Into the mailbag again

ive just read the article about tablebases for up to 6 pieces. i have a question concerning evaluation of a position. suppose the game gets to the endgame and the engine hops into its tablebases. now, if it "sees" all possibilities of positions and all the results, it plays perfectly. my question is, how does it evaluate the position? does it say something like (1-0, checkmate in 47 moves)? thats the good part, when the engine is winning. but suppose it is losing.

how does it go about deciding which moves to play if it "sees" that no counterplay can reach a draw if the oppenent plays it right? does it give up? playing a perfect endgame against a human player i would not give up even if i saw a checkmate in 20 for my opponent, because humans are not perfect in endgames.

thanks,
erik markus

Thanks for the questions, Erik.

The engine, if it's winning, will display something like "#47", meaning mate in 47, followed by the first move of the variation which leads to mate.

When it's losing, my experience has been that the engine makes the move which causes the longest mate in x value; in other words, it'll delay the mate as long as possible if there's no chance for a draw. Of course, if you have the "Resign" value in your program options set for "Early", the engine would likely give up at the point at which no draw is possible. -- SL

As I buy more and more software from ChessBase, at time I wonder if I have the software (ChessBase, Fritz9, Hiarcs,etc...) configured correctly. They ask for opening databases(what if I have multiple openings db's?), reference database (what if i have multiple db's?), etc.... I have end game databases. How do we add all these programs and have the programs in optimum configurations.

Robert Gonzales
Cypress, CA

The cool thing about the software is that you're not "locked" into any single configuration. Just use the opening book that's the most useful at the time (for example, if you're playing a game in the ChessProgram (i.e. Fritz) interface, just use the opening book for whichever engine you're currently playing against. As for databases, you'll usually want your largest database to be set as your reference database. However, I keep individual databases for each opening I play regularly (such as, say, the Ruy Lopez Worrall), so if I'm doing research work on the Worrall, I'll set that database as my Reference database to keep the searches shorter, save time, and have access to the additional notes and games I've added to that database.

So the short answer is that the optimum configuration is whatever is most pertinent and useful to you at that particular time; you can always change opening books or designate a different database as your Reference database later. -- SL

Tripled pawns were named the Irish Pawn Centre by Tony Miles when he saw a game won by the irish player Keogh. I believe that Smyslov once beat Botvinnik with a similar structure.

Hoskidi

I never knew it had a name! Cool! Thanks for the info. I didn't find either of the games you referenced, but here's one a fellow named Keogh lost with tripled pawns (which occur after 33...bxc3):

Burnett, Andrew - Keogh, Brian Grangemouth op, 1995

1. e4 e5 2. Nc3 Nf6 3. g3 Nc6 4. Bg2 Bc5 5. Nge2 d6 6. O-O Bg4 7. h3 Be6 8. Nd5 Ne7 9. Nxf6+ gxf6 10. d4 exd4 11. Nxd4 Qd7 12. Kh2 O-O-O 13. Be3 f5 14. exf5 Nxf5 15. Nxf5 Bxf5 16. Qf3 c6 17. Bxc5 dxc5 18. Rad1 Qe6 19. Rfe1 Qg6 20. h4 Rxd1 21. Rxd1 Rd8 22. Rxd8+ Kxd8 23. Qf4 Kc8 24. Qxf5+ Qxf5 25. Bh3 Qd7 26. Bxd7+ Kxd7 27. Kh3 b5 28. Kg4 Ke6 29. Kg5 a5 30. h5 c4 31. Kh6 b4 32. c3 a4 33. a3 bxc3 34. bxc3 Kd5 35. f3 1-0

Thanks again! -- SL

hi i install the new fritz9 on my pc and when i go to use the programm the engine don't load and the programm tell me that i must put the fritz cd in cd rom i put cd nothing happened what can i do?

Karim Monir

Make sure the CD drive letter displayed in that dialogue is the correct letter for the drive the CD is in. If not, type in the correct letter. Then click the "Check" button again. That should clear the dialogue and let you play. -- SL

I have heard it is possible to use the engine to analyze an opening book. I am just about to create a new opening book using Shredder and some other uci engines and I am eager to have it analyzed. Is this possible and how? I could not find any thing about it in the manual.

And, if I analyze my opening book, to what extent will Shredder take notice of it when it is using the book? Will the probability for playing the move rise if the position got a positive evaluation, or will it still just look at the statistics for victory or defeat for the whole game?

