CCC Practice Game

Discussion of computer chess matches and engine tournaments.

Moderator: Ras

CRoberson
Posts: 2094
Joined: Mon Mar 13, 2006 2:31 am
Location: North Carolina, USA

Re: CCC Practice Game

Post by CRoberson »

It seems to me that some of the negatives of a simple voting system
have been revealed. It seems ok to do this while in book, but not
when out of book.

Here is my suggestion for our decision making during the out of
book phase. We use ourselves and computers as one big virtual
distributed super computer. We can start trying this out in the
practice game.

Here is how it works:
1) The team receive a position (move made by opponent)

2) Captain assigns moves to analyze to each member.

3) Each member takes his move(s) and runs a very long
analysis with his computer(s).

4) Members report the programs resulting value back to the
forum. This should be done half way to our decision time limit.
However the team members should keep the analysis running.

5) We look over the results and decide on which are the best
two to four moves if not the best one. Then all go an analyze
this moves or the ones they are assigned (interest level can
be used on assigning them).

6) New results and analysis for each move are presented.

7) Once we have read all the analysis then we vote (not before).

8) Make our move.


# 7 solves the problem of people voting and saying later that
they should have considered....

Using our computers this way will add 1 ply of search to what ever
we do. That 1 ply of search could be worth between 50 and 100 Elo.
Its a guaranteed strength improvement. Basically, we would be
doing distributed computing and splitting at the root.
Steve B
Posts: 3697
Joined: Tue Jul 31, 2007 4:26 pm

Re: CCC Practice Game

Post by Steve B »

In Reply to Charles's suggestion..

while i also think that this would result in superior move recommendations..this would necessitate that the Captains know beforehand which members they can assign moves too
in other words ..a fixed team of members.. which i think we will not have..i imagine that members will come and go..some days vote and some days..not vote
i can also say that the simple voting system is the only system i have witnessed so far in forum correspondence games

however the system you mention can certainly be adopted if a certain minimum number of team players can commit to be here every day

Steve
Steve B
Posts: 3697
Joined: Tue Jul 31, 2007 4:26 pm

Re: Our 4th move is.....

Post by Steve B »

Christopher Conkie wrote:.......cxd4 by (un)popular demand.

Christopher
5.Nxd4
Date "2009.07.12"]
[Round "1"]
[White "Tasc R30(v2.2-1993)"]
[Black "CCC/CTF Forum Team"]
[Result "?"]
(Remove Blacks b Pawn)

1.e4 Bb7 2.Nc3 c5 3.Nf3 e6 4.d4 cxd4 5.Nxd4*
[d] rn1qkbnr/pb1p1ppp/4p3/8/3NP3/2N5/PPP2PPP/R1BQKB1R b KQkq

+.97 Regards
Steve
tano-urayoan
Posts: 638
Joined: Thu Aug 30, 2007 8:23 pm
Location: San Juan, Puerto Rico

Re: CCC Practice Game

Post by tano-urayoan »

CRoberson wrote:It seems to me that some of the negatives of a simple voting system
have been revealed. It seems ok to do this while in book, but not
when out of book.

Here is my suggestion for our decision making during the out of
book phase. We use ourselves and computers as one big virtual
distributed super computer. We can start trying this out in the
practice game.

Here is how it works:
1) The team receive a position (move made by opponent)

2) Captain assigns moves to analyze to each member.

3) Each member takes his move(s) and runs a very long
analysis with his computer(s).

4) Members report the programs resulting value back to the
forum. This should be done half way to our decision time limit.
However the team members should keep the analysis running.

5) We look over the results and decide on which are the best
two to four moves if not the best one. Then all go an analyze
this moves or the ones they are assigned (interest level can
be used on assigning them).

6) New results and analysis for each move are presented.

7) Once we have read all the analysis then we vote (not before).

8) Make our move.


# 7 solves the problem of people voting and saying later that
they should have considered....

Using our computers this way will add 1 ply of search to what ever
we do. That 1 ply of search could be worth between 50 and 100 Elo.
Its a guaranteed strength improvement. Basically, we would be
doing distributed computing and splitting at the root.
Also analysis will vary a lot depending on hardware and software used and the time dedicated to analysis. I just a have a single core computer, so if i want to make a good deep analysis I need to let it run longer thatn smp systems. Another problem is different time zones, as our opponent has already flag us twice, we need to be strict with our time management, and the process you suggest seems lengthy. And of course Steve B suggestion that for this we need permanent players.
Steve B
Posts: 3697
Joined: Tue Jul 31, 2007 4:26 pm

Re: CCC Practice Game

Post by Steve B »

tano-urayoan wrote:
Another problem is different time zones, as our opponent has already flag us twice, we need to be strict with our time management, and the process you suggest seems lengthy. And of course Steve B suggestion that for this we need permanent players.
that was my error Tano
i missed the Fact that Chris already made the official move for the team earlier today
i was waiting for the move to be made in a new post this evening
Game On Regards
Steve
tano-urayoan
Posts: 638
Joined: Thu Aug 30, 2007 8:23 pm
Location: San Juan, Puerto Rico

Re: Move 4

Post by tano-urayoan »

playjunior wrote:
Henrik Dinesen wrote: Or Qb6. if a6 all our moves except Bb7 will have been pawn moves. if Nxd4 a6 we have a kan paulsen type of sicilian again without b pawn. so how could we take advantage or minimize this disadvantage of the missing b pawn?
We shouldn't have played c5?

