People have recently asked the question: "How much stronger is Gothic Vortex Nucelar V than Gothic Vortex Gold IV?"
The diagram above should help answer that question.
Vortex V has the 5-piece Gothic Chess tablebases. These tablebases are probed in RAM as the program searches, they are not just used when you enter into a 5-piece position.
Therefore, Vortex can see that the trades from this position shown above will force black into a pawn-down ending of Queen and Pawn vs. Queen where the side to move (white) will have a mate in 268 moves. The key to the win is 2. Cxh7! and it is the only move to win.
Vortex can announce this win after only 1 second of search (even though the screen shot shows 6 seconds... I didn't capture the first mate announcement quickly enough).
You can imagine there are countless cases as the endgame approaches that the over 300 billion positions can provide a deep insight that will affect the play at the root of the game tree. This makes Vortex play endgames the way God would play them: perfectly, delivering checkmate as quickly as possible, and never making a mistake!
15 comments:
This is probably the wrong place to ask, but I didn't know where to do so, since I haven't seen any documentation online about it (to non-programmers, that is,) or any mailing address to H.G. Muller or the TSCP Gothic author.
I downloaded both WinBoard_F and TSCP Gothic, but I have difficulty using winboard to play against the engine. I put the engine in the Winboard directory, and modified the shortcut to immediately activating crafty to attaches to tscp instead. However, Winboard didn't recognize the game as Gothic Chess, and opened it up as a regular chess game, only to receive an illegal move (i1-h3) from the engine in the second move. What did I miss ?
The "target" line in the shortcut I created for the TSCP Gothic engine is this :
"C:\Program Files\WinBoard-4.2.7\winboard.exe" -cp -fcp TSCPGothic\tscpgothic.exe -fd TSCPGothic -scp TSCPGothic\tscpgothic.exe -sd TSCPGothic
I guess there's something I should add.
My apologies, again, for posting this in an irrelevant post.
You should add:
-variant gothic
Further note that the way you use the -fcp and -fd options, you should have the TSCP-Gothic executable (and opening book) in the folder C:\TSCPGothic, NOT in the winboard directory. Unless you also put the winboard.exe from WinBoard_F in the TSCPGothic directory. But in that case you are calling the wrong WinBoard. Are you sure the WinBoard_F executable did replace the one in C:\Program Files\WinBoard-4.2.7 ?
Be sure to set board size to Middling or Bulky, or you won't see A and C!
Also note that you can watch Gothic engines (amongst which TSCP Gothic) play each other live, on my website:
http://home.hccnet.nl/h.g.muller/battlehome.html
and then click the link to the viewer.
Sorry, I said something wrong: You only have the C:\ in fromt of the pathname of the winboard.exe, not before the engine names.
So the TSCPgothic.exe should be in a folder TSCPGothic, which again should be inside the WinBoard directory. It should not be in the WinBoard directory itself.
Note that the WinBoard version is recognizable through the help->about menu. For WinBoard_F this should say 4.3.12.
There's no "wrong place" to make a comment like that, and you might as well put it in a high-traffic area (as long as it is not SPAM I don't mind.)
H.G. your link is not working (http://home.hccnet.nl/h.g.muller/battlehome.html) so you might want to double check it.
Well, for me the link really works after copy-paste, so there can't be a typo. I guess at the moment you tried the site of my provider must have been down.
My own Chess PC has been up and running all the time (and I still have the same IP address), so the link in that page to my viewer should work too.
So, Ed, there now is a brandnew "Gothic Vortex Nucelar V", fine! As usual this engine will recognize moves far better than your GC online site, which was not even able to recognize a mating check in my last game of BetaSmirf today.
To HGM, your automated engine testing site is working fine. Even though there is playing a SMIRF version not having the current time management inside, our both engines seem to perform well. People will remain interested and often follow your 10x8 event!
The SMIRF I currently run did not seem to do too bad on time, there hardly were any time losses. The purpose of the current time-odds tournament was not so much to test Smirf, as well test the new WinBoard_F time-odds feature, and provide more equal opponents to Matthias' engines, so that I can estimate their rating better. So I decided to stick to the Smirf version I already characterized well in the previous tournament.
Next tourney I will start using the latest Smirf version.
Btw, if you are working on the Smirf time management: the addition of a classical and an incremental time-control mode would be quite useful, and probably not too much work. Currently Smirfoglot 'fakes' those mode, by entering N moves per M minutes controls on a move-by-move basis to the SmirfEngine as M/N minutes for the next move, and for M minutes + S seconds/move for the rest of the game as 60*M+20*S seconds for the rest of the game. The 20 above should be more realistically 40, but I was not sure how extremely Smirf might distribute time in a sudden-death game, and if most of the time given to it comes from the increment, it might count itself rich, and use that time for the first move, before it was actually awarded, and thus forfeit on time.
Harm, when I am watching some current games between SMIRF and TSCP, there are final time frames related 3:1, thus SMIRF will play far below of its possibilities at the end of such games. Well, this is a temporary weakness, hopefully erased in coming events. ;-)
I am not sure what exactly you mean by "related 3:1", but if you mean that Smirf had 3 times less time left in the late end-game than its opponent, despite the fact that it started with equal time or even more time than its opponent, you are right. Note that this is Smirf 1.71d, though, and that many games already start with one of the engines having upto 54 times as much time on the clock than its opponent.
But it shows how tricky a task time management is in sudden-death games. By just using just a little bit more time than the opponent (say 10% more), in a long game this can bring you in a position where he really has left a large factor more time than you have yourself, with disastrous consequences. I witnessed a game today were Smirf's flag fell in KQK, where Smirf just did't have enough time left to march his King all the way from the other side of the board to perform the mate.
Reinhard,
Just so you know, the Gothic Vortex program that I give away for free only does a 7-ply search and it does not have all of the evaluation function on.
I hope SMIRF can beat it. A 10-year old girl won 2 games against it in December of 2007.
The Vortex version that is for sale is much, much stronger.
Ed, I have not addressed your free Vortex program in my postings here.
Indeed, 7 ply is nothing. With that limitation it would most likely not even beat Chancellor and BigLion80. Although it might outsearch them by one ply in the middle-game, it would not be enough ahead to survive the end-game if it kept the depth fixed.
Post a Comment