which would take more moves to win

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
duncan
Posts: 11495
Joined: Mon Jul 07, 2008 8:50 pm

which would take more moves to win

Post by duncan » Wed Feb 03, 2016 10:39 pm

a top gm playing komodo with knight handicap

or 2 computers one with a knight handicap but both with access to a 32 piece tablebase

and how many moves is the former ?

bob
Posts: 20916
Joined: Mon Feb 27, 2006 6:30 pm
Location: Birmingham, AL

Re: which would take more moves to win

Post by bob » Thu Feb 04, 2016 7:11 pm

duncan wrote:a top gm playing komodo with knight handicap

or 2 computers one with a knight handicap but both with access to a 32 piece tablebase

and how many moves is the former ?
Clearly the latter, since they would, by definition, play perfectly. I think there is little doubt that the GM would squander much of that knight advantage via mistakes.

I'd be surprised if the distance to mate for the knight-odds game is not close to 4 digits if not more with perfect play.

jhellis3
Posts: 460
Joined: Fri Aug 16, 2013 10:36 pm

Re: which would take more moves to win

Post by jhellis3 » Thu Feb 04, 2016 8:03 pm

Clearly the former, since the result of the latter is already known.

The position would either be known as a draw or loss and adjudicated immediately (given any sort of sense by the operator(s)).

syzygy
Posts: 4600
Joined: Tue Feb 28, 2012 10:56 pm

Re: which would take more moves to win

Post by syzygy » Fri Feb 05, 2016 12:13 am

jhellis3 wrote:Clearly the former, since the result of the latter is already known.

The position would either be known as a draw or loss and adjudicated immediately (given any sort of sense by the operator(s)).
The two engine operators might keep it a secret for each other that they have access to 32-piece TBs (including the relevant 31-piece TB).

Much will then depend on what information the TBs actually store (WDL? WDL50/DTZ50? DTM?) and on how the engines implement swindle mode.

Even if both sides are aware that they both use TBs, the losing side might still hope he can trick a "perfect" DTM player into a 50-move draw and therefore play on. Of course this is useless if the winning side has access to DTZ50.

Uri Blass
Posts: 8771
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: which would take more moves to win

Post by Uri Blass » Fri Feb 05, 2016 1:21 am

bob wrote:
duncan wrote:a top gm playing komodo with knight handicap

or 2 computers one with a knight handicap but both with access to a 32 piece tablebase

and how many moves is the former ?
Clearly the latter, since they would, by definition, play perfectly. I think there is little doubt that the GM would squander much of that knight advantage via mistakes.

I'd be surprised if the distance to mate for the knight-odds game is not close to 4 digits if not more with perfect play.
I will be surprised if the distance to mate for knight handicap game is 3 digit (more than 99 moves).

I see no reason to believe that the distance to mate is more than the number of moves that computers need to mate today when both of them are not perfect.

bob
Posts: 20916
Joined: Mon Feb 27, 2006 6:30 pm
Location: Birmingham, AL

Re: which would take more moves to win

Post by bob » Fri Feb 05, 2016 4:53 am

Uri Blass wrote:
bob wrote:
duncan wrote:a top gm playing komodo with knight handicap

or 2 computers one with a knight handicap but both with access to a 32 piece tablebase

and how many moves is the former ?
Clearly the latter, since they would, by definition, play perfectly. I think there is little doubt that the GM would squander much of that knight advantage via mistakes.

I'd be surprised if the distance to mate for the knight-odds game is not close to 4 digits if not more with perfect play.


I will be surprised if the distance to mate for knight handicap game is 3 digit (more than 99 moves).

I see no reason to believe that the distance to mate is more than the number of moves that computers need to mate today when both of them are not perfect.
I can think of a thousand reasons why that is not true. We don't even know that a knight advantage is a forced win. If one can force a trade to KRN vs KR for example...

It won't be a short game in any case.

