Damage does not have (yet) the 7 Piece Endgame Databases

[BertTuyt]

When we do not count the sub-slice databases the 7P database set consists of 76 files in total:

2 - 5 18 ( 18 )
3 - 4 20 ( 8 )
4 - 3 20 ( 8 )
5 - 2 18 ( 18 )

In total I have now calculated 52 sub-databases (see the numbers in brackets).

So the bad news, I did not finish yet [img]images/smilies/icon_sad.gif[/img]
And the good news, the WDL counts for the 52 files are exactly the same as the input provided by Ed [img]images/smilies/icon_smile.gif[/img]

Keep you posted,

Bert

[TAILLE]

When we do not count the sub-slice databases the 7P database set consists of 76 files in total:

2 - 5 18 ( 18 )
3 - 4 20 ( 8 )
4 - 3 20 ( 8 )
5 - 2 18 ( 18 )

In total I have now calculated 52 sub-databases (see the numbers in brackets).

So the bad news, I did not finish yet [img]images/smilies/icon_sad.gif[/img]
And the good news, the WDL counts for the 52 files are exactly the same as the input provided by Ed [img]images/smilies/icon_smile.gif[/img]

Keep you posted,

Bert

Hi Bert,

As I undestand you do not intend (for the moment) to generate the 6-1 slice. I did not either generate this db but I will perhaps change my mind. The point is the following : after having generating the 8 pieces db I hope generating partial 9p db. If your objective is only to generate a partial 5x4 db everything is fine but if you would like to generate a partial 6x3db (especially 6men against 1king and 2men because a sacrifice of 3 men in order to have a kind is rather common) then you need the 6x2db and the 6x1db.
Do you have the intention in the future to generate a partial 6x3db ?

[BertTuyt]

Gerard,

first I focus on getting all 7P databases. Hereafter I need to verify that my faster DB program also works.
Then I will start with the 8P databases, which will keep me busy till mid 2010.
Hereafter I also want to go for the partial 9P databases mainly the 5 - 4 and 6 - 3 versions ( with max. 1 King on both sides).

Bert

[BertTuyt]

Gerard,

forgot to answer your other question, for the short term I will not calculate the 6 - 1 sub-databases.

Bert

[BertTuyt]

The 7p database generation has dramatically slowed down.

As few people know, I had 2 hardware crashes during the tournament.
For what ever reason these crashes occur now every day.
So in the morning i have to restart the PC due to a crash during the night, and when im back from work the same procedure has to be applied.

The positive news is that the time between these crashes is sufficient large that at least a few sub-databases are calculated.
But it doesn't give me confidence for a next tournament.

I think it has something to do with the auto overclock capabilities of my i7, but i just don't know.
And as this processor is rather new, and not widely sold, also the computer-shop doesn't have a clue so far.

I will give it a last try during the weekend, to find out, study all the systems errors as collected by Windows.
Also then I will decide if there is a way out.
If not , i have to cancel the next tournament, as it makes no fun to play with an older version, in stead of my latest engine.

Keep you all posted, all suggestions are welcomed.

Bert

[Ed Gilbert]

Hi Bert,

The i7 may be relatively new but that is not an excuse for your computer to crash. It sounds like you have a hardware problem. I suggest you run some stress test programs such as Prime95, and also a memory test like Memtest86+. If it cannot run these programs without giving memory errors or crashing then this is a strong indication that your computer has a hardware problem and you should get the shop you bought it from to repair or replace it.

-- Ed

[BertTuyt]

Despite the "in between" crashes i'm now at 64 7p databases calculated, out of 76 , so around 84%.

Unfortunately i have encountered a WLD-count difference (for the first time since I started in April) in the 2221w slice ( a little strange as the 2221b slice is correct).

According to Ed the count should be:
2221w: wins 231885784, draws 10071366000, losses 4086903333
Whereas my result was a :
2221w: wins 231885784, draws 10071365999, losses 4086903334

Hope Ed could provide some details in the 81 subslices, so I hope I don't have to redo all work (as I assume my result is wrong, and Ed is right).

Keep you posted,

Bert

[BertTuyt]

With a little help from my friends....................................

Thanks to Ed , we were able to find (in the end) the source for the error count in the DB. Ed wrote something interesting about this "quest", see topic Endgame Databases.

In the mean time I have started the DB generation, and finished the next 2 "DB-clusters" (each containing 81-slices). And to my relief no errors in the WDL count !!!

I have still 8 to go (so 4 pairs of 2), 68 finished out of 76 ( 89%).

Keep you all posted.....

Bert

[BertTuyt]

Today (8-Nov-2009) I finished the last 2 DB's.
So the 7p database work has now ended for me.
I will post, most likley in the cause of the next week, some overall learnings.

Herewith I would like to especially thank Ed for providing the WDL counts for all sub-databases.
And also for the support when i encountered a miscount , which we were able to localize and correct.

To my knowledge there are now 4 people who have generated 7p databases so far:
Michel (Dragon), Ed(Kingsrow), Gerard (Damy), and myself (Damage).

I also will not stop here, so someday i will also go for the next challenge the 8p databases, but not tomorrow .

Bert

[ildjarn]

Today (8-Nov-2009) I finished the last 2 DB's.
To my knowledge there are now 4 people who have generated 7p databases so far:
Michel (Dragon), Ed(Kingsrow), Gerard (Damy), and myself (Damage).

I assume Stef Keetman has generated them too, since he wrote an article (or more?) about 7-piece endgames in Dammen.

[BertTuyt]

Im not sure when Stef wrote these articles, i guess it is in the late 1990 (or beginning 2000) ?
If so, i think it is very difficult with the hardware then available (memory size, hard-disk size, clock speed, 32bit OS) to generate all 7p Databases.
Maybe he generated a few specific DB's, but i assume he didnt generate all.

Do you know the publishing date of these articles?
This might give a clue, about hardware in those days.

Bert

[ildjarn]

The articles 'absolute endgame knowledge' were published in Dammen 131/132, 133/134, 135, 139/140 and 142/143. The last one (142/143) is from 2000, and is about 5v2 endgames. Unfortunately I don't have the articles, so I don't know the exact details.

[TAILLE]

Hi Bert,

Very good Bert. How many disk space do you need for this 7 pieces db ?
As far as I am concerned I do not have the complete db because I did not generate the 6x1 db.

[BertTuyt]

In my "memory" also the latest articles were around 1999-2000.
Hope that someone has a copy , or some details.
Anyway, I really believe that "normal" PC hardware in these days were no capable in calculating all 7p databases.
Im not sure what available memory was around 2000, or hard-disk size, or clock-speed, but my first reaction would be, unsufficient to generate the full 7p DataBase set.

Hope that others can help here.

Bert

[BertTuyt]

Gerard, you are right I have to be more precise.
I calculated only the "interesting" subsets 5-2 and 4-3, but 6-1 is technically not a major problem, and can be generated relatively fast.

I have to check, but as i used raw databases (so no compression, and captures included), the disk space is (but i need to re-check) is somewhere between 300-400 GigaByte (i switched of the i7 computer now, so will verify tomorrow).

In the next weeks i will go through the compression steps, so this size will "dramatically" reduce.

Bert