REBEL-EAS

Peter Berger wrote:

Admin wrote:

Install REBEL-EAS.EXE in the same folder as Chess System Tal 2.0

Book are stored in the polyglot-books folder.

This doesn't work for me.

As Rebel EAS is working engine-wise I assume I have set up the engine correctly in principle.

Also the book is working with Rebel 16.2, so I assume the book is set up correctly, too.

I am out of ideas.

Happy Christmas time!
Peter

I figured it out. While the book file is actually named book.bin on my computer you have to give Book.bin (upper-case) to make it work - no idea why.

I have huge problems with Rebel EAS (and also Chess System Tal). While Rebel 16.2 is perfectly stable and has been running on my computers for very long times for several months without any trouble, Rebel EAS (and CS Tal) keep crashing at the long time controls I am mostly interested in. After like 1 hour thinking time at most they usually crash and Fritz tells me, that this is due to a problem with the engine.

Has anyone experienced the same? Can anyone give a tip how to solve this problem?

Happy new year!
Peter

Peter Berger wrote:

I have huge problems with Rebel EAS (and also Chess System Tal). While Rebel 16.2 is perfectly stable and has been running on my computers for very long times for several months without any trouble, Rebel EAS (and CS Tal) keep crashing at the long time controls I am mostly interested in. After like 1 hour thinking time at most they usually crash and Fritz tells me, that this is due to a problem with the engine.

Has anyone experienced the same? Can anyone give a tip how to solve this problem?

Happy new year!
Peter

When you experience a crash, can you post the data? EPD, mainlines, time control.

It would be interesting to know if it happens on "infinite" or "analyze" as well.

Here is a pretty reliable way to crash Rebel EAS in an acceptable amount of time on my computers.
I change threads to 4 and hash to 12GB, give the book file and set book depth to 100, all else default.
I start the program in analyze mode (it shows a book move) , then input 1. h3 and let it analyze further on its own.
It starts to output lines and seems to work fine - after some time (in my most recent try at 26:21 at depth 36/65) it crashes.
Fritz gives the following message:
"Chess-System-Tal-2.00" caused an exception.
This is a problem in the engine.
Win 10.0 x64
16123 MB
Intel(R) N200
N=1/P=1/C=4/L=4/G=1
Intel(R) UHD Graphics
Intel(R) UHD Graphics
Intel(R) UHD Graphics

I can add the last line given:

Neue Partie Line 0.0
rnbqkbnr/pppppppp/8/8/8/7P/PPPPPPP1/RNBQKBNR b KQkq - 0 1

Analysis by Chess-System-Tal-2.00:

1...e5 2.e4 Sf6 3.Sc3 Lb4 4.a3 La5 5.Sf3 0-0 6.Lc4 c6 7.La2 d5 8.0-0 dxe4 9.Sxe5 Dd4 10.Sc4 Lc7 11.Se2 Dd8 12.d4 exd3 13.Dxd3 Dxd3 14.cxd3 Lf5 15.Td1 Te8 16.Lf4 Sa6 17.Lxc7 Sxc7 18.Sg3 Le6 19.Sd6 Teb8 20.Sge4 Sfe8 21.Sxe8 Txe8 22.Sd6 Teb8 23.Lxe6 Sxe6 24.d4 Kf8 25.d5 cxd5 26.Txd5 Td8 27.Tad1 Td7 28.Se4 Txd5 29.Txd5 Ke7 30.Te5 Td8
= (-0.20) Tiefe: 36/65 00:20:53 607MN
Die Stellung ist ausgeglichen
(Berger, Berlin 31.12.2023)

Peter Berger wrote:

Here is a pretty reliable way to crash Rebel EAS in an acceptable amount of time on my computers.
I change threads to 4 and hash to 12GB, give the book file and set book depth to 100, all else default.
I start the program in analyze mode (it shows a book move) , then input 1. h3 and let it analyze further on its own.
It starts to output lines and seems to work fine - after some time (in my most recent try at 26:21 at depth 36/65) it crashes.
Fritz gives the following message:
"Chess-System-Tal-2.00" caused an exception.
This is a problem in the engine.
Win 10.0 x64
16123 MB
Intel(R) N200
N=1/P=1/C=4/L=4/G=1
Intel(R) UHD Graphics
Intel(R) UHD Graphics
Intel(R) UHD Graphics

I don't have Fritz any longer, I simply forgot to add a dvd-reader when I ordered my new PC's, but I have Arena running in Analyze mode. After 1.h3 the engine is still in book with 1..d5 but after 2.d3 it's now analyzing. After 4 minutes at depth 38, we will see what happens.

