Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Unique nodes from the start position

Author: Heiner Marxen

Date: 13:06:52 06/29/00

Go up one level in this thread


On June 29, 2000 at 14:03:32, Robert Hyatt wrote:

>On June 29, 2000 at 11:17:20, blass uri wrote:
>
>>On June 29, 2000 at 10:54:38, Andreas Stabel wrote:
>>
>>>On June 29, 2000 at 09:11:47, Robert Hyatt wrote:
>>>
>>>>On June 29, 2000 at 08:16:25, Andreas Stabel wrote:
>>>>
>>>>>I have made a program that calculates this and here is the result to ply 7.
>>>>>
>>>>>   |               | Unique nodes  | Unique nodes | Unique nodes  | Factor |
>>>>>Ply| Total # nodes | ep = pawn two | ep = opposite| ep = Only if  | prev.  |
>>>>>   |               |               | pawn can hit | ep is legal   | row    |
>>>>>---+---------------+---------------+--------------+---------------+--------|
>>>>> 0 |             1 |             1 |            1 |             1 |        |
>>>>> 1 |            21 |            21 |           21 |            21 |  21.00 |
>>>>> 2 |           421 |           421 |          421 |           421 |  20.05 |
>>>>> 3 |          9323 |          8023 |         5783 |          5783 |  13.74 |
>>>>> 4 |        206604 |        109262 |        77796 |         77796 |  13.45 |
>>>>> 5 |       5072213 |       1351950 |       898812 |        898812 |  11.55 |
>>>>> 6 |     124132537 |      15334851 |     10281864 |      10281862 |  11.44 |
>>>>> 7 |    3320034397 |     160373323 |    106193912 |     106193643 |  10.33 |
>>>>> 8 |   88319013353 |               |              |               |        |
>>>>> 9 | 2527849247520 |               |              |               |        |

Hey, thanks very much for these numbers, Andreas!  I had plans to do such an
analysis myself, some time.


>>>>>It is interesting to note how great the reduction is just by not
>>>>>setting the E.P. target square if there is not pawn to hit. This simple
>>>>>test reduses the number of unique positions by a third !
>>>>
>>>>
>>>>How is this happening?  If no ep is possible, how can you get more nodes when
>>>>you set the ep target, even though no pawn can make the capture???
>>>>
>>>
>>>Two positions will be different with and without the E.P. target square set
>>>because they will get different hash signatures. So if a chess program first
>>>searches one and then later in the search encounter the other it will have
>>>to search again because it will not realize that it is in fact functionally
>>>the same position. If you just chech each time a pawn is move two forward, if
>>>there exist a pawn of the oposite colour which may hit this pawn before you
>>>set the E.P. target square you will avoid this.
>>
>>I understand that you say that chess programs have different hash signature for
>>the position after 1.Nf3 Nf6 2.d4 and the position after 1.d4 Nf6 2.Nf3 because
>>they assume in the first position that the d4 may be captured next move.
>>
>>I believe that it is correct only for part of the chess programs.
>>I guess that the time that programs waste is bigger than the time that they save
>>by calculating the hash signature faster but I have no experience with it so I
>>may be wrong.
>>
>>It is not surprising that there are many positions when pawns get forward 2
>>squares and inspite of that they cannot be captured by the en passent rule and
>>it is not surprising that the starting position has many cases like this case
>>because there are many pawns that can move 2 squares.
>>
>>Uri
>
>
>All I can say is that I do this correctly internally.  I only set the ep flag
>if a pawn moves two squares _and_ it can actually be captured by an opposing
>pawn.

Yes, I do the same in Chest (I check for pseudo legality, not true legality,
i.e. I just check whether an opposing pawn is in the right place).

It may be noteworthy, that FEN and EPD require to indicate an e.p. target
for every double step of a pawn, regardless of opposing pawns being in
an appropriate position.  I consider that a bug in FEN/EPD.

Heiner



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.