Fritz 16

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Dan Cooper
Posts: 184
Joined: Sun Nov 01, 2015 2:15 am

Re: Fritz 16

Post by Dan Cooper » Mon Nov 13, 2017 7:49 pm

kasinp wrote:
Aser Huerga wrote:
kasinp wrote:It does say that "the new multiprocessor engine was developed by star programmer Vas Rajlich ("Rybka") and is up there with the global elite."

Sounds like a continuation of Rybka development to me.

PK
I guess they forget to add "a few years ago ...". I bet for the same Rybka 4.1, maybe slightly tweaked.
Just installed it, and you may be right :).

1. It feels almost identical to the Fritz15 engine
2. Parameters look exactly like they did in Rybka series
3. Hash seems still limited to 4GB
4. It is a UCI engine
5. It does use Syzygy tablebases.

no idea as to strength increase, yet.
PK
I have been assuming this entire time that Fritz 16 is Fritz 15 with no elo improvements.

peter
Posts: 1451
Joined: Sat Feb 16, 2008 6:38 am

Re: Fritz 16

Post by peter » Mon Nov 13, 2017 7:57 pm

Thanks for the info, David!

I'd be interested, if the engine now was able to use more than 4G hash, which F15 could not.

And or even Syzygy- usage now?

Both would at least be a clear sign to me, if the engine really had gotten relevant development indeed,
Peter Martan

kasinp
Posts: 169
Joined: Sat Dec 02, 2006 9:47 pm
Location: Toronto

Re: Fritz 16

Post by kasinp » Mon Nov 13, 2017 8:13 pm

peter wrote:Thanks for the info, David!

I'd be interested, if the engine now was able to use more than 4G hash, which F15 could not.

And or even Syzygy- usage now?

Both would at least be a clear sign to me, if the engine really had gotten relevant development indeed,
The engine still uses Nalimov Tb (only GUI reads Syzygy); 4GB hash limit is still there. I see Rybka (let me call it that) searching deeper than before, but at least for the moment it is getting kicked around rather badly in a quick bullet match against Shredder 13... improvement seems to be limited.

PK

peter
Posts: 1451
Joined: Sat Feb 16, 2008 6:38 am

Re: Fritz 16

Post by peter » Mon Nov 13, 2017 8:29 pm

Thanks a lot, Peter
:!:
Peter Martan

Modern Times
Posts: 2144
Joined: Thu Jun 07, 2012 9:02 pm

Re: Fritz 16

Post by Modern Times » Tue Nov 14, 2017 3:54 am

Is chess960 supported ? Given that Rybka supported it Vas certainly knows how to do it. Perhaps the Chessbase GUI has limitations.
.

Opinions expressed here are my own, and not necessarily those of the CCRL Group.

carldaman
Posts: 1545
Joined: Sat Jun 02, 2012 12:13 am

Re: Fritz 16

Post by carldaman » Tue Nov 14, 2017 7:41 am

Dayffd wrote:Available Tuesday, Nov 14 (Chessbase.com). Even has emoticons to show Fritz's mood. Very important, of course. :)
Apparently, there is a some kind of bug causing the GUI (not the engine) to keep grabbing more RAM the longer it runs until it overwhelms the system resources. No idea why it's happening, but it affects both the 64 and 32-bit
installations. The only workaround I see is to close the program before that happens, and relaunch.

Hopefully there will be a permanent fix. I only tested it on a Win 7 box.

tpoppins
Posts: 606
Joined: Tue Nov 24, 2015 8:11 pm
Location: upstate

Re: Fritz 16

Post by tpoppins » Tue Nov 14, 2017 11:10 am

Thanks to Peter Kasinski for the early sale headsup. Here are some observations about Ry... erm, Fritz 16.

1) It's UCI, like Fritz 15. Oddly, some posts in this thread seem to express doubts about such possibility.

2) Chess960 is not supported - see the UCI options log below. There are "Chess960" strings in the exec, apparently leftovers from Rybka. In Cutechess Fritz 16 disappears from the list of the list of the available engines when the fischerrandom game category is selected.

3) Hash table is still limited to 4096 MB.

