by Steve Lopez

Here is is, by popular demand...

Before we get started, I'll present the usual notes, caveats, whining, etc.

Over the next few weeks, we're going to examine the various features of Fritz6 and how you can use them to help improve your own chess abilities. We won't be looking at all of the features (I want this series to be marginally shorter than War and Peace), just the ones I think are the most significant. As always, your mileage may vary, but feedback is always welcome. My e-mail link is near the bottom of the page. If you don't get a personal reply to your comments about this series, please don't be offended. I currently get about twenty to thirty ChessBase/Fritz related e-mails a day and I expect the number will increase as this series progresses. Rest assured that I read all of my e-mail (except for the spam and ones from the half-dozen or so people I've killfiled).

We also won't be delving in-depth into any one specific feature; otherwise this series will never end. Many of Fritz' features have been covered in previous articles (for example, in 1999 I wrote individual articles on the "Spy" and "Show threat" functions). We'll hit the high points of using these features, but won't go into deep discussions of them; I'll try to remember to include references to previous ETN issues whenever applicable.

Many of the tips you'll see here will apply to any chess program that contains the relevant features. If you have an old freeware DOS chess program from the late 1980's that allows you to set levels and time controls or edit an opening book, you'll might find some of the information in this series useful. The specific procedural steps we'll look at will apply to the current Fritz6/Junior6 interface (as well as the next versions of Nimzo and Hiarcs, which will likely use the new interface).

Much of what we'll look at will be very basic information to some users. If you're an experienced user, please bear with me on this point. I get daily e-mails from folks who have purchased Fritz6 as their first chess program and are totally lost as to some of the basic functions (which is natural, by the way. I bought my first PC and chess program in 1992 and it took me a while to figure everything out. Nobody sprang from the womb possessing expert knowledge of how to use chess software; we were all chess software beginners at one point).

For information on how to install and set up your Fritz6, please see ETN for December 12, 1999.

We'll start by looking at Fritz' playing levels. The first thing you'll need to know to understand some of these levels is the meaning of the term ply. A ply is a half-move in chess. When you read a chess book, you'll see (for example) a game in which Capablanca drew Nimzovich in 23 moves. This actually refers to move pairs: White makes a move, then Black makes a move. In chess books, this is shown as one move. In computer chess terms, though, this is actually two plies. Each time a player makes a move and the board position changes, this is one ply. For example, after 1.e4, one ply (or half-move) has been played. After 1.e4 e5, two plies have been played. After 1.e4 e5 2.Nf3 Nc6 3.Bb5 a6, six plies have been played.

This is important in computer chess, because when a program shows you its search depth (how far it's looked ahead when considering a move), the number is given in plies, not moves (move pairs). If the program has searched five plies after you make a move, it's considered:

  1. Its reply to your move;
  2. Your possible replies to its first reply;
  3. Its possible replies to your possible second moves;
  4. Your possible replies to its possible second moves;
  5. Its possible replies to your third moves.

This list of moves and replies and counter-replies, etc., is known as the move tree. It keeps branching out with more and more and more replies as the search goes deeper. It's called a move "tree" because of you made a chart of all of these moves and stood it up on its side, it would resemble a tree. The trunk would be the move you just played. There would be just a few branches near the bottom of the tree, but the branches get much more numerous as you travel up the tree (that is, as the move search deepens).

The reason this is significant to you as a player is because the deeper a program searches, the more it sees and the stronger it plays. If you're a beginner, you don't want the program to play at its top level, in which it looks very deeply ahead (unless, of course, you're a masochist and you enjoy being destroyed repeatedly).

Fritz6 gives you a lot of different options for setting its playing strength. You want to set an option in which you get a good game but don't get slaughtered everytime you play.

Before you start playing, make sure you have an opening book loaded. We'll look at these in greater detail in a future article; for now, all you need to know is that Fritz uses the opening book to vary its play. If you don't have an opening book loaded, the program will tend to play the same things over and over.

You'll also need to set a level. Once the level is set, hit "New Game". Fritz gives you a large number of choices here and it's tough to know what to choose. We'll take a look at each of these and help you make a decision.

There are two "casual" levels in Fritz: fixed depth and fixed time. You access these by going to the Game menu, selecting "Levels", and choosing either "Fixed time" or "Fixed depth" from the submenu.

"Fixed depth" is exactly what the name implies. You type in a number of plies and the program will move after that ply depth has been reached. This is a decent way to handicap the computer and weaken its play; the lower the number you type, the weaker the program will play, because you're limiting how far ahead it can "see". If you set fixed depth to "3", for example, the program considers its reply to your move, your replies to its reply, and its replies to your possible next moves.