Magnar Totland

No you can't have an engine analyze a whole book, but this isn't really necessary -- the book is already chock full of statistical analysis based on the games the book was created from. The engine will decide on a move based on a combination of the statistical success percentage of the given moves in that position and any "weighting" figures which may be present (either because the user has tweaked them by hand or they've been changed by the program's previous success or failure in prior games when 'Book learning" was enabled). -- SL

Follow the yellow brick road and you reach --. If you're in Shredderville on the weg "10.ctg," the "!!'road leads one all over the place, e.g.:: 1.e4 e5 2.Nf3 d3 "!!" 3.d4 Nf6 "!!" 4.dxe5 Nxe4 5.Qd5 etc. For one thing I think we blew through a red light at the corner of 4.Nc3 and +/=.

ChessBase knows how the job is done. What is the explanation - what is the purpose?

Richard Runke

Despite the illegal 2...d3 you gave (I'm pretty sure you meant ...d6), I was able to follow the notation. The move 4.Nc3 is given in the opening book, so I assume that Shredder opted for the pawn capture in the game you played instead of the Knight move. The exclams designate a preferred move as set up by the opening book's creator; while the move is preferred, it's also not a guarantee that the program will play it, especially if the weighting factors in the book have been altered or the book option for a greater variety of play has been selected. I'm pretty sure that's what's happened here, since 4.Nc3 is also given the exclams, designating it as the preferred move. -- SL

Hi,can you please tell me how to uninstall the Nalimov tablebases?

Kevin Harrison

There are two options here:

1) In Windows, go to Start/Settings/Add-Remove Programs and see if you have a "Fritz Endgame Turbo" listing (I can't tell you for sure because I don't have the hard drive space to do a full install of the complete set of tablebases). If so, select this listing, click the "Add/Remove" button, and then follow the prompts;
2) Manually delete the .emd files from the tablebases folder using "My Computer" or "Windows Explorer".

That's all there is to it. -- SL

Regarding tablebases:

So when will we see 7 pieces and how big will that be?

What is a practical limit 8, 10, 12?

What kind of storage is needed for these?

And what is the point (10, 15 pieces)where access (a few seconds or more)time becomes significant?

Michael Roza

The crystal ball's in the shop for repairs, Michael, so anything I answer will be a guess or extrapolation. Taking your questions in order:

1) Considering that the full set of six piecers is still in the works, I imagine it'll be a while before we see seven piecers. But when/if they appear, they're gonna be huge. Considering that at least one of the six piece files won't even fit on a DVD, you can imagine how big the sevens will be. Then the question becomes: how will the end user obtain the files? Until DVDs are superceded by the new megadisk formats currently being developed, I think distribution is going to be a problem.

2) In theory, there's no limit to the number of pieces in a tablebase. In current practice, though, I think we've come pretty close to maxing it out until hardware (disk storage and processor speeds [for tablebase generation]) catches up a bit.

3) Once again, there's no way to tell with dead certainty (for me, anyway; I'm no mathematician), except that we're seeing a huge progression in the size of the files with each added piece. So you can extrapolate from the current sizes.

4) Define "significant". When just a second or two remains on the program's clock, current access times can be "significant". The definer here will be whether a tablebase is stored on a disk or in RAM; the latter is a lot quicker, but then there's the problem of how much RAM is physically available. And remember that the more RAM you allocate to tablebase caching, the less you have available for the engine's transposition tables. But the faster an engine can access the tablebases the better, all matters of relative "significance" aside.

Of course, the new "Shredderbase" format for tablebases changes the whole equation, because the new format is much more compact than the Nalimov version. The Processor That Ate Toledo will still be required to generate the things, but the storage becomes much easier -- especially for those of us who don't have multiple 100+ GB hard drives on our machines. And access times are much faster, too (more on this point next week).

In short, time will tell. For right now, I think anything higher than six-piecers are impractical for the average user's hardware. But that changes very quickly since new hardware is developed and released all the time; as I'm fond of saying, any hardware you buy is obsolete by the time you get it home and get it out of the box. So we'll see... -- SL

More questions are still stuffing the e-mail box, and we'll look at them a bit later. Until next week, have fun!

You can e-mail me with your comments on ChessBase Workshop. All responses will be read, and sending an e-mail to this address grants us permission to use it in a future column. No tech support questions, please.

© 2006, Steven A. Lopez. All rights reserved.