The Romi vs. Hamsters 0.2 saga almost over! :)

[Michael Sherwin]

RomiChessX64NG8 - Hamsters0.2 : 75.0/100 65-15-20 (=11111111=001101111=1111=1010101111=01110110011=11=11111==11=11==0111111=1111=1110=1=11=101=110=011=) 75% +191

Just installed Windows Vista 64 bit and it works very well, except that I have no sound. I will just buy a sound card that has a Vista 64 bit driver, when I find one. Arena (1.1) 32 bit runs just fine under 64 bit Vista and works perfectly with Romi x64. Hamsters 0.2 also seems to run just fine in this enviroment and if anything Hamsters node rate is improved. Arena loads faster and the chess engines that Arena starts, also load faster.

Well, this version of Romi has a small change that seems to have resulted in a stronger engine. And anyone that has been following the Romi vs. Hamsters 0.2 saga will remember that my goal for Romi against Hamsters 0.2 is 70%. Howevr, that goal is for the 32 bit version of Romi, so with the 64 bit Romi scoring 75% the 32 bit Romi may already be there. I just have a couple more test to make with the 64 bit version of Romi then I will see if the 32 bit version can score 70%. Then I will send both the 32 bit and 64 bit versions to Swami for posting!

[Alessandro Scotti]

Congrats Michael... I think 32-bit version will have no problems reaching the 70% goal with a good 5% margin!
I hope you will then choose another Hamsters for your tests, they like playing Romi a lot!

[Michael Sherwin]

Congrats Michael... I think 32-bit version will have no problems reaching the 70% goal with a good 5% margin!
I hope you will then choose another Hamsters for your tests, they like playing Romi a lot!

Thanks Alessandro! I am looking forward to using a new version of Hamsters to test against, very soon.

The 32 bit version of Romi was not quite there yet. Then after a lot of difficult tuning Romi 32 did manage to get a new record.

RomiChessNG8 - Hamsters0.2 : 69.5/100 58-19-23 (1111101=0=11111110====10111001111=1101=111110=10=10=101==1111==1111=111==01=0=0011=11=1110011=101110) 70% +147

My kingdom for half a point. Arena reports 70%, however, 69.5 is not 70 and therefore the saga continues. All Romi had to do was draw the last game and this phase of my work would have been finished. But no, Romi lost it rather quickly just before sending Fedor a wink. I'm going to have to keep a closer watch on those two!

Before, writing this version off as failing the 70% goal I am going to replay the match several more times to see if Romi used all of her luck or not. Also I told Prof. Hyatt that I would repeat a match a few times and post the results in the programming forum so that the luck factor can be demonstrated. I will compile a 64 bit version and do the same. Look for that, as that may prove interesting.

[Tony Thomas]

I am pretty sure that the latest Hamsters can score more than 50% against Romi.

[Michael Sherwin]

I am pretty sure that the latest Hamsters can score more than 50% against Romi.

You are probably right Tony and the new Hamsters is better than the latest RomiChess beta, but here is a little tease to show that it may still be interesting. The first two games of the standard hundred game match at 10+10 both ended in Romi's favor. Take note of Romi's 123rd move in game two!

[Event "Arena tournament"]
[Site "A13"]
[Date "2007.09.22"]
[Round "1"]
[White "Romi695x32"]
[Black "Hamsters5"]
[Result "1-0"]
[ECO "A70"]
[Opening "Benoni"]
[Time "22:56:39"]
[Variation "Classical (without 9.O-O)"]
[TimeControl "600+10"]
[Termination "normal"]
[PlyCount "115"]
[WhiteType "program"]
[BlackType "program"]

1.d4 Nf6 2.c4 c5 3.d5 e6 4.Nc3 exd5 5.cxd5 d6 6.e4 g6 7.Nf3 Bg7 8.Bd3 Na6
9.a3 O-O 10.O-O Bd7 11.Bf4 Qb6 12.Rb1 Nh5 13.Be3 Rfe8 14.Nd2 Rab8 15.Be2
Nf6 16.Bf4 Bf8 17.Qc2 Qc7 18.Rfc1 b5 19.b4 c4 20.h3 Bg7 21.Qd1 Bf8 22.Kh1
Qb6 23.Be3 Qc7 24.Qc2 Bg7 25.Nf3 Re7 26.Nd4 h5 27.Kh2 h4 28.Kg1 Ree8 29.
Rd1 Nh7 30.Rbc1 Qb7 31.Nc6 Bxc6 32.dxc6 Qxc6 33.Nxb5 Qxb5 34.Bxc4 Nxb4 35.
axb4 Qe5 36.Qa4 Re7 37.Bd5 Nf6 38.Bxa7 Rd8 39.Bb6 Rf8 40.Bd4 Qg5 41.Bxf6
Qxf6 42.b5 Qb2 43.Rb1 Qc3 44.b6 Qc8 45.Rdc1 Qb8 46.Rc4 Bf6 47.Qc2 Kg7 48.
b7 Be5 49.Rc8 Rxb7 50.Bxb7 Rxc8 51.Qxc8 Qxc8 52.Bxc8 Kf6 53.Rb7 Bf4 54.f3
Be5 55.Kf2 Bf4 56.Ke2 Bc1 57.Kd3 Bh6 1-0

