ELECTRONIC T-NOTES


CHESSBASE USA'S WEEKLY ON-LINE NEWSLETTER


FOR THE WEEK OF DECEMBER 9, 2001


NEW FRITZ7 OFFLINE FEATURES

by Steve Lopez

Now that Fritz7 has been installed (see ETN Dec. 2, 2001), we're ready to check out the program's new features. These features consist of two types -- the online features (that allow you to play chess online in realtime at ChessBase's new chess server) and the offline features (which are self-contained within the program). This week we'll have an overview of the latter set of features (many of which will be explained in further detail in later ETN issues).

Obviously, the big attraction is the new engine. To describe what makes this engine different requires a bit of background explanation, so please bear with me while I dredge up a little bit of computer chess history.

Many years ago, chess programmers thought that the best way to get strong play from a computer engine was to make it a very fast searcher. The engines didn't "understand" much about chess strategy, but derived their strength from the fact that they could see very far ahead -- i.e. they didn't possess much chess "knowledge" but were strong because they could see nearly everything out to ten ply depths or greater. The Fritz5.32 engine was the apex of this technique; it's probably the fastest searcher of any commercial PC chess program. These engines are known as "fast but dumb" -- they see a lot but don't really understand what they see. They're strong because they can see the consequences of their actions, not because they understand why an initial move is good or bad before analyzing said consequences.

But there was a hidden trap here, one that the programmers couldn't anticipate. As has been described numerous times in earlier ETN issues, every time you add a ply to the search depth, you increase exponentially the number of positions that an engine must evaluate before finishing that ply. Even though PC processors are getting faster and faster, they're still not powerful enough to analyze a typical chess middlegame position out past sixteen or seventeen plies in a reasonable amount of time. So engines seemed to be hitting a "wall" at that point, even though PC processors were increasing in speed annually.

To get around this "wall", programmers needed to take another tack. Instead of relying on brute force to make their engines stronger, they decided to add more chess "knowledge" to make their programs stronger. Some programmers (like Mark Uniacke, who writes the Hiarcs series of engines) had been doing this all along, creating what are known as "slow but smart" programs. These programs don't see nearly as far ahead as the super-fast engines, but are much more adept at picking out strong moves early in the search tree by relying on a store of chess knowledge built into the engine. They don't see as far ahead, but they don't need to -- they're good at picking out strong moves early in the search without relying on seeing the consequences.

Starting with Fritz6, Franz Morsch began to program such knowledge into the Fritz program. This is why Fritz6 analyzed much more slowly (in terms of positions per second) than did its predecessors but was a much stronger chessplaying engine. Deep Fritz continued the process by adding more of these chess rules of thumb (known as "heuristics"), and Fritz7 takes the process to the next level by further increasing the amount of chess knowledge available to the program.

This, too, is why some users have already noticed that Fritz7 analyzes more slowly than Fritz6. A set of heuristics is like a checklist that an engine must consult to see if any of the rules of thumb apply to the position currently being analyzed; the longer the "checklist", the longer the process takes. Consequently, the program doesn't analyze as many positions per second but "understands" much more of what it sees, making it play a better game. Fritz7 has a more extensive set of heuristics than Fritz6, so it analyzes more slowly but will also play better chess.

How much better? That's a lovely little can of worms to crack open and arguments abound. If you keep up with the Interrant chess message boards, you'll see a variety of opinions. While it's generally agreed that Fritz7 is stronger than 6, users disagree on the exact Elo difference. I've seen numbers ranging from 20 points to 75 points. While this doesn't seem to be a huge jump, keep in mind that ratings are calculated differently at the master and IM level -- it's harder to get a large jump in one's rating compared to what occurs at the class level. For example, a Class C player can win a four round event and get as much as 128 points added to his rating (depending on the level of his opposition), while a GM can sweep Linares and only see a two to six point jump in his Elo rating. So a twenty point increase in strength at Fritz' level is nothing to sneer at. I've been in the computer chess business for a long time and hard experience has taught me to be pretty conservative in my estimates, so I'm inclined to believe that the strength increase is in the 20 to 40 point range, rather than the higher 75 point figure I've seen claimed by a few people. Keep in mind, too, that I'm not strong enough as a player to make that determination from personal experience and that hardware does still make some difference (despite the fact that programmers are developing engines that are somewhat less hardware-dependent than in earlier years). But the early reports indicate that Fritz7 is stronger than 6 and, with my limited "insider's" knowledge of what's been going on with the engine's development, I'm inclined to believe it.

