ELECTRONIC T-NOTES


CHESSBASE USA'S WEEKLY ON-LINE NEWSLETTER


FOR THE WEEK OF MARCH 12, 2000


A USER'S GUIDE TO FRITZ6 -- PART 10 (ANALYSIS I)

by Steve Lopez

Many moons ago, during the "Fischer era", I was happy to discover that my town had a weekly chess club down at the YMCA. I packed my set under my arm and talked my dad into driving me down to the club. When I arrived, I saw about two dozen middle-aged to elderly men hunched over tables and pushing pawns. It didn't take long for me to find a game. I was demolished fairly quickly. I remember asking my opponent his opinion of what I'd done wrong; his reply was a curt (actually, it was more like rude) "I don't know" as he went off to find another game.

Making a long story short, this scene was repeated many times that night and over the next few weeks. My opponents would wipe me out and then refuse to give me any feedback as to how to improve my game. I studied chess books at home, went back to the club each week with hopes of at least drawing a game, but got killed each time. In fact, after a couple of weeks I discovered that many of the club's members wouldn't play me anymore; I suppose I wasn't worth their time. After about six weeks of many losses and no help from my opponents as to ways to improve, I stopped going to the club. It wasn't worth my time. What I discovered the hard way is that it's tough to improve at chess when you don't know what it is you're doing wrong.

You can study all the chess books in the world and study games from databases (and these pursuits will help you improve your game, believe me), but nothing helps you improve more quickly than personalized criticism. Self-analysis is very difficult for most people (in life, too, as well as in chess). Many of us don't have a local master or GM who will give lessons. Our best bet in this case is to use chessplaying software to steer us in the right direction for improvement. This is possibly the most important function of chessplaying software. And this function is what we'll examine in the next few issues of ETN.

Last week we looked at the methods for saving games into a database (both games you've played against the computer and games you've played against human opponents). The first step in analyzing a game is to choose a game you want the program to look at. While it's loads of fun to relish our victories, these aren't usually the best games for analysis. Obviously the best way to discover our weaknesses is to have Fritz analyze our defeats and point out our mistakes.

There are two methods for starting the analysis process in Fritz6. The first is to load a game by double-clicking on it in the game list. This kicks you back to the main screen of the program, with your game loaded (the moves are displayed in the notation pane). You then go to the Tools menu, select Analysis, and then the analysis method you want the program to use. However, I don't recommend you start the analysis process this way, for two reasons:

Consequently, here's the method I prefer to use when starting the analysis process (and the method I encourage you to use):

  1. In the main chessboard screen of Fritz6, click on the button to go to the database window;
  2. Switch to the database into which you've stored the game you want to analyze (see last week's ETN);
  3. Click once on the game you wish to have Fritz analyze -- this will put the grey cursor bar over the game's listing;
  4. Go to the Tools menu, select "Analysis", and then the specific analysis type from the submenu.

In this issue of ETN, we're going to examine "Full analysis". This is an especially good analysis mode for beginners, as Fritz will provide some text commentary as part of its analysis. A more advanced (and, in my opinion, more precise) analysis method is "Blundercheck", which we'll examine next week in ETN. If you want to get a jump on next week's class work, check out the ETN issue for November 21, 1999, in which I extolled the virtues of Fritz5.32's Blundercheck feature.

After selecting "Full analysis", you'll get the following dialogue box:

This is the box in which you'll set the analysis parameters.

The first thing we'll consider is "Calc. time". This refers to the average number of seconds Fritz will spend in analyzing each move. This has various effects similar to those encountered when playing a game using "Fixed time" (please see ETN for January 2, 2000 for more on this). For example, if you set the calculation time to 30 seconds per move, when Fritz hits the 30 second mark it will finish the ply it's currently analyzing before moving on to the next move of the game, no matter how long this takes. The ply depth it reaches in a given time is extremely hardware-dependent; it will take some experimentation on your part to discover what calculation time provides you with good analysis without tying your computer up for an extended period.

