Adding Transposition caused more nodes to be searched?

[AndrewShort]

my engine, which isn't particularly fast at all (and is far from ready for primetime, lacks many features, etc.), searches depth 26 in less than 1 second (which includes iterations 1 - 26).

Total nodes searched was something like 173,000 (my nodes are not counted if there is a usable hash hit).

In that position, null move is not used in my engine.

My hash replacement scheme is pure hash replace.

Within my engine search, twofold rep is a draw.

Here's my PV from a depth 26 search (note the black f5 pawn is captured on the 19th move of the PV)

1. Ka1-b1
2. Ka7-b8
3. Kb1-c2
4. Kb8-b7
5. Kc2-c3
6. Kb7-c7
7. Kc3-d3
8. Kc7-b7
9. Kd3-e2
10. Kb7-c8
11. Ke2-f3
12. Kc8-c7
13. Kf3-g3
14. Kc7-d8
15. Kg3-h4
16. Kd8-d7
17. Kh4-g5
18. Kd7-e8
19. Kg5xf5
20. Ke8-e7
21. Kf5-g5
22. Ke7-f7
23. f4-f5
24. Kf7-g8
25. f5-f6
26. Kg8-f7

[AndrewShort]

my engine, which isn't particularly fast at all (and is far from ready for primetime, lacks many features, etc.), searches depth 26 in less than 1 second (which includes iterations 1 - 26).

Total nodes searched was something like 173,000 (my nodes are not counted if there is a usable hash hit).

In that position, null move is not used in my engine.

My hash replacement scheme is pure hash replace.

Within my engine search, twofold rep is a draw.

Here's my PV from a depth 26 search (note the black f5 pawn is captured on the 19th move of the PV)

1. Ka1-b1
2. Ka7-b8
3. Kb1-c2
4. Kb8-b7
5. Kc2-c3
6. Kb7-c7
7. Kc3-d3
8. Kc7-b7
9. Kd3-e2
10. Kb7-c8
11. Ke2-f3
12. Kc8-c7
13. Kf3-g3
14. Kc7-d8
15. Kg3-h4
16. Kd8-d7
17. Kh4-g5
18. Kd7-e8
19. Kg5xf5
20. Ke8-e7
21. Kf5-g5
22. Ke7-f7
23. f4-f5
24. Kf7-g8
25. f5-f6
26. Kg8-f7

and I forgot to say my hash table is 128 MB...

[xsadar]

Also is there Fine 1-69 or 71... anywhere ? Where exactly are those from ?

I've always wondered where those came from as well. After a little googling, I found this book:
* Fine, Reuben (1941), Basic Chess Endings, McKay, ISBN 0-679-14002-6
* Fine, Reuben & Benko, Pal (2003), Basic Chess Endings (second ed.), McKay, ISBN 0-8129-3493-8

Is that where they come from?

[jwes]

Is there Fine 1-69 or 71... anywhere ? Where exactly are those from ?

It is from "Basic Chess Endings" by Reuben Fine. Most of the positions are not suitable for tests.

[Dek]

my engine, which isn't particularly fast at all (and is far from ready for primetime, lacks many features, etc.), searches depth 26 in less than 1 second (which includes iterations 1 - 26).

Total nodes searched was something like 173,000 (my nodes are not counted if there is a usable hash hit).

In that position, null move is not used in my engine.

My hash replacement scheme is pure hash replace.

Within my engine search, twofold rep is a draw.

Here's my PV from a depth 26 search (note the black f5 pawn is captured on the 19th move of the PV)

1. Ka1-b1
2. Ka7-b8
3. Kb1-c2
4. Kb8-b7
5. Kc2-c3
6. Kb7-c7
7. Kc3-d3
8. Kc7-b7
9. Kd3-e2
10. Kb7-c8
11. Ke2-f3
12. Kc8-c7
13. Kf3-g3
14. Kc7-d8
15. Kg3-h4
16. Kd8-d7
17. Kh4-g5
18. Kd7-e8
19. Kg5xf5
20. Ke8-e7
21. Kf5-g5
22. Ke7-f7
23. f4-f5
24. Kf7-g8
25. f5-f6
26. Kg8-f7

Hi,

Thanks for your answers. How do one collect the PV if there is a TransTable hit. Do you store the PV in the transtable ? Because if let's say you search to depth 25 and at depth 13 you have an exact hit from the transtable. How do you collect the PV moves from 14 to 25 ?

Thanks.

[plattyaj]

Thanks for your answers. How do one collect the PV if there is a TransTable hit. Do you store the PV in the transtable ? Because if let's say you search to depth 25 and at depth 13 you have an exact hit from the transtable. How do you collect the PV moves from 14 to 25 ?

Thanks.

You have a truncated PV - you'll see it all the time. Some engines specifically mark the PV to show it was a hash hit.

Andy.

[AndrewShort]

my engine, which isn't particularly fast at all (and is far from ready for primetime, lacks many features, etc.), searches depth 26 in less than 1 second (which includes iterations 1 - 26).

Total nodes searched was something like 173,000 (my nodes are not counted if there is a usable hash hit).

In that position, null move is not used in my engine.

My hash replacement scheme is pure hash replace.

Within my engine search, twofold rep is a draw.

Here's my PV from a depth 26 search (note the black f5 pawn is captured on the 19th move of the PV)

1. Ka1-b1
2. Ka7-b8
3. Kb1-c2
4. Kb8-b7
5. Kc2-c3
6. Kb7-c7
7. Kc3-d3
8. Kc7-b7
9. Kd3-e2
10. Kb7-c8
11. Ke2-f3
12. Kc8-c7
13. Kf3-g3
14. Kc7-d8
15. Kg3-h4
16. Kd8-d7
17. Kh4-g5
18. Kd7-e8
19. Kg5xf5
20. Ke8-e7
21. Kf5-g5
22. Ke7-f7
23. f4-f5
24. Kf7-g8
25. f5-f6
26. Kg8-f7

Hi,

Thanks for your answers. How do one collect the PV if there is a TransTable hit. Do you store the PV in the transtable ? Because if let's say you search to depth 25 and at depth 13 you have an exact hit from the transtable. How do you collect the PV moves from 14 to 25 ?

Thanks.

if the PV is truncated due to a succesful hash hit working its way back to the root, then you can try to reconstruct the PV from the hash table after the fact. Sometimes this works, sometimes it doesn't - depends on how/when your hash entries get replaced. Also, even if it does manage to retrieve a longer PV, it might not be the PV that gave you the score, again, due to replacements. So play with that at your own risk