4) Same problem with tablebases as in later Fritz 15 versions -- see this post for details and the solution (which works for F16 as well). Perhaps the TBAccess option would work if you put the TBs into the directories hardcoded into the exec (C:\autoTB\TB4;C:\autoTB\TB5;C:\autoTB\TB6).

5) Bishop underpromotion is still never considered during the search. E.g. in the following position (taken from a 10-year old Rybka forum thread and attributed to Troitzky)
[d]1n4b1/PP6/6p1/4p1p1/p5P1/3p2PK/3P4/7k w - - 0 7
F16 won't see that 7.a8B draws and reports a mate in two for Black (7.axbQ Bd5 8.Qxe5 Bg2#). If you force 7.a8B it suddenly drops the score to something like 0.02; however, if you go back one ply to the diagrammed position (without clearing the hash) it goes blind again.

6) Haven't used the new F16 GUI yet so am unable to confirm the problem Carl reported. There were a couple of freezes in the Deep Fritz 14 GUI when switching from one Let's Check position to the next that happened for a while and then went away; no problems in the F15 GUI so far.

UCI output with TBAccess.dll installed in \Program Files\Common Files\ChessBase\Endgames:

Code: Select all

id name Fritz 16
id author Vasik Rajlich
option name UCI_Query type button
option name System Query type button
option name Hash type spin min 2 max 4096 default 32
option name Log Dir type string default <empty>
option name Log type check default false
option name xLog type check default false
option name Max CPUs type spin min 1 max 2048 default 1
option name Random type spin min 0 max 4095 default 0
option name CPU Usage type spin min 1 max 100 default 100
option name Use TBAccess type check default false
option name Main Process Priority type combo default Inherit var Normal var BelowNormal var Low var Inherit
option name Child Process Priority type combo default BelowNormal var Normal var BelowNormal var Low
option name Use Large Pages type check default false
option name Always Score Main GetMove type check default false
option name Display Upperbounds type check default false
option name Preserve Analysis type check default false
option name Clear Hash type button
option name Hash File type string default <empty>
option name Save Hash type button
option name Load Hash type button
option name Ponder type check default false
option name MultiPV type spin default 1 min 1 max 256
option name MultiPV_cp type spin default 20000 min 0 max 20000
option name TC Buffer type spin default 3 min 0 max 120
option name TC Normal GetMove Time type spin default 80 min 1 max 10000
option name TC Max GetMove Time type spin default 100 min 1 max 10000
option name White Pawn millipawns type spin default 0 min -10000 max 10000
option name Black Pawn millipawns type spin default 0 min -10000 max 10000
option name White Knight millipawns type spin default 0 min -10000 max 10000
option name Black Knight millipawns type spin default 0 min -10000 max 10000
option name White Light Squared Bishop millipawns type spin default 0 min -10000 max 10000
option name White Dark Squared Bishop millipawns type spin default 0 min -10000 max 10000
option name Black Light Squared Bishop millipawns type spin default 0 min -10000 max 10000
option name Black Dark Squared Bishop millipawns type spin default 0 min -10000 max 10000
option name White Rook millipawns type spin default 0 min -10000 max 10000
option name Black Rook millipawns type spin default 0 min -10000 max 10000
option name White Queen millipawns type spin default 0 min -10000 max 10000
option name Black Queen millipawns type spin default 0 min -10000 max 10000
option name Score Offset millipawns type spin default 0 min -10000 max 10000
option name Rook Endgame Penalty millipawns type spin default 0 min -10000 max 10000
option name Bishops Are Opposite Colored Penalty millipawns type spin default 0 min -10000 max 10000
option name Opposite Colored Bishops Endgame Penalty millipawns type spin default 0 min -10000 max 10000
option name Rook Endgame Scaling type spin default 100 min 0 max 1000
option name UCI_LimitStrength type check default false
option name UCI_Elo type spin default 1200 min 1200 max 2400
uciok
UCI output with TBAccess.dll renamed/deleted:

Code: Select all

id name Fritz 16
id author Vasik Rajlich
option name UCI_Query type button
option name System Query type button
option name Hash type spin min 2 max 4096 default 32
option name Log Dir type string default <empty>
option name Log type check default false
option name xLog type check default false
option name Max CPUs type spin min 1 max 2048 default 1
option name Random type spin min 0 max 4095 default 0
option name CPU Usage type spin min 1 max 100 default 100
option name NalimovPath type string default <empty>
option name NalimovCache type spin min 1 max 256 default 1
option name NalimovProbeFrequency type spin min 0 max 256 default 36
option name Main Process Priority type combo default Inherit var Normal var BelowNormal var Low var Inherit
option name Child Process Priority type combo default BelowNormal var Normal var BelowNormal var Low
option name Use Large Pages type check default false
option name Always Score Main GetMove type check default false
option name Display Upperbounds type check default false
option name Preserve Analysis type check default false
option name Clear Hash type button
option name Hash File type string default <empty>
option name Save Hash type button
option name Load Hash type button
option name Ponder type check default false
option name MultiPV type spin default 1 min 1 max 256
option name MultiPV_cp type spin default 20000 min 0 max 20000
option name TC Buffer type spin default 3 min 0 max 120
option name TC Normal GetMove Time type spin default 80 min 1 max 10000
option name TC Max GetMove Time type spin default 100 min 1 max 10000
option name White Pawn millipawns type spin default 0 min -10000 max 10000
option name Black Pawn millipawns type spin default 0 min -10000 max 10000
option name White Knight millipawns type spin default 0 min -10000 max 10000
option name Black Knight millipawns type spin default 0 min -10000 max 10000
option name White Light Squared Bishop millipawns type spin default 0 min -10000 max 10000
option name White Dark Squared Bishop millipawns type spin default 0 min -10000 max 10000
option name Black Light Squared Bishop millipawns type spin default 0 min -10000 max 10000
option name Black Dark Squared Bishop millipawns type spin default 0 min -10000 max 10000
option name White Rook millipawns type spin default 0 min -10000 max 10000
option name Black Rook millipawns type spin default 0 min -10000 max 10000
option name White Queen millipawns type spin default 0 min -10000 max 10000
option name Black Queen millipawns type spin default 0 min -10000 max 10000
option name Score Offset millipawns type spin default 0 min -10000 max 10000
option name Rook Endgame Penalty millipawns type spin default 0 min -10000 max 10000
option name Bishops Are Opposite Colored Penalty millipawns type spin default 0 min -10000 max 10000
option name Opposite Colored Bishops Endgame Penalty millipawns type spin default 0 min -10000 max 10000
option name Rook Endgame Scaling type spin default 100 min 0 max 1000
option name UCI_LimitStrength type check default false
option name UCI_Elo type spin default 1200 min 1200 max 2400
uciok

Dayffd
Posts: 395
Joined: Wed Sep 30, 2009 3:30 am

Re: Fritz 16

Post by Dayffd » Tue Nov 14, 2017 11:33 am

kasinp wrote:
peter wrote:Thanks for the info, David!

I'd be interested, if the engine now was able to use more than 4G hash, which F15 could not.

And or even Syzygy- usage now?

Both would at least be a clear sign to me, if the engine really had gotten relevant development indeed,
The engine still uses Nalimov Tb (only GUI reads Syzygy); 4GB hash limit is still there. I see Rybka (let me call it that) searching deeper than before, but at least for the moment it is getting kicked around rather badly in a quick bullet match against Shredder 13... improvement seems to be limited.

PK
When it arrives I was intending on a preliminary short (maybe 10 - 12) games match (prob around 5 minutes/game) with Deep Shredder 13. I wonder if Chessbase is more interested in adding more bells and whistles to the GUI than significantly improving the engine. One item that seems to have been dropped since F 14 is an elo field and that is a bit of a nuisance.
David S.

noctiferus
Posts: 364
Joined: Sun Oct 04, 2009 11:27 am
Location: Italy

Re: Fritz 16

Post by noctiferus » Tue Nov 14, 2017 2:57 pm

"5) Bishop underpromotion is still never considered during the search ": the old Rybka's footprint... I remember the long discussions about it.

schack
Posts: 96
Joined: Thu May 27, 2010 1:32 am
Contact:

Re: Fritz 16

Post by schack » Tue Nov 14, 2017 2:59 pm

I have unofficially heard that it improves on Fritz 15 by roughly 60 elo points, but that remains to be verified through testing.

Note that Stockfish 8 is also automatically installed.

Post Reply