Fritz7 actually comes with two engines, by the way. There's a version optimized for the newer MMX processors, as well as a "non-MMX" version. If you're using the engines inside the Fritz7 interface, the interface chooses which engine to use by default (depending on your hardware), so you'll only see one Fritz7 engine listed when you hit F3 to get the engine list. But ChessBase 8 users will see both engines available in CB's engine list -- you can use either or both for analysis.

The program also contains many new "ease of use" features. While the user interface is based on that of Fritz6, there have been several new features added. Two of these involve keyboard shortcuts for activating various other program features. If you go to the View menu and select "Shortcuts", you'll get a scrolling list of the keyboard shortcuts available in the program:

This is a handy way to find a particular shortcut you may have forgotten, as well as a means of discovering shortcuts you weren't aware existed. This is a new feature suggested by many users over the last couple of years.

But what if you want to change a particular shortcut? Say (for whatever reason) that you don't like using CTRL-F to flip the board (so that Black plays from the bottom) and you'd like to make F5 the default key for doing this. Go to the Tools menu and select "Customize":


This brings up a dialogue displaying the various shortcuts that can be customized to better suit your needs. In this case, you'd scroll down the list to the entry reading "Flip board - Rotate the chess board". Click on it to highlight it and you'll see that the current hotkey shortcut is CTRL-F. Click in the "Input new shortcut" field to get a cursor, then hit F5. If the key (or key combination) is currently being used for another function, the dialogue will tell you so (under "Currently assigned to:"). In this case, F5 isn't being used for anything else. So you'd just click the "Assign" button, then click "Apply". From now on, the board will be flipped whenever you hit F5.

Another handy new feature concerns Shuffle Chess (also known as FischerRandom), in which the pieces are set up randomly on the back rank to start a game (eliminating the benefits of any pre-memorized opening theory). Fritz7 supports Shuffle Chess. Just got to the File menu, select "New" and then "Shuffle Chess". You'll see a dialogue that allows you to accept the Shuffle Chess position number the program has randomly selected or else you can type in your own position number. Click "OK" and let the fun begin.

If you go back to ETN for July 2, 2000, you'll recall the procedure for setting up endgame tablebases on your computer: copy the files to a folder on your hard drive and then manually edit the chssbase.ini file to tell Fritz where to find them. There was also a special tweak in that article for users who could get only the first three CDs on their hard drive but not the fourth. A new option in Fritz7 allows you to give the program this information from within the interface itself. Go to the Tools menu, select "Options" and then click on the "Tablebases" tab to get the following dialogue:


This dialogue allows you to specify up to three separate locations for your tablebase files (for example, you may have a partitioned hard drive without enough space for the whole set on any one drive. Using this dialogue, you could conceivably put the tablebases in folders on three different drives and then show the program the three locations).

Needless to say, this is considerably easier than editing the .ini file (especially for new computer users who had trouble even finding the proper file, much less editing it). And when you select the tablebase location in this dialogue, the setting will be recognized by all ChessBase interfaces (not just Fritz7), so your other playing programs and ChessBase 8 will also recognize the location(s) of the tablebases.

The database window has a few new features as well -- you can now assign any existing opening key to a database within Fritz, plus reclassify the keys at will. These features were formerly exclusive to ChessBase and will be covered in a future ETN issue.

Another cool new feature is the analysis board. This is not the analysis board that was available in the Engine pane in Fritz6 (although that one's still available). The new analysis board is accessed through the Window menu of the main chessboard screen -- go to "Panes" and select "Analysis board":

The analysis pane appears below (and slightly smaller than) the main gameboard. The analysis pane allows you to move pieces around (to help you analyze positions) without disturbing the actual game. More information on this feature will also be forthcoming in a future ETN.

And the peripherals have been beefed up as well in Fritz7. The program comes with a new opening book, optimized to take advantage of Fritz7's own style of play. Plus the size of the database that accompanies the program has been increased -- the total number of games now stands at 357,908, with games dated through 2001 being included.

There are other tweaks added to the program as well -- chief among these being the ability to play online chess on ChessBase GmbH's new server. That's a whole new bag of features and we'll look at those in later ETN issues.

Until next week, have fun!

You can e-mail me with your comments, suggestions, and analysis for Electronic T-Notes.