[Event "Arena tournament"]
[Site "A13"]
[Date "2007.09.22"]
[Round "2"]
[White "Hamsters5"]
[Black "Romi695x32"]
[Result "0-1"]
[ECO "A70"]
[Opening "Benoni"]
[Time "23:33:24"]
[Variation "Classical (without 9.O-O)"]
[TimeControl "600+10"]
[Termination "normal"]
[PlyCount "294"]
[WhiteType "program"]
[BlackType "program"]

1.d4 Nf6 2.c4 c5 3.d5 e6 4.Nc3 exd5 5.cxd5 d6 6.e4 g6 7.Nf3 Bg7 8.Bd3 O-O
9.O-O Bg4 10.Qb3 Nbd7 11.Qxb7 a5 12.Qb3 Bxf3 13.gxf3 Nh5 14.Bb5 Ne5 15.Ne2
Qh4 16.Ng3 Rfb8 17.a4 Nf4 18.Bxf4 Qxf4 19.Ra3 Bh6 20.Qe3 Qh4 21.Qc3 Bf4
22.Rb1 h5 23.Qc2 Kh7 24.Rd1 Qh3 25.Rc3 h4 26.Bf1 Qd7 27.Ne2 Bh6 28.Bg2 g5
29.Ra3 Rb4 30.Rb1 Rab8 31.h3 Bg7 32.Ra2 Ng6 33.Kh1 Be5 34.Kg1 Qb7 35.Nc3
Rc4 36.Rc1 Nf4 37.Bf1 Qd7 38.Kh1 Rcb4 39.Nd1 Nxh3 40.Rb1 Nf4 41.Ne3 f6 42.
Nf5 Kh8 43.Bb5 Qh7 44.Rg1 h3 45.Rg4 Kg8 46.Ne3 Qg6 47.Nc4 Kg7 48.b3 Ra8
49.Qd1 Ra7 50.Nxe5 fxe5 51.Bc4 Kh7 52.Qg1 Kh6 53.Rc2 Qf6 54.Kh2 Rh7 55.Rc3
Rb8 56.Bb5 Kg6 57.Rc2 Qd8 58.Rc1 Rh5 59.Qd1 Rb7 60.Qd2 Rbh7 61.Rh1 Kg7 62.
Qc3 Kf7 63.Qd2 Kg6 64.Qe3 Rg7 65.Bf1 Qb8 66.Bb5 Kh7 67.Rg3 Qd8 68.Kg1 Rb7
69.Rh2 Kg8 70.Qd2 Rbh7 71.Bf1 Rh8 72.Qc1 Qb8 73.Qe1 Qb4 74.Qxb4 cxb4 75.
Bb5 Kf7 76.Kf1 Rc8 77.Bc6 Rc7 78.Ke1 Kf6 79.Rg1 Rh8 80.Kd2 Rh4 81.Rgh1 Rh6
82.Ke3 Rh8 83.Rc1 Rcc8 84.Rb1 Rh5 85.Rg1 Rc7 86.Rb1 Rhh7 87.Kd2 Rc8 88.Rg1
Rh4 89.Rb1 Ke7 90.Rg1 Rh5 91.Rc1 Kf7 92.Rg1 Rc7 93.Kd1 Ra7 94.Kd2 Ke7 95.
Bb5 Rc7 96.Bc6 Kf8 97.Rc1 Kg8 98.Rg1 Kg7 99.Kd1 Kf7 100.Kd2 Rc8 101.Rc1
Rh7 102.Rg1 Kf6 103.Rb1 Rcc7 104.Rg1 Rh8 105.Rc1 Ke7 106.Rg1 Rh5 107.Ke3
Kd8 108.Kd2 Kc8 109.Rg3 Re7 110.Bb5 Kd8 111.Rg4 Rc7 112.Bc6 Rc8 113.Rg1
Kc7 114.Kd1 Rg8 115.Bb5 Rg7 116.Kc2 Kb8 117.Rg3 Rc7+ 118.Bc6 Re7 119.Bb5
Kc7 120.Rg1 Kd8 121.Kd2 Rg7 122.Ke3 Rc7 123.Bc4 Rxc4 124.bxc4 Kc7 125.Kd2
Kb6 126.Kc2 Kc5 127.Kb3 Rh6 128.Rxg5 Ne2 129.Kb2 Kxc4 130.Rg7 Nc3 131.Rc7+
Kd3 132.Kb3 Ne2 133.Rf7 Nd4+ 134.Kb2 Ke2 135.f4 exf4 136.Rxf4 Nf3 137.Rh1
h2 138.e5 Nxe5 139.Kc2 Rh3 140.Rf6 b3+ 141.Kb1 Nd3 142.Rxd6 Nxf2 143.Re6+
Kd2 144.Ree1 Nxh1 145.Rxh1 Kc3 146.Rc1+ Kb4 147.Rh1 0-1

[Alessandro Scotti]

The new Hamsters is a strange beast. It is very similar to 0.4 but I stumbled upon a set of parameters that noticeably improved performance in my tests. I run a couple of thousands games and noticed uneven results, but overall they seem consistently better than 0.4. For example, it came ahead of Pseudo in the last two Noomen matches and that never happened before. Yet there are definitely dark areas that need to be addressed, and it's currently behind 0.4 in CCRL blitz. If in the long run 0.5 doesn't come at least 30-40 elo ahead of 0.4 I'll have to revisit my test and add some more opponents to the mix.