Peter Berger wrote:

Here is a pretty reliable way to crash Rebel EAS in an acceptable amount of time on my computers.
I change threads to 4 and hash to 12GB, give the book file and set book depth to 100, all else default.
I start the program in analyze mode (it shows a book move) , then input 1. h3 and let it analyze further on its own.
It starts to output lines and seems to work fine - after some time (in my most recent try at 26:21 at depth 36/65) it crashes.
Fritz gives the following message:
"Chess-System-Tal-2.00" caused an exception.
This is a problem in the engine.
Win 10.0 x64
16123 MB
Intel(R) N200
N=1/P=1/C=4/L=4/G=1
Intel(R) UHD Graphics
Intel(R) UHD Graphics
Intel(R) UHD Graphics

12 GB Hash on a 16 GB-RAM machine? This must lead to problems (or crashes). Windows 10/11 itself needs at least 6 GB (better 8 GB) free RAM. So, my advice is: Retry with Hash=8 GB

pohl4711 wrote:

12 GB Hash on a 16 GB-RAM machine? This must lead to problems (or crashes). Windows 10/11 itself needs at least 6 GB (better 8 GB) free RAM. So, my advice is: Retry with Hash=8 GB

While that's true CSTAL with threads=4 automatically limits the hash table size of 12Gb to 8Gb. Runs fine under Arena, maybe Fritz decides otherwise. Peter should look into the task manager how much ram his system uses.

BTW, happy new year to all....

I had no idea that an OS could be THAT bloated, thanks for information.Anyway, much as predicted by Ed, lowering the hashtable size didn't help. As no one besides me seems to experience this problem and all other programs work well, I will have to accept, that I can't use ChessSystemTal and Rebel EAS.

Peter, do you have an onboard intel graphic card ?!

I think Peter may be the only one right now doing these tests, so not to assume it is not important.. I think Chris mainly tests with Cutechess CLI and is not really interested in these long rang games? Ed tests with Arena but that is not really standard UCI sorry Ed . Matejst and Mike Young probably have Fritz but I don't think they are very active?  Dann Corbit has his own scripts and is on Linux, mostly. Most chessplayers only know Stockfish and newcomers only know the Fritz programs and Fritz users are probably all very young these days if I look at Silvian's pictures... Well that's not a bad thing. Chess as part of the curriculum.

Nowadays, I see they are selling Rubik's cubes too that are meant especially as a challenge for young kids, where they only have to solve one side for instance. It's a nice idea. Also I never knew about all those new architectures for cubes. I bought a 4 x 4 real Rubik's 'Master' cube back in the early eighties but never learned to solve it by myself. I have all those books from Jan van den Craats about the 3x3 mainly , well maybe not all. I bought some of those American at the American Bookshop here I think, it does not exist anymore here in the North, only in The Hague I think. I must have at least ten books in Dutch too, I just collected them, and various of the real Rubik's cube. Took them apart to apply vaseline over the moving surfaces to make them faster and more stable. It wasn't a sport yet then. I used to be reasonably fast solver back then but that was before the speed contests I think.

Bought a Christmas present for myself, these are not so expensive as some other brands, compared to the real Rubik's cubes back in the eighties and correcting for inflation, I think about the same priceranges. But there are also much larger cubes that cost hundreds of Euros. Well, this is difficult enough for me right now.

I was in high school and we all competed against each other with rubik cube.
At that time we could solve it by heart in maybe 1 minute.

My son today is faster.
At the time we solved it there was no internet.
The key moves were all shown and traded either by showing or by written down moves on real paper.

There are people solving the cube today amazingly fast.

This was not possible in the years i learned it. We had much more moves to make and were amateurs.

Mclane wrote:

Peter, do you have an onboard intel graphic card ?!

Yes, all my three computers (I am getting pretty nerdish again ) do have an onboard intel graphic card. Well spotted, though I don't expect this piece of information will help too much.

Peter

Peter Berger wrote:

Mclane wrote:

Peter, do you have an onboard intel graphic card ?!

Yes, all my three computers (I am getting pretty nerdish again ) do have an onboard intel graphic card. Well spotted, though I don't expect this piece of information will help too much.