However, as with all things in life, there's a catch. Using "Fixed depth", you're limiting Fritz' brute-force search. This is the type of search where it looks at all of your possible replies, considers all of its own, etc., and (in the example above) cuts off after three plies. But Fritz is still using something called selective search. In the case of forcing moves (such as checks, captures, and vicious tactical shots), the search goes deeper, past the three ply depth you've specified. This keeps the program from being blind to tactics, King hunts, perpetual checks, etc. Fritz shows you in the analysis window how deeply it's searching; the entry reads "depth=" followed by some numbers. The number to the left of the slash is the brute force search depth (which will be equal to the number you designated). The number to the right of the slash is the selective search depth (in which it's considering the consequences of checks and captures). In the example we're using, the entry will typically read "depth=3/9". It's cutting off after three plies in its normal search, but going nine plies deep in forcing lines.

When you set "Fixed depth", it's best to use an odd number. This insures that a ply in which it's Fritz' move will always be the last one considered; the program won't make glaring material-losing errors that way. Some programs are very weak at even-ply search depths.

When using "Fixed depth", Fritz will play the strongest moves it can within the limits you've imposed upon its search. Naturally, the deeper the search, the more it will see, and the stronger it will play. Don't go nuts with high numbers here; with present-day hardware technology, the typical chessplaying program will only do a 15 or 16 ply search maximum on a typical middlegame position (and that's if you expect it to use a generous amount of time per move -- say twenty or twenty-five minutes). If you want to complete a game in one sitting, I would set it for a maximum of 9, 11, or 13 plies depending on your hardware.

"Fixed time" is another way to limit Fritz' strength. Instead of limiting the program to a set number of plies per move, you're limiting Fritz to an average number of seconds per move. The dialog box allows you to type in a figure corresponding to the number of seconds. This also limits Fritz' thinking (just as in "Fixed depth"), but "Fixed time" allows Fritz to be a bit more flexible. Fritz will see more in positions where there is a small number of moves per side than it will in positions where each player has a lot of choices.

Here's how it works. Let's say you give Fritz ten seconds per move. This does not mean that Fritz will instantly move once its search has lasted ten seconds. It means that when the ten second mark has been reached, Fritz will finish analyzing the ply it's currently working on before it makes its move. There are no hard-and-fast figures I can toss out to you here -- a lot of this is dependent on your computer's processor speed. But the rule of thumb is that when the time limit is reached, Fritz finishes the ply it's on before moving.

As an example, let's say that the initial position has a lot of possible moves for Fritz (let's say 35 legal moves). By the time ten seconds rolls around Fritz likely won't be too deep into its search, but it's looking at a lot of positions per ply. It might take 30 or 45 seconds to make its move. On the other hand, if your last move checked Fritz' King, that limits Fritz' replies to maybe three or four legal moves. The program will rev up to a pretty good search depth, but will likely move in ten or twelve seconds since its move choices are pretty limited.

So keep this in mind: "Fixed time" and "Fixed depth" are somewhat similar in approach, but the effects are different. "Fixed depth" cuts the search off at a specified depth no matter what, while "Fixed time" will cause fluctuations in the search depth depending on the legal number of moves on the board.

The effects of "Fixed time" are very hardware dependent: setting it for ten seconds on a Pentium III 450 will result in stronger play than on a Pentium 133. "Fixed depth" is a constant, though -- a seven ply search is a seven ply search regardless of hardware, though the more powerful computer will complete the search and make a move more quickly than the weaker one.

Note, too, that although you're constraining the computer by using "Fixed time" and "Fixed depth", you're operating under no such constraints. You might have your machine set for a fifteen second average search, but you can take an hour per move if you like. This weakens the program's play in a relative sense. Years ago, I talked to a strong club-level player who was beating his program game after game and was complaining that his program "wasn't very strong". It turns out that he had his program set to 5 seconds per move on his 286 computer, while he was taking up to fifteen or twenty minutes to make his own moves. I told him to crank that thing up to three minutes a move or play it in a timed mode (with equal time for both him and the computer) and to get back to me if he still won consistently. I never heard from him again about this subject.

Another way to tweak the playing strength of the program is to use "Handicap and fun" mode. This is probably the best level for kids and beginners to use.There are nine different preset "personalities" that you can play against, plus a number of sliders and check boxes to enable you to alter the program's style of play.

The most significant of these sliders is "Playing strength". This corresponds to the approximate FIDE rating at which you want the program to play. The slider will show a range of ratings, lowest at the left and highest at the right. Moving the slider will display the current approximate rating above it.

There are two things to keep in mind when using this slider. First, the ratings are FIDE approximations, not USCF ratings. USCF ratings tend to be inflated by about 100 points in comparison with FIDE ratings. If you're a USCF rated player and you set the slider to your rating, you'll be playing an opponent who is about 100 points stronger than yourself.

