Using TRUUS to analyze BoomstraDam games

[Ed Gilbert]

what is the plydepth you reach at move 32 (in how many seconds , 20?) ?

In the search log below, the depths are nominal/avg/max, where avg means the average leaf node depth, not the average depth of all nodes.

Code: Select all

best 32-28, value -4, depth 1/2.9/5, nodes 45, time 0.00, 45 kN/s
best 34-30, value -4, depth 3/4.8/11, nodes 408, time 0.00, 408 kN/s
best 34-30, value -4, depth 5/7.8/17, nodes 4692, time 0.00, 4692 kN/s
best 34-30, value -2, depth 7/9.7/19, nodes 22533, time 0.02, 1408 kN/s
best 48-42, value -2, depth 9/11.4/27, nodes 74056, time 0.02, 4629 kN/s
best 48-42, value -6, depth 11/14.1/29, nodes 258015, time 0.06, 4095 kN/s
best 48-42, value -6, depth 13/15.3/30, nodes 794780, time 0.17, 4621 kN/s
best 48-42, value -8, depth 15/17.5/33, nodes 2606337, time 0.50, 5213 kN/s
best 48-42, value -2, depth 17/19.9/35, nodes 15318304, time 2.86, 5356 kN/s
best 48-42, value -4, depth 19/22.2/39, nodes 44335378, time 8.25, 5374 kN/s
best 48-42, value -2, depth 21/24.7/41, nodes 169577215, time 31.80, 5333 kN/s

Ed, at move 49 you see a databasewin, which database did you use for that ( 6, 7, 8, 9 ?)

That is with the 9pc db, and it takes 90 seconds to see a db win value.

-- Ed

[Ed Gilbert]

Feike,

I see that you deepen your iterations in steps of 2 plies, unlike Truus and Flits. It's similar to Kingsrow and Chinook, but these programs iterate over the odd number of plies whereas you do the even numbers. Did you experiment with both methods? Ed, did you?

Rein, I started using odd plies about 8 or 9 years ago and never gave it much thought, except that I felt going in steps of 2 ply increments seemed about the right size. I would be surprised if even ply increments worked any differently than odd, at least in kingsrow. Those are only the nominal depths anyway, only achieved in the absence of any truncations or extensions.

-- Ed

[FeikeBoomstra]

There is some theoretical background for the 2 ply increment, it is claimed to be more efficient. I started with even plies, and to be honest never tried it the other way around.

Originally my thought was that if you had even plies, you get a more honest evaluation, because your opponent has the possibility to "correct" your actions.

But due to search extension and reduction, this is not always the case of course.

First I stopped my search only at the max-nodes, but that did not work well.

[Ed Gilbert]

Is there a way to make Truus use its games database as an opening book? When it plays games it shows all the positions in the database that match the current position, along with some statistics on which moves have been played in games and their results, but I cannot make Truus play any of these book moves. It always does a search to pick its moves. I have set the menu option Opties, Stel Openingsrepertoire In. What else is needed?

-- Ed

[Ed Gilbert]

I have another Truus question. I have been playing test games between kingsrow and truus. In these game I have configured kingsrow to use only its 6-piece database, and I think I have truus set to use its 6pc db also, but I'm questioning if I set it correctly. Below is the tail end of a game in which kingsrow won playing with white. The start position shown below (move 49) is the first position in the game where kingsrow returned a search score of database win. But truus did not show its equivalent of database win until after move 60, where there are 5 pieces remaining. From moves 49 to 60 it shows scores of approximately 1.5 to 2.5. This makes me think that it is not using its 6pc database. I have used the menu "Opties", "Stel Overige Opties In", and in this dialog I have the correct Eindspel Database Directory, and a check in the box "6 Stukken database (anders 5)". Is there anything else I need to set?

-- Ed

[FEN "W:W16,21,25,27,30,35,43:B7,13,14,19,20,24,33."]
49. 21-17 7-12 50. 17x8 13x2 51. 16-11 33-39 52. 43x34 24-29 53. 34x23 19x28 54. 11-6 28-33 55. 6-1 33-38 56. 27-22 38-42 57. 22-17 42-48 58. 30-24 20x29 59. 1x34 48x30 60. 25x34

[GuidoB]

I have another Truus question. I have been playing test games between kingsrow and truus. In these game I have configured kingsrow to use only its 6-piece database, and I think I have truus set to use its 6pc db also, but I'm questioning if I set it correctly. Below is the tail end of a game in which kingsrow won playing with white. The start position shown below (move 49) is the first position in the game where kingsrow returned a search score of database win. But truus did not show its equivalent of database win until after move 60, where there are 5 pieces remaining. From moves 49 to 60 it shows scores of approximately 1.5 to 2.5. This makes me think that it is not using its 6pc database. I have used the menu "Opties", "Stel Overige Opties In", and in this dialog I have the correct Eindspel Database Directory, and a check in the box "6 Stukken database (anders 5)". Is there anything else I need to set?

-- Ed

[FEN "W:W16,21,25,27,30,35,43:B7,13,14,19,20,24,33."]
49. 21-17 7-12 50. 17x8 13x2 51. 16-11 33-39 52. 43x34 24-29 53. 34x23 19x28 54. 11-6 28-33 55. 6-1 33-38 56. 27-22 38-42 57. 22-17 42-48 58. 30-24 20x29 59. 1x34 48x30 60. 25x34

Try "Opties", "Bewaar Settings" and restart Truus. At my PC, it sees a database win within 15 seconds.

[Ed Gilbert]

Try "Opties", "Bewaar Settings" and restart Truus. At my PC, it sees a database win within 15 seconds.

I tried this and let it search for at least 5 minutes, but the search score only gets to 2.4. In another test, I set up a 6-piece position and can see in the Eindspel window it gives the correct db result of all the moves, so I think it is correctly accessing the db. I'm guessing it just does not change the search score when it see a db win during a search.

-- Ed

[Rein Halbersma]

Try "Opties", "Bewaar Settings" and restart Truus. At my PC, it sees a database win within 15 seconds.

I tried this and let it search for at least 5 minutes, but the search score only gets to 2.4. In another test, I set up a 6-piece position and can see in the Eindspel window it gives the correct db result of all the moves, so I think it is correctly accessing the db. I'm guessing it just does not change the search score when it see a db win during a search.

-- Ed

Ed,

Weird bug! Just as Guido, I also get a score of 9.99 within a minute of searching. There are several idiosyncracies with the interface of Truus, though. First, as Guido mentioned, settings are not persistent across sessions, so you need to explicitly set them. Second, once a winning move has been found, the search stops but does not always display the winning variation correctly.

On "C:Program Files\Truus" there is a manual file "handleid.rtf". YOu can run that file through Google translate to learn more about all the options.

Rein