Stockfish 8 versus Excalibur Saber 4

This forum is for general discussions and questions, including Collectors Corner and anything to do with Computer chess.

Moderators: Harvey Williamson, Steve B, Watchman

Forum rules
This textbox is used to restore diagrams posted with the fen tag before the upgrade.
Post Reply
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Stockfish 8 versus Excalibur Saber 4

Post by spacious_mind »

Just reading the results that were posted for Stockfish 8, made want to do a comparison to Excalibur Saber 4.

Stockfish 8 = 64 Bit Program = 1163 KB in size = speed for argument sake 4 GHz = @ 3400+ ELO = double in strength to Saber 4.

Saber 4 = 4 Bit Single Chip = 1.5 MHz = 8 KB in size = 1705 ELO (Colditz test that Paulwise recently carried out with it)

So if you take 1163 KB and multiply it by 16 (64bit/4bit) = 18,608 (4 Bits)
Divide that by 8 (Nelson) and Stockfish 8 program is 2,326 times bigger than Saber 4! Which means in order to achieve a doubling in ELO's from 1700 to 3400 it needs to be 2,326 times more programming than what Ron Nelson achieved with his 4 Bit program!!

Stockfish has a team of about 200 people working on it. Ron Nelson was one person!

Wow it really makes you wonder what's going on with this picture ( :P <---- tongue in cheek laugh)

Saber 4 plays at 1.5 MHz.... I am not even going to go there in the multiplication factor of 4 GHz over 1.5 MHz for a measly doubling of ELO improvement ( :P <---- another tongue in cheek laugh)

Ron Nelson's 4 bit program has to be the World Champion of 4 Bitters never to be beaten and a historical programming genius. (Hall of Famer for sure)

ps.. Ron Nelson's 8 KB also included a small opening book for which you would have to add another couple of MB for Stockfish!

Best regards
Nick
herO
Member
Posts: 342
Joined: Wed Nov 12, 2014 8:49 am

Post by herO »

Thank you for your comparison between those engines :)
User avatar
paulwise3
Senior Member
Posts: 1505
Joined: Tue Jan 06, 2015 10:56 am
Location: Eindhoven, Netherlands

Post by paulwise3 »

Hi Nick,

Nice comparison! :-P
Good to see I didn't spent this testing time for nothing... :-)

Edit: I just finished the Schachcomputer Testreport for the Saber IV

Best regards,
Paul
2024 Special thread: viewtopic.php?f=3&t=12741
2024 Special results and standings: https://schaakcomputers.nl/paul_w/Tourn ... 25_06.html
If I am mistaken, it must be caused by a horizon effect...
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

paulwise3 wrote:Hi Nick,

Nice comparison! :-P
Good to see I didn't spent this testing time for nothing... :-)

Edit: I just finished the Schachcomputer Testreport for the Saber IV

Best regards,
Paul
Hi Paul,

Yes I just looked at it. Unfortunately its an apples to oranges comparison if Saber 4 cannot do the same time as the other tested computers because of its 144 max seconds would not compare to remotely close to the rest of the computers receiving a 2hrs/40 setting. Besides I really doubt that this setting is consistent for all computers tested. I have enough computers where I know that if I set up a position and then set it to 2hrs/40 that a lot of computers would run well beyond 3 minutes of the first move in some cases going as far as 5 or 6 minutes. This kind of test is really only useful if all the computers were to be set at an exact fixed time and not something like 2hr/40 where some computers take advantage of their built in average time calculations which could vary for move 1 from anything from 1-2 minutes to 5-6 minutes.

Best regards
Nick
User avatar
paulwise3
Senior Member
Posts: 1505
Joined: Tue Jan 06, 2015 10:56 am
Location: Eindhoven, Netherlands

Post by paulwise3 »

Hi Nick,

In this (Testreport) case I disagree with you. True, it is a disadvantage that Saber IV's max. time/move is 144 seconds, but if I let it compute longer it does not score much better. Only test 2 (tactical, within 4 minutes) and test 24 (endgame, within 6 minutes) are solved as extra. So tactical it scores reasonable, but positional/strategic and in endgame it scores very poor. It has only very basic knowledge.

