Delays

Seems like I'm going to need a few more days. I've manged to split the move generation so that I now incrementally generate moves and search them (as appose to generate all the moves at once). This seems to be paying off. However, I'm having some transposition table problems again!

I'm hoping for end January.

Resurrected? ...or just out of the coma?

I guess life happened while I was busy working on my chess engine. I few months ago, I dumped the entire engine and started fresh. Then after a month or so, I dumped that idea and went back to the original codebase. I started integrating the sections of code that I rewrote and improved. To be frank, large sections of code in current release of Vicki is a terrible mess and these needed to be rewritten. I'm almost done with that.

I have already rewritten the move generation, which resulted in a speed increase of around 10%. I’ve also changed the representation of the move (I use a 32 bit integer for this) so that a full 8 bits are available for move ordering (this use to be 6). This change will greatly enhance the move ordering capabilities of Vicki.

I am now working on the actual search function and the transposition table. Apart from fixing a nasty bug with the transposition table (it returns a mate score, but just chases the king around with trivial end game positions), I also need to rethink all the pruning and speed-up techniques individually. I threw all these together without giving any much thought. I paid the price.

I'm also planning pondering, opening book learning and better interface support.

The next release is scheduled for the 4th of January 2010.

Vicki is back... and smelling blood!

Frustration

Even though I haven't posted in a while - Vicki is not dead. I'm having huge issues with Vicki at the moment which is taking up a great deal of time. The biggest problem at the moment is the transposition tables that are not working as it should. Every time I implement it, the engine plays weaker...substantially weaker. This can only be as the result of a bug. I hope to have this sorted out in the next week or so...

The transposition table is a very important component and I want that working first, before spending any more time with move ordering and evaluation as these elements share some common ground

Refactoing & revisiting

There are few things as annoying as a dormant blog - and here I am doing the exact same thing. I've recently revisit Vicki and I am in the process of refactoring the engine. I have found some bugs with the repetition detection and I am spending some time to get the engine stable. Once this is complete, I'll start working in getting the engine up to scratch.

One element that has been missing from Vicki is decent testing. For the time being I've identified 28 engines, ranging from weak to rather strong. The idea is to put Vicki in a gantlet against them and work with the final score as an indication of strength (2 rounds, from the starting position). Later on I'll try the test suites again.

Let see how it goes...

New Vicki almost ready...

I've been promising a new release for some time now... Fortunately, my thesis was accepted with little change and I can now focus that attention on my chess engine too. Vicki has been a bit a of a disappointment and I am working hard on getting her ready for WBEC Ridderkerk's 15th edition. Hopefully we'll see an engine around 2100 in stength :-)

Time-off...

Hi all

I need to take some time off from Vicki for a few weeks or so. My thesis has just arrived back from the moderators and I have some minor changes to make before I hand it in on the 24th of August... then there is also a research paper that needs to be submitted before the end of this month. Then apart from that, I also have to work to pay for everything...

So, Vicki will have to wait a while...

Sorry guys :-(

Null moves

I finally got null moves to work in Vicki. I had some trouble with the capturing of the king (which caused Vicki to hang), but I eventually got it fixed. I am utterly amazed at the effect of null moves in the engine. With null moves Vicki searches (on average) 1-2 plies deeper and even more later in the game. In late-middle games, depths of 12 plies are reached! That is without any other form of pruning or transposition tables. The new version of Vicki is already playing on FICS and I am conducting tests against the older version to evaluate the improvement.

Why didn't I have this ready for the WRCC?