The second thing to remember is that loading a different engine will change the range of ratings displayed on the slider. This range is an approximation based on the computer hardware you're using. Here's a sample list of the differences, taken from a Pentium II 266 MHz machine with 32 Mb RAM:

Thus to get a game at a pretty weak level, use a weaker engine (in fact, this is precisely why engines like ExChess2.51 are included in the package).

Most of the other sliders in "Handicap and fun" mode are fairly self-explanatory, but here are some brief descriptions:

Obviously, tweaking one of these parameters may have an effect on the other parameters (sometimes in unexpected ways). This is not an exact science. Just have fun with it. To get some ideas, click on the preset personalities and watch what the sliders do. To set the sliders back to the original levels, click the "Default" button.

Two others of the untimed levels are extremely valuable to the developing chessplayer. These are "Sparring" mode and "Friend" mode.

One of the best (and fastest) ways to improve your game is to study tactics. Learning to recognize the basic patterns and then spotting them in your own games is pretty much a sure path to improvement. But one of the major pains of using a chess computer for training purposes has always been the lack of tactical oppotunities. You can study tactics for hours a day, sit down to play a chess computer, and never get the chance to spring a tactical shot. This is because a computer can be dang near omniscient to average players like you and I. It sees all and knows all, often long before you do. A computer will spot that a certain line of play will allow you the chance for a Knight fork of its Queen and Rook two moves from now and immediately decide to stay away from that line of play. You never get close enough to the fork potential to even sense it -- the computer spotted it already and won't let you anywhere near that position.

"Sparring" mode makes the engine work in the opposite manner that we've come to expect from chess computers. In this mode, when the computer spots your potential Knight fork two moves down the road, it will actually steer toward it rather than away from it, giving you the chance to practice your tactical knowledge. It looks for tactical ideas for you to use against it and sets these up for you.

There are five levels to "Sparring" mode, ranging from easy (one move forks or skewers) to really hard (long combinations). Checking the "Point out wins" box means that Fritz will signal you (with a red light near the board) when the combination is there on the board, telling you that the time to strike is now! This is a pretty radical departure from the behavior of your average chess program. This feature was in the works for over fours years before it was introduced in Fritz5. I think this is one of the greatest features ever in a chess program and it goes a long way toward alleviating the frustration many people have felt down through the years when playing silicon opponents. We humans finally have the chance to practice tactics against our computers!

In "Friend" mode, Fritz becomes an "adaptive opponent" which is a grandiose way of saying that it will attempt to match your level of play. The idea is to provide you with an opponent of slightly better playing strength, so that you're challenged instead of slaughtered. Obviously, it won't do this the first time you play it in this mode. But over a series of games, Fritz will adjust its strength upward and downward to find a level at which you win about one out of three or four games against it.

When you enter "Friend" mode, a dialog box appears which shows how many games you've played in this mode, the percentage of games you've won, and a material handicap. You'll see this handicap change from game to game. The handicap is expressed in 1/100th pawn increments. The better a player you are, the lower the handicap will be. A total beginner might see a handicap as high as 900, meaning that Fritz needs to spot him a Queen to be able to give the player an even game. A grandmaster will typically see a negative number as his handicap (Fritz can't afford to give up any material against such a player). The average club player will likely see a handicap between 150 and 300 (1.5 pawns to a minor piece).

The idea of "odds" games was discussed previously in Electronic T-Notes (August 15, 1999) and there's a paragraph about "Friend" mode near the end of the article. Briefly, odds games in the 19th century were a popular way to give a weaker opponent a fighting chance. These games have fallen out of favor in recent years; in fact, many players feel humiliated by the mere offer of an odds game.

"Friend" mode is a sneaky way for Fritz to offer you an odds game. If your handicap is 300, for example, this means that Fritz will attempt to lose the equivalent of three pawns (ex. a minor piece, three pawns, or two pawns and a few tempi) early in the game. The program will then play at full strength once the "odds" have been given. This is why you may have noticed Fritz playing like a fish in the early middlegame but fighting like a tiger in the later stages of the game when you have it set for "Friend" mode. It didn't take a Bishop off the board and play at full strength from the game's start; it hung the Bishop early on and then played full-tilt once the game was imbalanced in your favor (this is also how it plays in "Handicap and fun" mode when you set the slider to a low rating level).

In "Friend" mode, Fritz will keep track of the number of games you've played and your winning percentage (which will likely stay in the 25% to 35% range after a few games) and you'll see your handicap rise and fall depending on how you do against the program. "Friend" mode is my preferred way of playing games against Fritz; I get a nice competitive game without getting my head torn off. I can also gauge my progress as a player (or lack of same) by watching my handicap value go up and down from game to game.

For masochists who actually prefer being slaughtered time and again by their computers, there are timed play modes in Fritz. There's also a rated game mode for players who want to be tested under serious "tournament" conditions. We'll look at these playing levels 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.