bookup.com

For decades Chess Openings Wizard has displayed the "best line."

That means the second best line often starts with the same move.

That's all about to change as COW will begin to mimic what the chess engines do in their multiPV modes. Alternative variations (if available) always begin with a different move.

Both approaches have their advantages and drawbacks. Often the second best move is ridiculous, with forced recaptures being a primary example.

Why the switch?

I'm working on a more sophisticated way to expand an opening repertoire automatically. It's based on the Theory Machine approach that has been at the heart of developing repertoires in COW since engines were first added. It was born out of ideas from Ron Taylor who pointed out that Backsolving has become obsolete with the arrival of nearly instant high quality analysis. The project is codenamed "Epic."

The idea of the Epic Theory Machine is having the engine be allowed to visit each position in the opening repertoire and then add moves that should be considered. This approach can (should?) include moves that are seen by the engines as slightly worse than the best move. Hence the need for second and third best moves.

In the next few weeks I will be rolling out updates to COW Pro for Windows and Macintosh followed by Express updates.

I'm also testing a cloud based chess engine server for the mobile (iPhone, iPad, Android) versions of COW. This server will also support multiPV the way the engines do it.

Good idea? Bad idea? Something I may have missed? Let me know.

- Mike

Why would you want to allow the engine to add slightly worse moves to your opening book?

The short answer is that they are playable and therefore you must prepare for them. There are players who make a practice of using engines to find offbeat moves that are not the best - but are reasonably sound as far as the engines are concerned. To defend against this kind of preparation you must have the engine point out moves to you that are not the best.

Are you planning to remove Backsolving?

I remember Zarkov, but could not find my copy of it.

I did find Chess Genius and Fritz 2.

I Think this a good idea, in principle. Sorry for appearing possibly stupid here, but will we be "told" by the analysis which are the added variations?

C-Level wrote: ↑Sun Apr 07, 2024 11:26 am Are you planning to remove Backsolving?

I remember Zarkov, but could not find my copy of it.

I think I still have a Zarkov manual in my boxes somewhere.

No, Backsolving remains handy in some circumstances. It's not going anywhere. The EPD analysis that supports parts of Backsolving is impacted by the engine changes but in a good way.

Marc Wakeham wrote: ↑Sun Apr 07, 2024 12:01 pm I Think this a good idea, in principle. Sorry for appearing possibly stupid here, but will we be "told" by the analysis which are the added variations?

Yes, of course. How that will be done is still in the design phase. The newly added candidates could be color coded for example. The real trick is how to consume the changes. I'm currently thinking of adding a collection of bookmarks leading to the positions where the engine has added a new candidate. You would cycle through those newly added bookmarks to see the new ideas.

It sounds interesting. I look forward to trying it.

MikeAtBookup wrote: ↑Sun Apr 07, 2024 12:37 pm

Marc Wakeham wrote: ↑Sun Apr 07, 2024 12:01 pm I Think this a good idea, in principle. Sorry for appearing possibly stupid here, but will we be "told" by the analysis which are the added variations?

Yes, of course. How that will be done is still in the design phase. The newly added candidates could be color coded for example. The real trick is how to consume the changes. I'm currently thinking of adding a collection of bookmarks leading to the positions where the engine has added a new candidate. You would cycle through those newly added bookmarks to see the new ideas.

Excellent. Colour coded is cool!

This sounds like a great addition. Looking forwqard to it..

MikeAtBookup wrote: ↑Sun Apr 07, 2024 9:29 am I'm working on a more sophisticated way to expand an opening repertoire automatically. It's based on the Theory Machine approach that has been at the heart of developing repertoires in COW since engines were first added. It was born out of ideas from Ron Taylor who pointed out that Backsolving has become obsolete with the arrival of nearly instant high quality analysis. The project is codenamed "Epic."

This is simply fantastic.
I believe that backsolving is not obsolete, but in practice never worked unless all the holes in tree were filled.
And finding and filling them manually is really painful, and to be repeated periodically.
Aquarium and BanksiaGUI made attempts at automatically growing a tree, but an hybrid mode like this would be way better.
I believe that a filter should be added to never add moves, to add only better moves, or moves within a max range, for both sides.
This would be a good complement to multiPV.
This would be useful also importing updates from pgn (like the new TWIC).
What about WDL?