Chess Genius Pro

[Yarc]

We do have a small sample already of the differences between the London and Hiarcs books

Lexman's first 5 games used the London book and the CGPRO scored an abysmal 1.5/5
after switching to the Hiarcs book the next 5 games produced a dominating performance of 4/5

earlier in this thread..Swarm mentioned a 4 game match MCGP vs Tasc R30 version 2.5 and the score was 2-2 at that point..i think the London book was being used (but i am not 100% certain)
50% Vs the R30
my 10 games with the H book produced a 50% score Vs the R40 v2.5 (stronger then the R30 by about 50 Elo)

i guess i could play another 10 game match Vs the R40 using the London book but i will wait until your matches are completed
if you obtain the same stark differences in the results of the books then i will be satisfied that the optimal settings for the MCGPRO is.. ECO OFF and ....

Hiarcs Human book ON Regards
Steve

Hi Steve,

I have been following the results and its very interesting but think we need more results to make a decision. For example, we might be seeing that against certain computer opponents the Human book is better but then with others, the London book may be more effective. This would confuse things a little as to which is the best option.

I would like to think that the MCGP plays just as well with either book and any losses it has against weaker opponents are just mistakes in the program, but landslide victories are quite good evidence even with the small numbers of games being played.

This London vs Human book scenario is certainly making some interesting conversation!

I am definitely pleased with my MCGP, for a relatively small cost I now own one of the strongest dedicated machines, even if it does not look as good as the TASC R30/40.

Game 3 is in progress but too early so say which side has the edge. I'll post the results soon.

The Book Debate Regards

Ray

[Yarc]

Game 3 of my 10 game match between MCGP and Stardiamond is still in progress. Currently on move 90 and there has been much shuffling around. I'll let them play a bit more but it's looking like a draw. Well done SD!

[Sargon1972]

Hello Ray ,


As i did see , Steve "defend" a little the Hiarcs Book what is not needed, i did say the book is not bad but the CGP do not understand many lines of it ,

And yet we have 17 games posted here where the CGP using the Hiarcs book scores 10.5/17 ....against three different opponents
so ..that sounds like a pretty good result to me ...I guess what you are saying is the results would be EVEN BETTER if we had the Hiarcs book and a different engine other then Lang's instead ...?
the main question for me is this ..is the Hiarcs book producing better results then the London book??
do you have any results comparing games played with the Hiarcs book and then London book against the same opponents?

Hello Steve, as i wrote before the Lang program do not understand many lines of the Hiarcs Book , also the CGP is the android version so a old stripped version of the Roma i believe ,many chess Knowledge is gone here unfortunately
Well the Book of Hiarcs is a modern book so i think a modern program would handle it quite well
So the best vs other programs is indeed the Hiarcs Book , play u vs a other Lang program i would take the Londen Book , that is what i did discover during testing ,nerveless the CGP have big problems with later Lang Programs .Not saying that it can not win games vs them
I did played vs Res Ruffian 10 games , 7,5-2,5 for Ruffian with the Hiarcs Book , i thought this can not be , so another 10games and now CGp wins with 7-3
At the end of the testing result over 100 games a elo of 2403! ,but i also wrote that this would end 2350+-
mine test games are on schachcomputer.info
Also i want to say that after a period i did get the final version ,so all previous test games with Prototype 1 are out , so i had to start over new again
Time was not mine partner ,so i play mostly 30 sec a move chess , at the end i did play 40-120 games vs R30 2.23 and V11
SD did lost against all 2 versions with 10-0 on active chess
Also the Atlanta and Magallan did not win a game in 30 games
Funny is then to see on schachcomputer.info that the Master Chess on 27Mhz wins some games

Kr,Hans

Comparing Regards
Steve

[Sargon1972]

Hello ,

All games where vs CGPro
Results active chess

CGPro -Res I Ruffian 9,5-10,5

CGPro- Star Diamond 10-0

CGPro- Atlanta 7,5-2,5

CGPro -Magellan -7,5-2,5

CGPro -V11 5,5-4,5

CGPro - CM Madrid 3.1 Aggr. 7-3

CGPro- R30 2.23 Aktiv 4,5-5,5

CGPro-Montreux Aktiv 5-5

CGPro- Risc 2 1MB 6-4

tournament
CGPro-R30 2.23 4,5-5,5

here the games scroll down
http://www.schachcomputer.info/forum/sh ... 01&page=15

http://www.schachcomputer.info/forum/sh ... 01&page=18
Mine name there is Sargon there are others to who played games of course all to find in this thread,enjoy

Kr,Hans

[Steve B]

All games where vs CGPro
Results active chess

CGPro -Res I Ruffian 9,5-10,5
CGPro- Star Diamond 10-0
CGPro- Atlanta 7,5-2,5
CGPro -Magellan -7,5-2,5
CGPro -V11 5,5-4,5
CGPro - CM Madrid 3.1 Aggr. 7-3
CGPro- R30 2.23 Aktiv 4,5-5,5
CGPro-Montreux Aktiv 5-5
CGPro- Risc 2 1MB 6-4

tournament
CGPro-R30 2.23 4,5-5,5

Thanks Hans
Which Book Regards?
Steve

[Sargon1972]

Hello Steve,

Your favourite Book
Kr,Hans

[Yarc]

The result of game 3 was a draw by adjudication. After 90 moves no progress was being made therefore decided to let the computers play until 100 moves and then adjudicate if required.

