Important IvanHoe Settings

[kingliveson]

I like the following settings combined with RobboTripleBases:

Repeating Time Control

Code: Select all

TimeEasyFactor 14
TimeOrdinaryFactor 64 
TimeAbsolutePercent 24
BufferTime 2000

Sudden Death Time Control

Code: Select all

TimeEasyFactor 13
TimeOrdinaryFactor 48 
TimeAbsolutePercent 20
BufferTime 2000

Incrementing Time Control

Code: Select all

TimeEasyFactor 15
TimeOrdinaryFactor 60 
TimeAbsolutePercent 20 
BufferTime 2000

These time management settings are based not just on sample games posted, but also through analysis and observation. Chances are they are not optimal. Also, only tested blitz games with ponder off. Am most unsure about the repeating time control -- still working on it.

AllowInstantMoveFromHash should be off -- only good for hyper-bullet matches in my opinion. It is designed for ponder off games, but I think it has a negative effect.

Pawn Hash is another important setting; it should be 1/16th of total hash size. For example, if Hash size is set to 512 MB, Pawn Hash should be 32 MB. You can get away with 1/8th, but anything larger will have a negative effect.

[kingliveson]

Edited the Incrementing Time Control settings. Also, I noticed the Repeating Time Control setting performs better on my Phenom compared to slower Athlon. If anyone finds these settings to be awful, not optimal, or has better configuration, please let us know.

[BTO7]

Thanks king ...great stuff here ! Appreciate the tips.

BT

[benstoker]

I like the following settings combined with RobboTripleBases:

Repeating Time Control

Code: Select all

TimeEasyFactor 14
TimeOrdinaryFactor 64 
TimeAbsolutePercent 24
BufferTime 2000

Sudden Death Time Control

Code: Select all

TimeEasyFactor 13
TimeOrdinaryFactor 48 
TimeAbsolutePercent 20
BufferTime 2000

Incrementing Time Control

Code: Select all

TimeEasyFactor 15
TimeOrdinaryFactor 60 
TimeAbsolutePercent 20 
BufferTime 2000

These time management settings are based not just on sample games posted, but also through analysis and observation. Chances are they are not optimal. Also, only tested blitz games with ponder off. Am most unsure about the repeating time control -- still working on it.

AllowInstantMoveFromHash should be off -- only good for hyper-bullet matches in my opinion. It is designed for ponder off games, but I think it has a negative effect.

Pawn Hash is another important setting; it should be 1/16th of total hash size. For example, if Hash size is set to 512 MB, Pawn Hash should be 32 MB. You can get away with 1/8th, but anything larger will have a negative effect.

What does 'buffertime' setting do exactly?

[kingliveson]

Incrementing Time Control modified slightly to:

Code: Select all

TimeEasyFactor 15
TimeOrdinaryFactor 66 
TimeAbsolutePercent 22 
BufferTime 2000

[kingliveson]

I have settled with these for now:

Repeating Time Control

Code: Select all

TimeEasyFactor 14
TimeOrdinaryFactor 64 
TimeAbsolutePercent 24
BufferTime 2000

Sudden Death Time Control

Code: Select all

TimeEasyFactor 13
TimeOrdinaryFactor 48 
TimeAbsolutePercent 20
BufferTime 2000

Incrementing Time Control

Code: Select all

TimeEasyFactor 18
TimeOrdinaryFactor 69 
TimeAbsolutePercent 23 
BufferTime 2000

[kingliveson]

Putting it altogether; small sample data on how these settings were derived is available here: http://www.open-chess.org/viewtopic.php?f=4&t=313

Repeating Time Control: 4'/40+4'/40+4'/40 (4 minutes for first 40 moves, then 4 minutes for next 40 moves, and so on).
Sudden Death Time Control: 1+0, 10+0, 60+0, etc.
Incrementing Time Control: 1+1, 15+10, 60+30, etc.

