Your Country Needs You!

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
Josef
Member
Posts: 178
Joined: Mon Dec 12, 2016 12:56 pm

Post by Josef »

Yes, but I have an older version of Smallfish because of my older iPad cannot install OSX 10. New Smallfish has Stockfish 8 engine.
Josef
Member
Posts: 178
Joined: Mon Dec 12, 2016 12:56 pm

Post by Josef »

I compared iPad Stockfish 5.0 versus Revelation II Stockfish 6.0. iPad is 6 times faster than Revelation and iPad didn't beat the Revelation in 30s time per a move.

5 games:

1-0
1/2-1/2
0-1
1/2-1/2
1/2-1/2
Josef
Member
Posts: 178
Joined: Mon Dec 12, 2016 12:56 pm

Post by Josef »

This is final Test 5 compared with Mephisto V. Please note weak moves of Revelation II - 20.c6 and 26.Qc5, Mephisto played these moves much better!

This is the final result of Revelation II Stockfish 6.0, you can see, Revelation II is very strong machine!

Image

Image
JeffB
Member
Posts: 420
Joined: Sat Mar 07, 2015 6:36 pm
Location: Seattle, WA USA

Post by JeffB »

How does a mere human player stand a chance? (Is it sacrilege to even ask that question?)
Cheers,

Jeff B.
Josef
Member
Posts: 178
Joined: Mon Dec 12, 2016 12:56 pm

Post by Josef »

There's no chance whatsoever, the best players have ELO 2800, best chess machines have ELO 3400-3600.

http://www.computerchess.org.uk/ccrl/404/
JeffB
Member
Posts: 420
Joined: Sat Mar 07, 2015 6:36 pm
Location: Seattle, WA USA

Post by JeffB »

My point exactly. :(
Cheers,

Jeff B.
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

I thought I would show the results of Test Game 4. This was a game originally played in the year 1783 between Thomas Bowder and Francois Andre Philidor. Bowdler is one of the few documented opponents that Philidor played without spotting the opponent a piece.

Below is the list of all the programs that were tested with this game. I also included Bowdler's and Philidor's performance score in order to compare them against the tested chess programs.

Image

This game was played 235 years ago! No computers, at a time in England where people lived by eating meat, more more meat, bread and cheese. Very few vegetables (frowned upon, poor man's food) were eaten by the upper class and from the time they got up in the morning until they went to bed at night they drank Beer, more Beer, Wine, more Wine and of course the bottled hard stuff like Whiskey, Gins, Port, Sherries, Cognac etc. Water was undrinkable and therefore no one drank it.

So in summary the middle/upper class person went through life from infant to deathbed being permanently half drunk and with guts full of meat and more meat. Whatever moved and was caught was eaten or put into pies.

Yet here are these two gentlemen living under these conditions performing better than most of the chess programs tested 200 plus years later. Philidor ranked 70th out of 245 tested chess programs in the above list with his game rating of 2339 ELO and Bowdler ranked 116. Even their combined rating of 2251 ELO ranks 84!

You just have to look at the programs that are below Philidor to be amazed at his ability in the year 1783 while playing half drunk!! :P

He was truly a super grandmaster back then and probably would easily be one today too!! :P

Just imagine what the Philidor's of the past would be able to achieve nowadays while playing sober!! LOL

Best regards
Nick
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

With Tasc R30 it often unclear which setting is the strongest, so here are a few comparisons as a result of these tests:

Tasc R30 - King 2.5 - 1995 - Johan de Koning

Image

Most people use the Active setting for R30 King 2.5 which the above tests seemed to confirm. But I am really not so convinced yet. Active shows 17 ELO points better but mostly it is as a result of game 2 where it out performed Normal setting by 279 points. Normal however beat Active in 3 out of 5 tests. So I think it is still unclear what the best setting is.

In the past I leaned towards Active for King 2.5 but more recently I am leaning towards Normal setting.


Tasc R30 - King 2.2 - 1993 - Johan de Koning

Image

For R30 King 2.2, Normal rated higher overall by 10 points. Normal also scored best in head to head wins. But still it is misleading a little because it beat solid 5-0 yet only by 4 points in games 1 and 4.

Also if you study the above, I doubt if the programmers really knew what type of style is defensive and what type of game is offensive. So it's just a name and really does not reflect a type of game played. You just have to look and compare style results between King 2.5 and King 2.2 to see what I mean. One style result of one program is almost the exact opposite of the other program.

Tasc CM32 - King 2.2 - 1993 - Johan de Koning

Image

With the fastest 32 MHz Chessmachine card it is a little more clear cut that Normal setting is the best for King 2.2.

Tasc CM32 - King 2.54 - 1995 - Johan de Koning

Image

Same with King 2.54 on the 32 MHz Chessmachine Normal setting is overall best.


Tasc CM30 - King 2.54 - 1995 - Johan de Koning

Image

On the slower CM30 card it is more inconclusive even though Normal setting seems to be best.

Tasc CM32 - Gideon Madrid 3.1 - 1992 - Ed Schroeder

Image

With Ed Schroeder it is also not so easy to definitively determine the best setting. In these tests Aggressive setting came out on top for the faster 32 MHz card.


