ELECTRONIC T-NOTES


CHESSBASE USA'S WEEKLY ON-LINE NEWSLETTER


FOR THE WEEK OF FEBRUARY 20, 2000


A USER'S GUIDE TO FRITZ6 -- PART 7 (OPENING BOOKS I)

by Steve Lopez

The term "opening book" in relation to Fritz6 doesn't have a dang thing to do with the hordes of (generally useless) chess books bending the bookshelf of every serious chessplayer, tomes with such impressive titles as Winning with the Beavis and Butthead Gambit or Mastering the Dilbert Defense. So if you've been frantically clicking on items in Fritz' menus, searching for some grand compendium of opening instruction, you can relax now. You're not overlooking anything.

An opening book (or opening tree -- the terms are used interchangably) is, in its most basic form, a way to keep a chess program from playing the same opening over and over. Years ago, I had a chess program called Cyrus. It had a decent 3D display, a couple of tweakable options, and gave me a competitive game without killing me -- in short, it was fun for about a month. The problem was that the copy I bought was missing the opening book ("never buy a piece of unpackaged shareware from a 'mom and pop' computer store" is the moral here). Cyrus always (and I mean always) played the Petroff Defense as Black. This was OK for me, as I got a lot of practice with the Cochrane Gambit. But I never knew that the program was supposed to have an opening book until I played my dad's copy on his computer and was startled to find myself in a Ruy Lopez instead of a Petroff.

The problem was that my copy didn't have an opening book, so every time I pushed 1.e4, the computer had to start thinking of a reply. It didn't have a standard library of openings from which to choose a response. It always came up with 1...e5. I played 2.Nf3 and after a bit of deliberation, Cyrus always replied 2...Nf6 -- the Petroff, every stinking time.

An opening book gives a chess program a library of standard openings from which to pick and choose and thereby vary its opening play (plus the program doesn't have to reinvent the wheel every time it plays a game). A randomizing element in the program will essentially "roll the dice" to choose a response from the book (much like a "random events table" which will be fondly remembered by old wargaming grognards).

Fritz6 comes with an opening book called general.ctg, which is a set of openings which maximize the strengths of chess computers while minimizing their weaknesses (preferring open tactical games while avoiding closed strategic ones). You can easily access this opening book if the program isn't doing so already. In the main chessboard screen of Fritz6, click the File menu, select "Open" and then select "Openings book" from the submenu that appears off to the side. The Windows file select dialog will appear -- use to to go to the CD and select "general.ctg" and click the "Open" button (more info is available in the ETN issue for December 12, 1999).

Once you have this sucker loaded, you can have a look at the book's contents by clicking the "Openings book" tab in the "Notation" pane of the main screen. When you do, you'll get something that looks like this:

What you're seeing is Fritz6's opening library -- in other words, what Fritz6 "knows" about the chess openings. This book was created by merging a whole slew of games (that's an English slew, not a metric slew) together into a single "tree". The tree was then edited (or "tuned") by a human to maximize the strengths of chess computers. Fritz (or any program using this book) will strive for tactical openings and try to avoid strategic openings or unsound openings. As we've discussed previously in ETN, a computer's main strength lies in tactical calculation, while closed positional games that require long-term planning are where humans excel.

I previously beat the subject of statistical game trees to death in the first few issues of ETN (March and April 1997), so I won't go into the subject in depth here. We'll hit the high points here; more information than you could possibly want can be found in those first few ETN issues.

The new (as of autumn 1997) opening book format in Fritz and ChessBase is superior to the old .fbk opening book format, in that the new format recognizes transpositions. Most chess positions can be reached via a variety of move orders. For example, the standard Ruy Lopez move order is 1.e4 e5 2.Nf3 Nc6 3.Bb5. However, if the players (for some bizarre reason) elect to play 1.e4 Nc6 2.Bb5 e5 3.Nf3 they'll reach the identical standard Ruy Lopez position. This is called a transposition of moves; the .ctg book format recognizes these transpositions, while the old .fbk format didn't. All Fritz would recognize using an .fbk file was an exact move order. If 1.e4 Nc6 2.Bb5 e5 3.Nf3 wasn't included in a .fbk opening book (and chances are that it wouldn't be), Fritz wouldn't know that this was a Ruy Lopez and would have to compute its own reply to 3.Nf3 instead of responding with a known book move from its library.

It's pretty easy to navigate through the game tree. You can click on a move in the list to have it made on the board, or use the four cursor keys on your keyboard to go up and down the list and step forward and backward through the moves.

Let's have a look at the elements of the opening book display.