But still it is a good competitor, I tried a few games at 1 minute/move (level 32), and had do do my best to get the win.
The pieces are like glued on the board, so very suitable for use on a holiday trip :D.

Best regards,
Paul
2024 Special thread: viewtopic.php?f=3&t=12741
2024 Special results and standings: https://schaakcomputers.nl/paul_w/Tourn ... 25_06.html
If I am mistaken, it must be caused by a horizon effect...
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

paulwise3 wrote:Hi Nick,

In this (Testreport) case I disagree with you. True, it is a disadvantage that Saber IV's max. time/move is 144 seconds, but if I let it compute longer it does not score much better. Only test 2 (tactical, within 4 minutes) and test 24 (endgame, within 6 minutes) are solved as extra. So tactical it scores reasonable, but positional/strategic and in endgame it scores very poor. It has only very basic knowledge.

But still it is a good competitor, I tried a few games at 1 minute/move (level 32), and had do do my best to get the win.
The pieces are like glued on the board, so very suitable for use on a holiday trip :D.

Best regards,
Paul
Hi Paul,

It is not necessarily Saber 4 that I am concerned about. The different computers when you set them up in a position at 2hrs/40 will all move at different times for their first move. Some programs are set up to take more time early in the game which is why you often see 5 or 6 minutes for a first move out of book. These computers tend take more time as they get closer to 40 moves because they have plenty of time left on the clock.

Other programs do the opposite of this.

By setting up a position the clock starts with move #1 out of 40.

It is why I think fixed time is maybe a better way to accurately compare all the programs from a fixed position starting point.

Best regards
Nick
User avatar
paulwise3
Senior Member
Posts: 1505
Joined: Tue Jan 06, 2015 10:56 am
Location: Eindhoven, Netherlands

Post by paulwise3 »

Hi Nick,

Yes, you are right about the time usage. I just finished a 4 game match 30 secs/move between the Saber IV and the Saitek Sensor XL. The Saber played at level 20, where it mostly moves between 25 and 30 seconds, and sometimes longer, up to 45 seconds. The XL often takes about 60 secs/move in the opening and the early middlegame, and hardly ever moves faster then 30 secs/move, except for a few direct answers. So apart from maybe a better program, it has this time usage advantage. Plus that it ponders...

Anyway, the Saber lost 3,5 - 0,5. In two games the Saber made very obvious blunders, I tried those also at level 24, but even then it made the same bad move.

Best Regards,
Paul
2024 Special thread: viewtopic.php?f=3&t=12741
2024 Special results and standings: https://schaakcomputers.nl/paul_w/Tourn ... 25_06.html
If I am mistaken, it must be caused by a horizon effect...
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

paulwise3 wrote:Hi Nick,

Yes, you are right about the time usage. I just finished a 4 game match 30 secs/move between the Saber IV and the Saitek Sensor XL. The Saber played at level 20, where it mostly moves between 25 and 30 seconds, and sometimes longer, up to 45 seconds. The XL often takes about 60 secs/move in the opening and the early middlegame, and hardly ever moves faster then 30 secs/move, except for a few direct answers. So apart from maybe a better program, it has this time usage advantage. Plus that it ponders...

Anyway, the Saber lost 3,5 - 0,5. In two games the Saber made very obvious blunders, I tried those also at level 24, but even then it made the same bad move.

Best Regards,
Paul
Yes I know I know, remember that Saber 4 is only a 4 bit program so its not going to be fantastic.

It is very confusing with the levels. If you look at the RadioShack manual for what probably is the exact same program as it has the same levels:

http://www.spacious-mind.com/html/trave ... ckers.html

Manual

http://support.radioshack.com/support_g ... /63368.pdf

RadioShack clearly defines the levels and shows 30 seconds as Level 25. Lately I have been sticking to this level for 30 second games as a result of this.

Best regards
Nick
Nick
Post Reply