg4. It develops a piece and attacks the White Queen. ChessBase recommends 10.Ne2 once again, blocking the attack. Black also doesn't do well in this line, scoring a poor 37%.

However, the third move is pretty interesting, because Black gets a 50% in this line: 9...Bc5. This, too, attacks the d4-Knight. But White has a counter-attack: 10.d3 with a discovered attack on the Black Queen from the dark-squared Bishop on c1. To find out why Black does (relatively) all right in this line, I'll again need to play through the games.

Note that the more games you have as a basis for the opening report, the more information you'll receive. You'll see suggested piece placements in the variations ChessBase suggests, as well as more alternatives in the recommended moves it presents to you. With a small thirteen game pool, the information presented is pretty limited and some of the features that normally appear in the opening report weren't generated by ChessBase in this instance.

So what do I have here? I have three recommended moves for Black that I can expect to face, recommended moves that I should play in reply, and shortcuts to important games that I should examine in the search for themes and ideas.

At this point, I know I'll want to come back to these thirteen games, so I bring the Database window back up on top and create an empty work database. I return to the opening report, click the link on the first screen (which gives me the Clipboard containing all thirteen games), bring the Database window back to the top, and then drag and drop the Clipboard over to the new database's icon to copy the games into it.

Rather than save the opening report into Mega Database 99, I can just run it a second time on this smaller work database later to get the same information. I can then save the opening report into that database instead of the Mega Database.

At this point a great deal of the preliminary work has been completed. I have an opening report that shows some ideas for both sides after 9.Bf1. The entire line at this point doesn't seem bad to me, but there's still a lot of work to be done. We'll continue our research into this position and examine some more uses for computers as a chess research assistant in next week's ETN. Until then, have fun!

You can e-mail me with your comments, suggestions, and analysis for Electronic T-Notes. If you love gambits, stop by the Chess Kamikaze Home Page and the Yahoo Chess Kamikazes Club.