The first thing you'll notice is a nice picture of a tree in the upper left corner. This isn't terribly significant, but what's directly under it is: the name of the opening book you're using (in this case, general.ctg). This is important because Fritz lets you create your own opening books or use books created by others (so you're not locked into using just one opening book). You can see at a glance which opening book you're using.

The next item to the right is "N". Remember I told you that an opening book is created by merging together a number of games? This column tells you how many games a particular position appeared in. The upper entry in a column gives you the figure for the database as a whole, while the number after a specific move shows you how many games that individual move was played in. The opening book that comes with Fritz was derived from a total of 86,734 games. The move 1.d4 was played in 35,312 of them.

The next column shows you a percentage. This is the success rate for a given move or position. This is arrived at from a straight averaging function. In the games from which general.ctg was derived, White has a 56% success rate. This does not mean that White won 56% of the games. It means that White received 56% of the total number of points available to the players of those games (remember that in tournament chess, a win is worth a point while a draw is worth a half-point). Basically, the higher the percentage, the better the move was for White. The lower the percentage, the better the move was for Black.

When the averaging function is internally invoked at the time an opening book is created, a White win is scored as 100. A Black win is scored as 0. A draw is scored as 50. There are a number of other percentages used for averaging purposes when a game in the database was incomplete (just an analysis line ending in an Informant symbol evaluation. You can find a list of these percentages in one of those early issues of ETN I previously mentioned.

Making a long rant short, these percentages should be used by you as guidelines as to which moves work and which ones don't, but raw statistics can lie in many ways. To read the full rant, once again refer to those first three or four ETN issues.

The next column is marked "Av.". This shows you the average Elo rating of the players who chose a particular move. "Perf" is the average performance rating of these players (this is a function of how well the player did, derived by comparing his rating to his opponent's rating).

The last columns deal with Fritz' "weighting" functions, which determine how often Fritz will play a particular move, modified by the success rate of the move in grandmaster play in addition to how well Fritz has done in that same position in games Fritz has played on your computer. We'll likely cover this in a later ETN issue.

There are a number of ways to modify Fritz' display of the opening tree. Right-click in an empty gray area of the opening tree display. A menu will appear with various options. Note that some of these options will be unavailable (will be "greyed out") if you are using an opening book on a CD instead of on your hard drive.

"Search games" is a shortcut that allows you to search for all the games containing the current board position in whichever database you're using at the time (see ETN for the last two weeks for more on databases and searching). "Sort" lets you change the sorting order of the move list in the display:

"Delete whole tree" wipes the whole tree out, so be very careful with this.. "Weed tree" was covered in ETN for August 22, 1999 (with a slight correction given in the following issue for August 29, 1999); it's used when you import a bunch of new games into an existing tree.

"Allow move adding" enables you to add moves to a tree by hand (and we'll cover this later on). "Choose font" is pretty self-explanatory, except that you must use a "Figurine" truetype font supplied with Fritz6 in order for the tree to display properly. "Close book file" is a shortcut to let you close the current opening book and load a different one. "Close" completely closes the "notation" pane -- go to the Window menu and select "Panes" to get it back.

The other command from this menu is "Properties". This allows you to select additional information to be displayed in the tree:

"NP=" shows you the total number of positions in the opening book. General.ctg contains nearly 2 million individual positions. Checking "Unplayed transpositions" displays moves that aren't actually part of the book but lead to positions that are in the book (very useful for postal players looking for a surprise transpositional weapon). Checking "Retro moves" displays the previous move or moves that led into the currently displayed board position. Unchecking "Show Elo numbers" eliminates the "Av" and "Perf" columns from the display; this is useful if you're using a low graphics resolution and want to see the Fritz weighting percentages rather than the Elo averages on monitors that don't show the full opening tree display (without shrinking the board to the size of a postage stamp).

The last item is "Statistics". This enables a useful bar graph display which provides win, loss, and draw totals for individual moves in the tree:

For example, knowing that the move c2-c3 in a certain position scores 42% (somewhat better for Black) is certainly useful information, but the bar graph shows exactly how many White wins, draws, and Black wins there were in games in which this position occurred. Note that the bar graph refers to the move that's currently highlighted in the tree, not to the current board position. In this particular case, we see that White won only a third of the games while Black won half of them. White's best bet is to avoid c2-c3 and play something else.

There are a lot of other functions of the opening book and we'll look at them 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 and sacrificial play, stop by my Chess Kamikaze Home Page and the Yahoo Chess Kamikazes Club. Currently at the Chess Kamikazes Club, you can read some interesting anecdotes about the last Dutch GM Lodewijk Prins, view other members' ideas about the Jerome Gambit, and participate in these discussions (or even start one of your own) by joining the club.