by Steve Lopez

As you're likely aware by now, ChessBase USA no longer offers five day a week tech support. We've learned that the vast majority of our software's users have Internet access and over the last year I discovered that most of the tech questions we received were being sent by e-mail rather than being phoned to us. Consequently, it was no longer cost-effective to offer full-time phone support. Telephone tech support is now offered on Friday only; however, questions can be e-mailed to us at any time.

This brings us to an important point: if you e-mail a technical question to us, you must provide a valid return e-mail address in order for us to respond to your question. Don't laugh. This sounds pretty basic, but it's evidently not obvious to a few people. About 10% of the e-mail responses I send out get bounced back because of an invalid e-mail address. If you "munge" your e-mail address in the header of your e-mail, we have no way to respond to your question -- the reply will simply bounce back to us and you won't receive it . So if you've e-mailed us over the last few weeks and have failed to receive a reply, this is the likely reason.

Many questions sent via e-mail have been answered in past issues of Electronic T-Notes. If you have a question about major functions of the programs (the hash tables in the playing programs and the repertoire database function in ChessBase 7 are particularly popular topics for tech questions), it's likely been covered in ETN. For our international readers who have been reading ETN only since July 1998 at the ChessBase GmbH web site, there's an additional 15 months' worth of ETN (from March 1997 to June 1998) archived at the ChessBase USA web site. There's also a FAQ (Frequently Asked Questions) page at the ChessBase USA site.

If you do send in a question that's been covered in a past issue of ETN, the likely response will be for me to send you the URL link for the article. Two or three people have expressed varying levels of offense at what they perceive to be an "impersonal" response. Sorry, no offense is intended. The entire reason for ETN to exist is to provide useful information in a freely-accessible public forum and prevent me from spending 20 hours a day writing individual responses to the same identical questions over and over again.

There have been a few requests for updates/revisions/rewrites of past ETNs that are archived in the .zip files at the ChessBase USA site. I have no immediate plans to revise the 10-part series on learning a new opening (though I'll likely get around to it someday), but the shorter pieces are prime candidates for such revision. In fact, this week's ETN is an update of an article from September 1997. So, without further ado...


by Steve Lopez

I still receive a great many questions about the graphic analysis display in Fritz and the other playing programs we offer, despite the article which appeared over two years ago in ETN. So I've decided that it's time to revisit this topic.

When you start a game against one of the programs, the graphic display looks like this:

Before we start a game and begin getting information in this window, let's have a look at an empty one and learn what we can about it.

Anyone who has ever sworn at an algebra teacher will recognize this as a basic graph. The black line across the center of this window is a move number display, denoting the move numbers in the game. The move numbers (2, 4, 6, 8, etc.) appear at the bottom of the window; there's also a hash mark on the center line which corresponds to each of these moves.

There are two numbers at the upper left and right corners of this window. The one to the upper left denotes material while the one to the upper right denotes thinking time. The default for material when you start a game is 250cp -- this stands for 250 centipawns. A centipawn is 1/100th of a pawn (100 cp = 1 pawn). Thus 250 centipawns is equal to 2.5 pawns. The number to the upper right is set to 0s which means zero seconds.

Let's take a look at the window in action. First here's the graphic display from a game I played against my old nemesis Doctor? 2.0 (longtime readers of ETN are no doubt laughing at me right about now):

First we'll take a look at the jagged black and blue lines running across the window. These lines are a graphic display of how much time each player used for his/its moves at particular points in the game. The black bar shows Doctor's time usage while the blue bar shows how I used my time. We can see right away that Doctor had a very long "think" at move 17 -- the line peaks at the top of the graphic display window. But how do we know exactly how long it pondered at that point?

This is where the number at the upper right of the display comes in handy. It shows 623s, meaning 623 seconds. This means that the highest peak on the graph corresponds to 623 seconds. Doctor thought for over ten minutes before playing its 17th move.

Note that I used considerably less time throughout the game than did Doctor. I used less than a minute for each of my moves, which is why the blue line barely registers on the graph. The closer a time line is to the center line, the less time the player thought at that point, while the closer a "peak" is to the top of the display, the longer the player took to consider his move. My two longest "thinks" were at moves 10 and 12, though both were still under a minute.

Now take a look at the bar graph portion of the display. Green bars appear above the center line and mean that the human player was ahead. Red bars appear below the center line and mean that the human player was losing.

I was carefully examining the numerical analysis Doctor was providing as I was playing the game. Throughout the opening, the evaluation went no higher or lower than 0.15 pawns either side of 0.00. Consequently, we see almost no bar graph activity in the early going. If you look closely, you can see a small green bar at moves 4 and 5 -- this means that I had a very slight edge (about 0.10 pawn) during those two moves.

Later in the game, we see that I started to suffer a bit. Starting at move 11, the evaluation stopped being equal and Doctor began to get a slight edge. During this time I was starting to organize an attack against its castled King while it counterattacked strongly in the center. So I was just repositioning pieces while Doctor begam pressing forward in the center and started to cramp my position a bit.

Finally, I was ready (or so I thought). At move 16 I threw my Queen into a hole on Doctor's Kingside. I saw the big dip in my evaluation, but thought it was due to the fact that I was playing in "Friend" mode --- Doctor was deliberately misevaluating the position as a means of matching my play and giving me a chance. It turns out I was wrong.

On move 18 I sacrificed a Bishop to blow open the Kingside. But Doctor had a defensive resource that I'd missed. The evaluation took a nosedive and, after looking at the variation Doctor was displaying, I decided to throw in the towel (again, as I always do against Doctor? 2.0 -- stop laughing please).

You can plainly see the sudden evaluation drops at moves 17 and 18. How big were they? That's where we look at the number in the upper left of the graphic display window. This number changes periodically (as does the time figure in the upper right corner) to show what the largest bar on the graph is worth in 1/100ths of a pawn. In this case we see that it reads 150cp; the largest bar in the graph (the one for move 18) is equal to 150cp -- one and one-half pawns.

The rules of thumb are as follows:

Here's another (depressing) example of a completely losing game played by yours truly. This one was against Nimzo7.32 set for "Aggressive" level in "Friend mode":

Some "friend"; Nimzo flogged me mercilessly. I wasn't in too bad shape until around move 10. Things then started to go downhill. From moves 14 through 18 I held my own, made a slight comeback for a couple of moves, then went deeper into the tank at move 21. I finally got smacked at move 24 (the big red bar; you can see that the corresponding evaluation is 300cp, the equivalent of three pawns or a minor piece. The interesting part is that the material on the board was still even, but because of a sudden Knight fork, I was about to lose a pawn. The remaining 200cp deficit was due to my generally crappy position).

Perhaps someday I will learn to stop playing gambits against computer programs. Then, when I revise this article again two years from now, I might have a few displays containing green bars to show you as examples.

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 my Chess Kamikaze Home Page and the Yahoo Chess Kamikazes Club.