Ares 1.005 available

[CRoberson]

After a long break, I've been working on Ares again. The new version is improved and is released as a full 64 bit binary instead of 32 bit for Ares 1.004. My tests indicate about a 200 Elo improvement over Ares 1.004.

I have a website for it. The licensing is at the bottom and is simple. You can use it, but not copy, reverse engineer, modify or any other copy method. I don't allow redistribution either.

The site is http://aigames.net/AresChess1-005.html

There are two verison:
1) 64 bit with popcnt
2) Plain 64 bit.

If anybody needs a 32 bit version, let me know. If anybody needs a Linux version, let me know.

Charles

[Graham Banks]

Excellent news. Thanks Charles.

[AdminX]

After a long break, I've been working on Ares again. The new version is improved and is released as a full 64 bit binary instead of 32 bit for Ares 1.004. My tests indicate about a 200 Elo improvement over Ares 1.004.

I have a website for it. The licensing is at the bottom and is simple. You can use it, but not copy, reverse engineer, modify or any other copy method. I don't allow redistribution either.

The site is http://aigames.net/AresChess1-005.html

There are two verison:
1) 64 bit with popcnt
2) Plain 64 bit.

If anybody needs a 32 bit version, let me know. If anybody needs a Linux version, let me know.

Charles

Thanks Charles, what opening books do you recommend?

[Ferdy]

Thanks for sharing.
It seems there is a problem in displaying pv, an illegal move is shown at the end of pv line. This is in Arena 3.5.

Code: Select all

702613 >first : ucinewgame
702613 >first : isready
702645 <first : readyok
702660 >first : position startpos moves d2d4
702660 >first : go wtime 300000 btime 300000 winc 0 binc 0 movestogo 40
702660 <first : info depth 1
702660 <first : info depth 1 score cp -54 nodes 2 pv b8a6 
702660 <first : info depth 1 score cp -31 nodes 4 pv b8c6 
702676 <first : info depth 1 score cp 0 nodes 22 pv d7d5 
702676 <first : info depth 2
702676 <first : info depth 2 score cp -62 nodes 97 pv d7d5 a1a1 
702676*moveerror* start piece<0 and target piece<0 !...!
702676 <first : info depth 3
702691 <first : info depth 3 score cp 0 nodes 327 pv d7d5 g1f3 a1a1 
702691*moveerror* start piece<0 and target piece<0 !...!
702691 <first : info depth 4
702691 <first : info depth 4 score cp -61 nodes 648 pv d7d5 g1f3 g8f6 a1a1 
702691*moveerror* start piece<0 and target piece<0 !...!
702707 <first : info depth 5
702707 <first : info depth 5 score cp 0 nodes 4812 pv d7d5 g1f3 g8f6 b1c3 a1a1 
702707*moveerror* start piece<0 and target piece<0 !...!
702707 <first : info depth 6
702707 <first : info depth 6 score cp -48 nodes 5909 pv d7d5 g1f3 g8f6 b1c3 b8c6 a1a1 
702707*moveerror* start piece<0 and target piece<0 !...!
702723 <first : info depth 7
702723 <first : info depth 7 score cp -4 nodes 22049 pv d7d5 c1f4 c8f5 b1c3 b8c6 c3b5 a1a1 

Also the UCI_ShowCurrLine is not working although it is shown in the engine option list. No info currline is sent to gui. This can be seen below engine window in arena if this working.

Code: Select all

option name UCI_ShowCurrLine type check default true

[CRoberson]

Ted,

What are you doing awake? You are in the same time zone I am in; you are supposed to be sleeping! The wife will be jealous.

I don't know for this version. I've always tried to make my programs more positional and better at closed positions: starting with d4 or c4. This version has had quite a bit of eval tuning, so I am not sure what is best.

I tried some online testing against Rookie with my old book and another match with Rookie using a more varied book (mostly e4) that I made from the CCRL database and Ares performed a bit better with the old d4 book.

Charles

[AdminX]

Ted,

What are you doing awake? You are in the same time zone I am in; you are supposed to be sleeping! The wife will be jealous.

I don't know for this version. I've always tried to make my programs more positional and better at closed positions: starting with d4 or c4. This version has had quite a bit of eval tuning, so I am not sure what is best.

I tried some online testing against Rookie with my old book and another match with Rookie using a more varied book (mostly e4) that I made from the CCRL database and Ares performed a bit better with the old d4 book.

Charles

Thanks Charles, this time the wife has to blame our two year old son.

[Ferdy]

One more observation, when go infinite is sent, it sends a weird nps value.

2534205 >first : ucinewgame
2534205 >first : isready
2534237 <first : readyok
2534252 >first : position startpos
2534252 >first : go infinite
2534252**Time for starting analysis:78
2534252 <first : info nps -2147483648
2534268 <first : info time 0
2534268 <first : info depth 1
2534268 <first : info depth 1 score cp 39 time 0 nodes 2 pv b1a3
2534268 <first : info depth 1 score cp 62 time 0 nodes 4 pv b1c3
2534283 <first : info depth 1 score cp 94 time 0 nodes 22 pv d2d4
2534283 <first : info depth 1 score cp 100 time 0 nodes 26 pv e2e4
2534299 <first : info depth 1 score cp 100 time 0 nodes 40 pv e2e4
2534299 <first : info nps -2147483648
2534299 <first : info time 0
2534315 <first : info depth 2
2534315 <first : info depth 2 score cp 0 time 0 nodes 85 pv e2e4 a1a1

[CRoberson]

I tested mostly in Fritz and SCID. I did some gauntlet testing in Arena 3.0 and lots of testing in cutechess-cli. Haven't tried Arena 3.5.

I see the odd nps value. It happens up until 1 seconds has elapsed. Then it is fine. The trailing pv a1a1 didn't bother the other interfaces, but you'd think it should if a variation board is turned on. I'll fix that and I'll download Arena 3.5 and try it.

[CRoberson]

This is one of the best news recently. Thank you, Charles.

I recall you used to test NoonianChess. I have a new version of that too. It needs more testing and I am still modifying it.

[Dr.Wael Deeb]

Thanks for the new release

What is the estimated Elo strength of the program according to CCRL rating list
Dr.D