They followed the Ruy Lopez, Marshall, main line and the SD was out of book at move 14 whilst the MCGP was out of book at move 16. I use ChessGenius3 on my mobile to store the games and this showed the incorrect book opening name, it had English Symmetrical.


Here is the game:

[Event "40m/2h H book"]
[Date "2016.09.26"]
[Round "3"]
[White "SD"]
[Black "MCGP"]
[Result "1/2-1/2"]
[ECO "C89"]

1.e4 e5 2.Nf3 Nc6 3.Bb5 a6 4.Ba4 Nf6 5.O-O Be7 6.Re1 b5 7.Bb3 O-O 8.c3 d5 9.exd5 Nxd5 10.Nxe5 Nxe5 11.Rxe5 c6 12.d4 Bd6 13.Re2 Qh4 14.g3 Qh3 15.Nd2 Bf5 16.Ne4 Be7 17.f3 Qh5 18.Qd3 Rfd8 19.a4 Qg6 20.Bd2 Rdb8 21.Qc2 Re8 22.Rae1 Red8 23.a5 c5 24.Bxd5 Rxd5 25.dxc5 Rxc5 26.Qb3 Be6 27.Qb4 Rc7 28.Qd4 Rd7 29.Qe3 Qh5 30.Bc1 b4 31.Ng5 Bxg5 32.Qxg5 Qxf3 33.cxb4 Rb8 34.Qc5 Qb7 35.Bd2 Rc8 36.Qe5 Rd5 37.Qe3 Rcd8 38.Bc3 h6 39.Rd2 Rxd2 40.Bxd2 Bh3 41.Qf2 f6 42.Re3 Bg4 43.Bc3 Bh3 44.Qc2 Rd1+ 45.Re1 Rxe1+ 46.Bxe1 Qf3 47.Qf2 Qd3 48.Bc3 Kh7 49.Bd2 Bd7 50.Qf4 Qc2 51.Qd4 Qb1+ 52.Kf2 Bb5 53.g4 Qf1+ 54.Kg3 Bc6 55.Qf2 Qd3+ 56.Be3 Kg8 57.h4 Qe4 58.Bd4 Qd5 59.Bc5 h5 60.gxh5 Qxh5 61.Kh2 Qg4 62.Bd4 Bd7 63.Bc3 Qh3+ 64.Kg1 Bc6 65.Qh2 Qf3 66.Be1 Qg4+ 67.Qg3 Qd4+ 68.Qf2 Qd1 69.Kh2 Qd5 70.Bc3 Qc4 71.Qe1 Qf4+ 72.Qg3 Qf1 73.Qg1 Qe2+ 74.Kg3 Qf3+ 75.Kh2 Bd5 76.Be1 Qe2+ 77.Qf2 Qe5+ 78.Kg1 Qe4 79.Qg3 Bc6 80.Bc3 Kh7 81.Kf2 Bb5 82.Qh3 Qf4+ 83.Ke1 Bc4 84.Bd2 Qe4+ 85.Kd1 Bf7 86.Bc3 Bh5+ 87.Kc1 Qe2 88.Qd7 Bg6 89.Qd2 Qe4 90.Qf2 Qb1+ 91.Kd2 Qd3+ 92.Ke1 Bh5 93.Qg2 Qc4 94.Qc2+ Kg8 95.Qf2 Qd3 96.Qg2 Qe3+ 97.Kf1 Kh7 98.Qf2 Qh3+ 99.Kg1 Be8 100.Qf4 Bc6 1/2-1/2

Final position was:

[fen]8/6pk/p1b2p2/P7/1P3Q1P/2B4q/1P6/6K1 w - - 80 101[/fen]

It's a shame for the MCGP because at move 76 there was a mate in 6 but it was not able to see far enough ahead for that:

Mate in 6 for black:
[fen]6k1/6p1/p4p2/P2b4/1P5P/5q2/1P5K/4B1Q1 b - - 31 76[/fen]

So after three games the MCGP and SD are level with 1.5 points each. The SD is definitely putting up a good fight.

What will game 4 have in store?