Peter

Usually onboard graphic cards can be adjusted in bios concerning the amount of GB they get from the absolute GB the machine has.

Eelco wrote:

I think Peter may be the only one right now doing these tests, so not to assume it is not important..

There probably is some general truth in this.
I am having a deeper look in the German world of correspondence chess currently - the most interesting conclusion is how tiny it is. A very conservative (safe) estimation is that we are talking 500 active players at most - the real number is probably way lower.
Who else is looking at chess positions for a long time, and which tools are they probably using? Top human players do, but what will they use? Stockfish for obvious reasons, maybe Leela or Dragon for sake of variety.
This leaves the computerchess world. There are the guys who do rating lists. They have become pretty scientific (loads of games, careful setup). Which means - they are basically doing the same kind of tests the developers are doing themselves.
And if you find a position where a specific engine does something stupid - what would it be good for to know? Let's say, the engine plays a stupid sacrifice. With the NNUE no one basically cares too much, as no one really knows why it went for it - and there is no obvious change to make it stop to do so.
The guys who still do long games and actually look at them are more like wine wine connoisseurs - they do it for their personal enjoyment. They may enjoy some engine as it suits their personal taste - but this is nothing someone else can profit from.
So, yes: it is a good question about what kind of user numbers we are even talking about when it is about chess engines below the top 3.

Peter Berger wrote:

There are the guys who do rating lists. They have become pretty scientific (loads of games, careful setup). Which means - they are basically doing the same kind of tests the developers are doing themselves.

The guys who still do long games and actually look at them are more like wine connoisseurs - they do it for their personal enjoyment

The good news is: Today, you can have and do both:
I do my "pretty scientific (loads of games, careful setup)" with my UHO-Top15 Ratinglist. Each engine plays 15000 games (15 opponents * 1000 games). But that does not mean, that I do not look at any games or do not enjoy them. But to do so, I needed to write some tools, to filter the interesting games out of the huge amount of games. But, I managed to do my Interesting Wins Search Tool and now, I filter the games very fast and find the best ones. And I sent some of these games (and some, I filtered out of the Blitz- and Bullet enginegames on chesscom) to the german Youtuber TheBigGreek (just 2 examples here):
https://youtu.be/_XbDLQZPEfo?si=2w9tWpWx9oD0DGHS
https://youtu.be/JleRxjFRRIU?si=JGQNjq_bJY8TPUUJ

And part of the truth is: Really spectacular enginegames are rare. And this means, the more enginegames you have, the more spectacular games you will get. And the probability of seeing a spectacular game "live", when a enginematch is running, is very small.
So, my conclusion here is: Play many, many enginegames, the more the better. And filter them, to find the really cool games. My tools will do the work for you and they are very, very fast! And much enginegames means (additionally), you get better and reliable statistics.
So, for me, there is today absolutely no reason to play slow enginegames and watch them live. You will see only a few cool games (most of them are high-level chess, but boring) and get no proper statistics out of the games. And the higher chess-level, using longer thinking-times, does absolutely not matter, because top enginechess is so far beyond all human chess-skills, that +50 or +100 Elo gaining with long thinking-times, are meaningless for any human.


https://www.sp-cc.de/files/interesting_wins_search_tool.7z
https://www.sp-cc.de/files/engines_aggressiveness_statistics_tool.7z
https://www.sp-cc.de/files/sacrifice_games_search_tool.7z

In recent years I have mostly done games with pretty uneven setups (one of the engines being clearly stronger than the other one). This often led to very interesting games to watch.
I enjoy watching games live, as I don't know the result yet. So I can look at the development of the games, think about them for myself, and then watch how they play out and experience the surprises. I especially enjoy watching the opening phase.
But it is true that with the development of the NNUE engines, this has become more difficult. They are too  strong for the previous generation, so that there are too few upsets in matches to keep the thrill.
And when you pit them against each other in even setups, they mostly draw, often in very uneventful ways. Currently this is still kind of interesting for me as my opening knowledge is outdated, and I see some interesting "novelties" (for me) in the opening.
Recently I tried some slow games Rebel 16.2 - Stockfish. They were very boring indeed.
You pointed out a possible way to deal with this development. This may be the way to go. I'd still prefer to watch games live when I have the time. But I haven't found a good solution for the problem you mentioned yet.