Needless to say, the longer a calculation time you give Fritz, the better analysis you'll receive. But there is a law of diminishing returns here. As we've already seen in earlier articles in this series computer software tends to hit a "wall" around ply 15 in most middlegame positions in which there are a lot of legal moves. Around that point there are so many positions that a computer has to consider to complete the ply that it's just not worth the extra time required to squeeze out that extra half-move. As an example, on one of my computers (a Pentium II with 32 Mb RAM) it takes Fritz about 20-25 minutes to reach a 15-16 ply search depth. This is just too much time, considering that it reaches 13 plies in under 10 minutes. It's not worth it to me to make the program work another 15 minutes just to get another full move out of it.

In a typical middlegame position with around 30 possible moves, you'd like to see Fritz give an eleven to thirteen ply analysis. This gives you good analysis without tying your computer up for twelve hours or more.

It's also a very good idea to have Fritz analyze a game overnight while you sleep. Even though Fritz is a Windows application (and, in theory, allows multitasking), Fritz is extremely resource-consumptive. You don't want to be running other programs while Fritz is analyzing (in fact, it's a good idea to disable any TSRs you have running in the background before starting an analysis session with Fritz). Let the program analyze your games overnight when you won't be needing your computer for anything else.

This brings us back to the sticky question: what setting should you use for "calc. time"? Again, this depends primarily on the speed of your machine. As a general rule of thumb, anything under 30 seconds is too brief and provides superficial analysis. The lowest I go on a Pentium I with 64 Mb RAM is 45 seconds (and that's if I'm in a hurry). I typically use no less than 60 seconds and usually go for three minutes (180 seconds) per move. Again, depending on your hardware, your mileage may vary. If you have a fast PIII computer, 180 seconds per move might tie your computer up for many hours (because of the "wall" effect I described earlier). Play around with this setting and see what works best on your machine.

The next setting is "Threshold". This is a value given in 1/100ths of a pawn. The default value is 30. In "Full analysis" mode, Fritz doesn't display its numerical evaluations, but it still uses them internally (obviously). As an example, let's say that you use the default value (30). Fritz analyzes a move. If its evaluation of the best line of play in that position shows that the best move results in an improvement of less than 30/100ths of a pawn over what was actually played in the game, it won't display its analysis of that move to you. If the best line it discovers is more than 30/100ths of a pawn better than the actual game move, it will show you that variation in its analysis.

This allows you to tweak the program's output to your taste. If you set "threshold" to 100, you won't see anything but improvements that are better by a pawn or more. This results in purely tactical analysis being presented to you (you'll only see the points in the game where you blew it by losing a pawn or more, with Fritz showing what you should have played instead). This is a pretty good setting for beginners (who aren't ready to start considering subtle positional points and are still overlooking purely tactical shots or are simply hanging pieces). If you knock it down to 15, you'll get a lot more analysis in Fritz' end product, most of which will be gradual positional improvements. The reason why 30 is the default is that gaining or losing a tempo is considered to be worth about 30/100ths of a pawn (and a special "thank you" goes out to Bob Pawlak for teaching me that point).

Here again, people always ask me "What do you use as a setting?" In most cases, that depends on the character of the game being analyzed. If it's a very tactical game, I set it for 50. If it's very positional, I set it for 25. If I can't make up my mind, I set it for 30. I'm not saying that this is absolutely etched in granite -- I'm just offering some personal guidelines.

"Last move" is not available when you access "Full analysis" through the database window. You do get this option when starting it from the main chessboard screen. I never use this option myself, so I'll bypass it in this article.

The next area to consider is "Storage". This tells Fritz what to do with the game when it's finished analyzing it. "Replace" means that Fritz will replace the original game in the database with its annotated version. "Append" means that Fritz will add the annotated game as a new game at the end of the database, while leaving the original version alone. This is why I suggested accessing "Full analysis" from the database window instead of the main chessboard window -- if you do it from the main screen, the "storage" options aren't available to you and you'll need to save/replace the game manually.

Please note that if you have Fritz analyze a game that's already annotated, it removes the original annotations. So, for example, if you want to have Fritz analyze an already-annotated game from the database that comes with the program, you'll likely want to use "Append" as the storage option -- if you don't, the original annotations are gone.

Note, too, that the "Storage" options will not be available if the game you're analyzing is from the database on the CD (the analysis information can't be written to the CD, since it's a permanent medium).