Regards
Ray

[Cubeman]

Hello Ray. If MCG Pro failed to ''see'' a checkmate win for Black at move#76 in Game#3 at 40 moves/2 hours time control, that indicates a major flaw in MCG Pro's programming because any strong chess computer should easily be able to calculate a forced checkmate in 6 moves even at a short time control such as 60 seconds per move............Best Wishes, Mike.

To be fair to the MCG the forced mate in 6 is not a typical sequence of checks then mate. The winning move is 1..g5! I tested some Mobile devices and PF4 with Hiarcs 13.1 took 35 seconds and Chess Genius on Nokia Lumia 920 took 92 seconds to come up with mate. The position makes quite a good test option for dedicated computers as I bet many strong units will struggle to find the mate in a short time.

[Yarc]

Hello Ray. If MCG Pro failed to ''see'' a checkmate win for Black at move#76 in Game#3 at 40 moves/2 hours time control, that indicates a major flaw in MCG Pro's programming because any strong chess computer should easily be able to calculate a forced checkmate in 6 moves even at a short time control such as 60 seconds per move............Best Wishes, Mike.

To be fair to the MCG the forced mate in 6 is not a typical sequence of checks then mate. The winning move is 1..g5! I tested some Mobile devices and PF4 with Hiarcs 13.1 took 35 seconds and Chess Genius on Nokia Lumia 920 took 92 seconds to come up with mate. The position makes quite a good test option for dedicated computers as I bet many strong units will struggle to find the mate in a short time.

As this was on move 76 it was getting close to the required 40 moves in 2 hours, so we must also consider what time was left on the clock. At this point in the game, both the MCGP and SD were only reaching 9 ply in their calculations so they would have needed to extend their search by another 3 ply to find it. Ill try the position without time constraints to see what happens, but I agree with Cubeman it's not a simple mate for lower powered computers.

Regards
Ray

[swarm]

The result of game 3 was a draw by adjudication. After 90 moves no progress was being made therefore decided to let the computers play until 100 moves and then adjudicate if required.

They followed the Ruy Lopez, Marshall, main line and the SD was out of book at move 14 whilst the MCGP was out of book at move 16. I use ChessGenius3 on my mobile to store the games and this showed the incorrect book opening name, it had English Symmetrical.


Here is the game:

[Event "40m/2h H book"]
[Date "2016.09.26"]
[Round "3"]
[White "SD"]
[Black "MCGP"]
[Result "1/2-1/2"]
[ECO "C89"]

1.e4 e5 2.Nf3 Nc6 3.Bb5 a6 4.Ba4 Nf6 5.O-O Be7 6.Re1 b5 7.Bb3 O-O 8.c3 d5 9.exd5 Nxd5 10.Nxe5 Nxe5 11.Rxe5 c6 12.d4 Bd6 13.Re2 Qh4 14.g3 Qh3 15.Nd2 Bf5 16.Ne4 Be7 17.f3 Qh5 18.Qd3 Rfd8 19.a4 Qg6 20.Bd2 Rdb8 21.Qc2 Re8 22.Rae1 Red8 23.a5 c5 24.Bxd5 Rxd5 25.dxc5 Rxc5 26.Qb3 Be6 27.Qb4 Rc7 28.Qd4 Rd7 29.Qe3 Qh5 30.Bc1 b4 31.Ng5 Bxg5 32.Qxg5 Qxf3 33.cxb4 Rb8 34.Qc5 Qb7 35.Bd2 Rc8 36.Qe5 Rd5 37.Qe3 Rcd8 38.Bc3 h6 39.Rd2 Rxd2 40.Bxd2 Bh3 41.Qf2 f6 42.Re3 Bg4 43.Bc3 Bh3 44.Qc2 Rd1+ 45.Re1 Rxe1+ 46.Bxe1 Qf3 47.Qf2 Qd3 48.Bc3 Kh7 49.Bd2 Bd7 50.Qf4 Qc2 51.Qd4 Qb1+ 52.Kf2 Bb5 53.g4 Qf1+ 54.Kg3 Bc6 55.Qf2 Qd3+ 56.Be3 Kg8 57.h4 Qe4 58.Bd4 Qd5 59.Bc5 h5 60.gxh5 Qxh5 61.Kh2 Qg4 62.Bd4 Bd7 63.Bc3 Qh3+ 64.Kg1 Bc6 65.Qh2 Qf3 66.Be1 Qg4+ 67.Qg3 Qd4+ 68.Qf2 Qd1 69.Kh2 Qd5 70.Bc3 Qc4 71.Qe1 Qf4+ 72.Qg3 Qf1 73.Qg1 Qe2+ 74.Kg3 Qf3+ 75.Kh2 Bd5 76.Be1 Qe2+ 77.Qf2 Qe5+ 78.Kg1 Qe4 79.Qg3 Bc6 80.Bc3 Kh7 81.Kf2 Bb5 82.Qh3 Qf4+ 83.Ke1 Bc4 84.Bd2 Qe4+ 85.Kd1 Bf7 86.Bc3 Bh5+ 87.Kc1 Qe2 88.Qd7 Bg6 89.Qd2 Qe4 90.Qf2 Qb1+ 91.Kd2 Qd3+ 92.Ke1 Bh5 93.Qg2 Qc4 94.Qc2+ Kg8 95.Qf2 Qd3 96.Qg2 Qe3+ 97.Kf1 Kh7 98.Qf2 Qh3+ 99.Kg1 Be8 100.Qf4 Bc6 1/2-1/2