Ponder Off

Repeating Time Control

Code: Select all

TimeEasyFactor 14
TimeOrdinaryFactor 64 
TimeAbsolutePercent 24
BufferTime 2000

Sudden Death Time Control

Code: Select all

TimeEasyFactor 13
TimeOrdinaryFactor 48 
TimeAbsolutePercent 20
BufferTime 2000

Incrementing Time Control

Code: Select all

TimeEasyFactor 18
TimeOrdinaryFactor 69 
TimeAbsolutePercent 23 
BufferTime 2000

Ponder On

Repeating Time Control

Code: Select all

TimeEasyFactorPonder 31
TimeOrdinaryFactor 64 
TimeAbsolutePercent 24
BufferTime 2000

Sudden Death Time Control

Code: Select all

TimeEasyFactorPonder 28
TimeOrdinaryFactor 48 
TimeAbsolutePercent 20
BufferTime 2000

Incrementing Time Control

Code: Select all

TimeEasyFactorPonder 39
TimeOrdinaryFactor 69 
TimeAbsolutePercent 23 
BufferTime 2000

Additional Ponder on/off settings:

AllowInstantMoveFromHash: Off
Pawn Hash: 1/16th of total hash size

Note: There is good probability these time management settings are not optimal. The ponder-on settings have not been tested at all -- they are based on a formula which is simply scaling ponder-off TimeEasyFactor to default TimeEasyFactorPonder. Am still working on it, and will update if something better is found.

Edit: IvanHoe version used in all these tests is unmodified from source, and is available here: http://chess.cygnitec.com/engine/ivanhoe/

[kingliveson]

Code: Select all

Time Management 5401  2010 

				                    1	         2		      3            4
1  IvanHoe 9.54b-TM10 x64  ***********  26.0 - 24.0  27.5 - 22.5  25.5 - 24.5    79.0/150   
2  IvanHoe 9.54b-TM00 x64  24.0 - 26.0  ***********  24.5 - 25.5  26.5 - 23.5    75.0/150   
3  IvanHoe 9.54b-TM11 x64  22.5 - 27.5  25.5 - 24.5  ***********  25.0 - 25.0    73.0/150  0.00  
4  Deep Rybka 4 SSE42 x64  24.5 - 25.5  23.5 - 26.5  25.0 - 25.0  ***********    73.0/150  0.00

Missing from the list is IvanHoe 9.54b-TM01 x64 which will be thrown in along with Stockfish 1.8 next. Rybka was included to avoid incestuous results. 50 games between each pair of engines definitely are not enough -- more games ahead.

IvanHoe 9.54b-TM00 x64 + Neither Option
IvanHoe 9.54b-TM01 x64 + Option 2
IvanHoe 9.54b-TM10 x64 + Option 1
IvanHoe 9.54b-TM11 x64 + Option 1 & 2

Option 1

Code: Select all

#if 0 /* about the 0.6 elo for the hasslings */ /* measure to UCI? */
  if (!mtg)
    {
      rat = ((double) (TIME + 1)) / ((double) (OPP_TIME + 1));
      if (rat > 2.0)
	rat = 2.0;
      if (rat < 0.5)
	rat = 0.5;
      if (rat > 1.0)
	{
	  mult = ((int) (1024.0 * rat)) - 1024;
	  DESIRED_TIME += (mult * DESIRED_TIME) >> (10 + 2); /* max 1/4 */ 
	}
      else
	{
	  mult = ((int) (1024.0 / rat)) - 1024;
	  DESIRED_TIME -= (mult * DESIRED_TIME) >> (10 + 2); /* max 1/4 */ 
	}
    }
#endif

Option 2

Code: Select all

#if 0
  if (ROOT_DEPTH && ROOT_SCORE < -50)
    DESIRED_TIME += DESIRED_TIME >> 3;
  if (ROOT_DEPTH && ROOT_SCORE < -100)
    DESIRED_TIME += DESIRED_TIME >> 3;
#endif

The 1+0 (25 positions) ponder-off tournament used default settings for all engines. AllowInstantMoveFromHash default is off by compile. No EGTB was used. The code snippets are additional time management functions -- disabled by default in control.c.

Time Losses
IvanHoe 9.54b-TM00 x64 - 3
IvanHoe 9.54b-TM10 x64 - 4
IvanHoe 9.54b-TM11 x64 - 5
Deep Rybka 4 SSE42 x64 - 3

Tournament data available upon request.

Additionally, special thank you goes out to IvanHoe developers for listening to complaints about poor time management. A great effort was put in beginning from 999958-beta release to remedy the issue. Thank you!

[robbolito]

Hi kingliveson,
I am glad that you got good scores against Rybka.
The versions I have used so far did not do so well.
When can we expect the release of your compiles.
Thanks.

[kingliveson]

Hi kingliveson,
I am glad that you got good scores against Rybka.
The versions I have used so far did not do so well.
When can we expect the release of your compiles.
Thanks.

If you are interested, you can test the attachment.