The next area is "Annotations". Fritz offers four special annotation types. "Verbose" means that it will provide some commentary in plain text. It's a neat feature -- if you want to see some plain text, make sure you have this box checked. Admittedly, some of the commentary can be repetitive (especially the "menagerie" comments, such as "Doesn't get the bull off the ice" and "Doesn't get the cat out of the tree"). You'll also notice that once a game is dead lost for one side but goes on for many more moves, Fritz continues to provide some "better" lines of play with text caveats that they still don't help anyway. Lost is lost.

The "Graphical" box means that Fritz will add colored squares and arrows to the board at relevant points in the game. You'll see this frequently when an isolated pawn becomes the object of attack and defense. "Training" allows Fritz to add timed training questions to the game; when you replay the game, a timed "test" will appear at certain points in the game. Fritz is using these to quiz you on tactical motifs. Note, however, that you don't always receive graphical and training commentary. I'll typically get training questions once in every ten or twelve games that Fritz analyzes. Also note that setting a longer calculation time increases your chances of seeing these types of annotations.

"Opening reference" is a special case. It's normally in half-tone (greyed out). This is because you need to specify an opening database in order for this function to become available. Click the "Reference-DB" button. You'll get the standard Windows file select dialogue which will let you choose a database as a reference database. This can be any database of your choosing. In most cases, you'll use the database that came with Fritz. Advanced users (particularly ChessBase 7 users) will find that one gets even better results from this feature if one specifies a keyed database containing nothing but games on the specific opening in question.

Once a reference database is selected, the "Opening reference" option becomes available. Put a check in this box and Fritz will search the database for other games using the same opening as the game it's analyzing. It'll drop references to a couple of these games as variations into the analyzed games (just like the ones you see in a lot of chess books and magazines).

Finally, the "Side" box allows you options for limiting Fritz' displayed analysis to one player's moves if you wish. The options are self-explanatory. I prefer to select "both". I want to see ways in which my opponent could have improved his game -- this shows me things I missed while I was playing the game.

Once you've finished selecting your analysis options, click "OK" and the program will kick you back to the main chessboard screen. If you've selected the "Opening reference" option, Fritz will immediately drop in some references from other games. After a few moments, Fritz will then jump to the final position of the game and begin analyzing it. It works from back to front when it analyzes a game -- anyone who has annotated a chess game knows that it's easier to comment on moves if you know what's going to happen. Fritz is no different; it stores analysis of later moves in memory to speed up the process of analyzing earlier moves.

If you like, you can watch Fritz analyze for a while; it shows you what it's presently thinking in the analysis pane. I typically have Fritz analyze games overnight. If you have an older monitor, you might want to turn the monitor off during the night to prevent "burn-in". However, newer monitors (from the last five years or so) shouldn't have this problem.

How do you know if Fritz is finished analyzing when you come back in the morning? When Fritz is finished, it'll kick the display back over to the database window. Look at the line for the game you had Fritz analyze -- you'll see some letters (corresponding to the analysis types it generated) in the far right-hand column. For an explanation of these abbreviations, see ETN for January 11, 1998.

Double-click on the game to load it. If any training questions were created, you'll be asked if you want to do the training or ignore it. Your game will now contain variations and commentary on better moves the players could have made.

As you play through your game, you'll see the variations Fritz has added. Many of these end in special evaluation symbols. You'll also sometimes see these symbols added to moves that were actually played in the game. These symbols are used to tell you who's winning (or at least has an edge) in that position. Below are some of the symbols and what they mean:

You can play through the game and when you come to a point at which Fritz offers an improvement, you can choose that move and play through the variation to see what might have happened with better play (and usually see an evaluation symbol at the end of the variation). This is exactly the kind of feedback I was looking for (and not receiving) when I was going to the local chess club all those years ago. If only I'd had a computer and Fritz back in those days I'd be a much stronger player today. [Insert deep sigh here]

In this article we've seen how to activate "Full analysis". Next week we'll look at "Blundercheck". After that, we'll examine some of the nuts and bolts of what this analysis stuff actually means and how you can use it to improve your game. Until next week, be having fun please!

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.