Final position was:

[fen]8/6pk/p1b2p2/P7/1P3Q1P/2B4q/1P6/6K1 w - - 80 101[/fen]

It's a shame for the MCGP because at move 76 there was a mate in 6 but it was not able to see far enough ahead for that:

Mate in 6 for black:
[fen]6k1/6p1/p4p2/P2b4/1P5P/5q2/1P5K/4B1Q1 b - - 31 76[/fen]

So after three games the MCGP and SD are level with 1.5 points each. The SD is definitely putting up a good fight.

What will game 4 have in store?

Regards
Ray

Is that the correct position for mate I tried to solve it could not i usually can solve those and so I put on stockfish has not found it and its been several minutes so far ..maybe we both need some coffee hahaha UP date Stock fish wanted g5 had a score of +.90 I did the move shoes mate n 5 I took move back hit analyse shows mate n 6 now ..idk hahaha OK cleared everything did over it took 11 seconds on my duo core 2.8 laptop .. so not sure but prob take aprox 9 minutes to solve for stockfish 7 at single core 120 mhz ?

[Steve B]

It's a shame for the MCGP because at move 76 there was a mate in 6 but it was not able to see far enough ahead for that:

Mate in 6 for black:
[fen]6k1/6p1/p4p2/P2b4/1P5P/5q2/1P5K/4B1Q1 b - - 31 76[/fen]

Seems to me both computers were a bit overwhelmed by the position
after all....the SD.. although in a bad position . blunders with 76.Be1?? to put itself into the position to be mated in 6 moves
almost any other move was better

there is enough ....

Blame To Go Around Regards
Steve

[Yarc]

Seems to me both computers were a bit overwhelmed by the position
after all....the SD.. although in a bad position . blunders with 76.Be1?? to put itself into the position to be mated in 6 moves
almost any other move was better

there is enough ....

Blame To Go Around Regards
Steve

I agree. Have diverted from starting game 4 to see how long both machines take to solve this mate in analysis mode. The SD solved it in one hour fifty minutes and the MCGP solved it in one hour nine minutes. So MCGP was forty one minutes quicker.

By the way, in analysis mode the MCGP announces mate without having to press INFO.

A tricky position Regards
Ray

[paulwise3]

Although the checkmating sequence of moves at move#76 of the game between MCG Pro and Novag Star Diamond is not a typical checkmating sequence that occurs in most chess games and even though the sequence was deep into the game where each chess computer had not a lot of time left on their game clocks at 40 moves/2 hours time control, the fact remains that MCG Pro operates at a very fast (compared to most dedicated chess computers) 120 MHz clock speed and therefore should have easily calculated that winning sequence since the sequence was only 6 moves deep! Clearly MCG Pro has a major flaw in its' endgame programming especially when one considers that the time control for the game was 40 moves/ 2 hours!!................Mike.

Hi Mike,

6 moves deep means 12 ply, which is a depth that is not easily reached for dedicated chess computers. And you cannot compare the speed to that of a mobile phone or a tablet. A rule of thumb is that for each extra ply-depth you need about a factor 6 or more in speed (or time) to accomplish that. Or you must have a very smart selective search mechanism.

Ply depth regards, Paul

[paulwise3]

Hi Mike,

I guess your laptop is (at least!) about a factor 20 faster then the CG Pro, and Stockfish will use much more hashtable space then the poor 160 mb the cgp has. So...

I think you made my point
Regards, Paul

[ricard60]

Although the checkmating sequence of moves at move#76 of the game between MCG Pro and Novag Star Diamond is not a typical checkmating sequence that occurs in most chess games and even though the sequence was deep into the game where each chess computer had not a lot of time left on their game clocks at 40 moves/2 hours time control, the fact remains that MCG Pro operates at a very fast (compared to most dedicated chess computers) 120 MHz clock speed and therefore should have easily calculated that winning sequence since the sequence was only 6 moves deep! Clearly MCG Pro has a major flaw in its' endgame programming especially when one considers that the time control for the game was 40 moves/ 2 hours!!................Mike.

This could be a time management problem. When that position of checkmating came up what time was left on the machine?

Checking clocks regards
Ricardo