Tasc CM32 - Gideon 3.0 - 1992 - Ed Schroeder

Image

The earlier chess program version Gideon 3.0 does show a marked difference in the Test Rating score in favor of the Aggressive setting. Yet only wins 3 tests to two.

So if you look at this mixed bag of scores what really is aggressive and what is defensive with these programs?

I would consider game 4 of all these test games the most passive/defensive type of game. Yet I cannot see how any of these programs really reflect anything from their game styles between versions.

What do you see?

Best regards
Nick
User avatar
Dave C
Member
Posts: 460
Joined: Fri Jul 17, 2015 1:50 am
Location: SoCal, US

TASC Settings

Post by Dave C »

Hi Nick,

Interesting analysis of the variety of playing styles available on the R30 - 2.5. I am curious about test game 2. What is it about game 2 that the Normal setting did not play nearly as well as the Active setting? Any ideas?

Concerning the other test games: Normal played better in 3 of the other 4 test games which might mean it is slightly better. But, what about when you are playing and that occasional game occurs that is similar to test game 2 and Normal may play significantly weaker than Active.

Seems like Active is a little more consistent.

Best regards,
Dave
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Re: TASC Settings

Post by spacious_mind »

Dave C wrote:Hi Nick,

Interesting analysis of the variety of playing styles available on the R30 - 2.5. I am curious about test game 2. What is it about game 2 that the Normal setting did not play nearly as well as the Active setting? Any ideas?

Concerning the other test games: Normal played better in 3 of the other 4 test games which might mean it is slightly better. But, what about when you are playing and that occasional game occurs that is similar to test game 2 and Normal may play significantly weaker than Active.

Seems like Active is a little more consistent.

Best regards,
Dave
Hi Dave,

Yes it requires more tests for sure. I look at it this way:

King 2.5:

wins 1 game by 137 ELO
wins 1 game by 62 ELO
wins 1 game by 14 ELO
loses 1 game by 11 ELO
loses 1 game by 279 ELO

so losing 1 badly and winning one cancels each other out.
winning by 14 and losing by 11 cancels each other out.
that leaves 1 win with a superiority of 62 ELO and I'll take that !! :)

Especially since its predecessor showed normal as best setting too. There has to be a reason why the programmer thought normal setting is best (this is what they typically think when they create a default or normal setting :)

Best regards
Nick
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Here is an interesting match comparison that was done in Selective Search in 1995:

Image

This was obviously King 2.2 because King 2.5 had only just come out.

It also seems to show that Normal is slightly better in those test matches.

Regards
Nick
User avatar
Dave C
Member
Posts: 460
Joined: Fri Jul 17, 2015 1:50 am
Location: SoCal, US

TASC style comparison

Post by Dave C »

spacious_mind wrote:Here is an interesting match comparison that was done in Selective Search in 1995:

This was obviously King 2.2 because King 2.5 had only just come out.

It also seems to show that Normal is slightly better in those test matches
Interesting chart...thanks. Against a variety of competitors this chart seems to confirm the thought that Normal is slightly better...with version 2.2. And, 2.5 being a later development of the 2.2, the comparison of playing styles is consistent with your testing of 2.5.

Many thanks,
Dave
Four.nine
Full Member
Posts: 911
Joined: Tue Mar 21, 2017 7:30 pm

Post by Four.nine »

Spacious:

No specific comments just a general observation:
Your dedication to chess computing and chess program analysis is astonishing and outstanding.
Great work!

As for me, I will be leaving the chess computer scene soon for my annual 7 month Ben Dover "sabbatical".
You see, my dedication to my pool and deck, yard and lawnmowers, canna lily and watering, obligatory/perfunctory repairs and upgrades is astonishing and outstanding.

Keep the excellent analysis going...see you next winter.
User avatar
blaubaer
Full Member
Posts: 935
Joined: Thu Jul 28, 2011 12:53 pm
Location: Bavaria, the centre of Mysticum
Contact:

Post by blaubaer »

Hi Nick,

every year when we meet in Klingenberg (also April 20th to 22nd this year - Klingenberg is located near Frankfurt!!!) we talk about your absolute dedication to chess computers, computer chess and chess in general. This thread is an excellent example but there are lots of others... 8)

When I saw your huge test game 4 list above first time, I though I ate 2 pounds of ribeye steak without any vegetables - absolute fantastic and almost incredible... :P

What do you do besides chess? If I would not know the answer I would ask myself that question every second day... :wink:

Then I got a really bad feeling when I saw the first entry of the list - you know why....:roll:

See you or some other board members in Klingenberg to discuss this topic and drink some beer!

Adoring Regards, Michael
Josef
Member
Posts: 178
Joined: Mon Dec 12, 2016 12:56 pm

Re: Your Country Needs You!

Post by Josef »

Some time ago I bought an interesting CT800 chess computer, so I tested the game power against the published result from the Reflection Vancouver chess module. I tested the CT800 in two mods, 10s and 30s. The CT800 proved to be a very strong opponent, especially in the 30s setting.
Attachments
chess1.jpg
chess1.jpg (94.86 KiB) Viewed 389 times
chess2.jpg
chess2.jpg (95.36 KiB) Viewed 389 times
chess3.jpg
chess3.jpg (96.19 KiB) Viewed 389 times
Post Reply