As of now, I vote for Nf6. We need to complicate the position. After cxd4 Nxd4 Bb4 Nb5 I have trouble seeing what do we do next.
Well we are playing a sicilian without a b pawn and no obvious compensation, so maybe c5 was not the path to choose.

What could we play after Nxd4? a6, Nf6, Bb4, d6,

I have not check with a computer but what happens if Bb4, Ndb5 Qb6?
Henrik Dinesen
Posts: 877
Joined: Wed Mar 08, 2006 9:52 pm
Location: Denmark

Re: Move 4

Post by Henrik Dinesen »

tano-urayoan wrote:
playjunior wrote:
Henrik Dinesen wrote: Or Qb6. if a6 all our moves except Bb7 will have been pawn moves. if Nxd4 a6 we have a kan paulsen type of sicilian again without b pawn. so how could we take advantage or minimize this disadvantage of the missing b pawn?
We shouldn't have played c5?

As of now, I vote for Nf6. We need to complicate the position. After cxd4 Nxd4 Bb4 Nb5 I have trouble seeing what do we do next.
Well we are playing a sicilian without a b pawn and no obvious compensation, so maybe c5 was not the path to choose.

What could we play after Nxd4? a6, Nf6, Bb4, d6,

I have not check with a computer but what happens if Bb4, Ndb5 Qb6?
Then the best answer is Be3.
Henrik
Henrik Dinesen
Posts: 877
Joined: Wed Mar 08, 2006 9:52 pm
Location: Denmark

Re: Our 4th move is.....

Post by Henrik Dinesen »

So, now I think 5.-, Nf6 can be played:

6. Ndb5, a6 7. Nd6+, Bxd6 8. Qxd6 and we can take on e4.
6. e5, Nd5 7. Nxd5, Bxd5

Looks better than 5.-, a6 to me.
Henrik
tano-urayoan
Posts: 638
Joined: Thu Aug 30, 2007 8:23 pm
Location: San Juan, Puerto Rico

Re: Our 5th move is.....

Post by tano-urayoan »

Nf6 maybe just Bd3, what about Bc5? in this systems sometimes the g knight goes to e7 and later g6. the problem with Bc5 should be something like Qg4. And d6 looks pasive but how could we complicate this position?
Henrik Dinesen
Posts: 877
Joined: Wed Mar 08, 2006 9:52 pm
Location: Denmark

Re: CCC Practice Game

Post by Henrik Dinesen »

CRoberson wrote:It seems to me that some of the negatives of a simple voting system
have been revealed. It seems ok to do this while in book, but not
when out of book.

Here is my suggestion for our decision making during the out of
book phase. We use ourselves and computers as one big virtual
distributed super computer. We can start trying this out in the
practice game.

Here is how it works:
1) The team receive a position (move made by opponent)

2) Captain assigns moves to analyze to each member.

3) Each member takes his move(s) and runs a very long
analysis with his computer(s).
Some members will have various tasks for their computer(s) already, and may not be willing to spend that much time on each move.
4) Members report the programs resulting value back to the
forum. This should be done half way to our decision time limit.
However the team members should keep the analysis running.

5) We look over the results and decide on which are the best
two to four moves if not the best one. Then all go an analyze
this moves or the ones they are assigned (interest level can
be used on assigning them).

6) New results and analysis for each move are presented.

7) Once we have read all the analysis then we vote (not before).

8) Make our move.


# 7 solves the problem of people voting and saying later that
they should have considered....

Using our computers this way will add 1 ply of search to what ever
we do. That 1 ply of search could be worth between 50 and 100 Elo.
Its a guaranteed strength improvement. Basically, we would be
doing distributed computing and splitting at the root.
Overall a good suggestion, which in theory I like a lot. But I believe, unfortunately, that Steve's remarks are correct.
On the other hand, if we're able to, at some point, to create a team with fixed members, we should IMO look into an approach like the one you suggest.
Just, hoping that while the use of computers can be extensive, that people still will make use of their minds too. Good suggestions made to the programs may very well end up being the programs prefered line after some time of analysis.
Henrik