User avatar
Laskos
Posts: 10339
Joined: Wed Jul 26, 2006 8:21 pm
Full name: Kai Laskos

Re: which would take more moves to win

Post by Laskos » Fri Feb 05, 2016 7:41 am

bob wrote:I'd be surprised if the distance to mate for the knight-odds game is not close to 4 digits if not more with perfect play.
I would be very surprised if the distance is close to 4 digits. In fact, it seems a reasonable guess it to be closer to 10^2 than to 10^3.
bob wrote: I can think of a thousand reasons why that is not true. We don't even know that a knight advantage is a forced win. If one can force a trade to KRN vs KR for example...

It won't be a short game in any case.
Well, the insights are quite conclusive. Going from ultra-short time controls with weak engines to ultra-long time controls with strong engines, the following phenomena can be observed:

1/ Standard opening position -- draw rate increases
2/ Knight advantage -- win rate increases

If the perfect chess is just a continuation of the path from very low quality games to very high quality games, the perfect chess is:

1/ Standard opening position -- draw
2/ Knight advantage -- win

Do you ever use such things like "common sense", "intuition", "meaningful extrapolation", "empirical data"?

Uri Blass
Posts: 8771
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: which would take more moves to win

Post by Uri Blass » Fri Feb 05, 2016 2:18 pm

bob wrote:
Uri Blass wrote:
bob wrote:
duncan wrote:a top gm playing komodo with knight handicap

or 2 computers one with a knight handicap but both with access to a 32 piece tablebase

and how many moves is the former ?
Clearly the latter, since they would, by definition, play perfectly. I think there is little doubt that the GM would squander much of that knight advantage via mistakes.

I'd be surprised if the distance to mate for the knight-odds game is not close to 4 digits if not more with perfect play.


I will be surprised if the distance to mate for knight handicap game is 3 digit (more than 99 moves).

I see no reason to believe that the distance to mate is more than the number of moves that computers need to mate today when both of them are not perfect.
I can think of a thousand reasons why that is not true. We don't even know that a knight advantage is a forced win. If one can force a trade to KRN vs KR for example...

It won't be a short game in any case.

I do not agree that we do not know that a knight advantage is a forced win.

Maybe you do not know but I know like I know that if I leave a coin in the ear then it is going to fall to the floor and not stay in the air.

In both cases no proof for it but I get conclusions about the future based on the past.

syzygy
Posts: 4600
Joined: Tue Feb 28, 2012 10:56 pm

Re: which would take more moves to win

Post by syzygy » Sat Feb 06, 2016 1:21 pm

Uri Blass wrote:Maybe you do not know but I know like I know that if I leave a coin in the ear then it is going to fall to the floor and not stay in the air.
That depends on the shape of the ear. (I know, you made a typo :-))

I fully agree that perfect DTM-play from both sides in a knight odds game will most likely take fewer moves to reach mate than a typical knight odds game between humans and/or engines.

Bob seems to think that the losing side is somehow likely to find amazing defensive resources that enable it to postpone the inevitable by hundreds of moves. But there is no rational basis for such expectation. The winning side starts with a huge advantage that it can and will use to increase its advantage at every move. The winning side will have so many more options at its disposal than the losing side that it is sure to find a mating sequence that is much shorter than one would see in a game between fallible opponents.

duncan
Posts: 11495
Joined: Mon Jul 07, 2008 8:50 pm

Re: which would take more moves to win

Post by duncan » Sun Feb 07, 2016 2:17 am

bob wrote:
duncan wrote:a top gm playing komodo with knight handicap

or 2 computers one with a knight handicap but both with access to a 32 piece tablebase

and how many moves is the former ?
Clearly the latter, since they would, by definition, play perfectly. I think there is little doubt that the GM would squander much of that knight advantage via mistakes.

I'd be surprised if the distance to mate for the knight-odds game is not close to 4 digits if not more with perfect play.
and what would your estimate be how long it takes a 32 piece database to defeat rook and queen handicap ?

Post Reply