Stockfish 8 UCI engine

[Luigi Vogliobene]

Hi to all,
I just downloaded, from the official Stockfish repository, the updated sources (16/11/2016) and compiled a new version of Stockfish 8 for arm7.
If you would like to give it a try all you have to do is :

1. save the original engine a-stockf (version 7) of picochess :
    'cd /opt/picochess/engines/armv7l'
    'cp a-stockf stockfish7'

2. copy the attached file to the folder '/opt/picochess/engines/armv7l'

3. make it executable with
    'sudo chmod 755 a-stockf'

4. Rebuild the engines .ini:
    'cd /opt/picochess'
    'sudo python3 ./build_engines.py'

5. reboot the RPi 
    'sudo reboot'

I played only one game but seems more stable than the previous version.

Luigi

[Al]

Hi Luigi,

Thanks for providing the New Stockfish 8 Engine, however it loses on time to me on every game played at Fischer 5,3 and in a tournament against Stockfish 7, it drew the 1st game and lost the 2nd on time in a BPP against R ending.

So I'm sorry to say it still has time issues.

Cheers,

Al.

[Luigi Vogliobene]

Hi Al,
I can only confirm your assessment, yesterday, after playing a single game (and lost w/ Stockfish8) I hope this release fixed the time issues but tomorrow it loses on time twice out of four games.
I revert back to stockfish7, sadly ... waiting for the next release :)
Thank you
Regards
Luigi

[Al]

Hi Luigi,

Although your Stockfish8 compile is slow moving in PicoChess and SCID tournaments at quicker time controls, I ran a mini 4 round tournament of Stockfish7 v Stockfish8 at a longer time control and Stockfish8 is much stronger ....

It won 3.5 to 5 & in all 3 wins swapped a rook for a minor piece & pawn to gain a much better end game. If only we could cure this time problem at blitz levels ....

The pgn is attached.

Cheers,

Al.

[Al]

The score should read 3.5 to 0.5

Al.

[Luigi Vogliobene]

Hi Al,
this new Stockfish 8 build was given to me from a guy of the Immortal Chess Forum.
If you would like to give it a try, may be a tournament ...
It is only 245 Kb
Thank you
Luigi

[Shivkumar Shivaji]

You can also try stockfish directly from github.

"git clone https://github.com/mcostalba/Stockfish"

"git checkout master"

and then cd src

"make build ARCH=armv7"

or

"make build ARCH=general-32" if above does not work.

--
You received this message because you are subscribed to the Google Groups "PicoChess" group.
To unsubscribe from this group and stop receiving emails from it, send an email to picochess+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Al]

Hi Luigi & Shiv,

I'll give it a go when my current tournament ends. Is it any better on Picochess?

Shiv, the last Stockfish 8 build was compiled exactly like your 1st suggestion, and the build before exactly like your 2nd suggestion by DJ.
The latest build on the 16th seems very strong but has time issues on shorter time controls.

Cheers,

Al.

[Al]

Hi Luigi,

I'm sorry to say it's worse than your own version. In a 4 round SCID tournament against Stockfish 7, it won 1, drew 1, lost 1 & lost the remaining game on time in a
K+R v K+B+P ending (which was totally drawn).

I did notice a slight speed improvement in PicoChess on Fischer level 5,3 but that may have been due to me adding 2 parameters in the 'a-stockf.uci' file:

Hash = 256
SlowMover = 1

I'm going to try the same parameters in your earlier version, as I'd already tried SlowMover values from 100 down in 10s but had not tried lower than 10.

Cheers

Al.

[Luigi Vogliobene]

Hi Al,
thank you very much fot testing.
Yesterday also Stockfish 7 lost on time (level 4 30' + 15" fischer time)  1 out of three games, today also level 5 w/ hash = 384, I am a bit disappointed about the time management of this engine in RPi
I'll give a try again to my own build of stockfish 8 the I'll revert to others engines ...
Good evening :)
Luigi

[Al]

Hi Luigi,

No it's not my parameters changes, it doesn't make any difference with the SlowMover options.

Your newer Engine is definitely faster in PicoChess than your own compile but weaker ....

I wonder if you can contact this guy who supplied the latest compile to see what he did, maybe some sort of faster move parameter can be included in the compile?

Cheers,

Al.

[Luigi Vogliobene]

Hi Al,
I just wrote him :) now I'll wait for his reply and I'll inform you :)
Luigi

[Luigi Vogliobene]

Hi Al,
the guy replies to me :

"it's a standard build from git repository

make profile-build ARCH=armv7


make ARCH=armv7 COMP=gcc config-sanity
make[1]: Entering directory '/home/pi/chess/Stockfish/src'

Config:
debug: 'no'
optimize: 'yes'
arch: 'armv7'
bits: '32'
kernel: 'Linux'
os: 'GNU/Linux'
prefetch: 'yes'
popcnt: 'no'
sse: 'no'
pext: 'no'

Flags:
CXX: g++-4.8
CXXFLAGS: -Wall -Wcast-qual -fno-exceptions -fno-rtti -std=c++11 -pedantic -Wextra -Wshadow -DNDEBUG -O3 -flto
LDFLAGS: -Wl,--no-as-needed -lpthread -Wall -Wcast-qual -fno-exceptions -fno-rtti -std=c++11 -pedantic -Wextra -Wshadow -DNDEBUG -O3 -flto "

Have a nice day.

Luigi

Il giorno venerdì 18 novembre 2016 20:03:49 UTC+1, Al ha scritto:

[Luigi Vogliobene]

Hi Al,
this is my attempt to build a new stockfish 8, this is my make command:

sudo make profile-build ARCH=armv7 COMP=gcc config-sanity

Attached you will find the executable

Luigi

Il giorno venerdì 18 novembre 2016 20:03:49 UTC+1, Al ha scritto:

[Lutz]

Hi Luigi,

trying this on a RPi3 (after "git clone https://github.com/mcostalba/Stockfish" etc.) I get

/usr/include/arm-linux-gnueabihf/gnu/stubs.h:7:29: fatal error: gnu/stubs-soft.h: No such file or directory

In that directory is a stubs-hard.h but no stubs-soft.h.

Is there an easy way to handle this?

Thanks for an answer!

Lutz

[Luigi Vogliobene]

Hi Lutz, I downloaded the sources from the Stockfish repository ( https://github.com/official-stockfish/Stockfish ) and the make I posted works good for me.
Luigi

[Luigi Vogliobene]

I am just doing some tournament test w/ Arena 1.0 on my RPi3 between the Stockfish7 official of Picochess and the Stockfish 8 build I compiled today : "my" build run almost three times faster than Stockfish7 (200 vs 590 kN/sec) reaching eventually 4 to 8 more plies in deep.
I will post the results when the tour will end (since now two draws)
Luigi

[Al]

Hi all,

I've been out on a family Birthday all day so have had a few beers, I will test the latest engine tomorrow. I was aware arena was out as DJ let me know, but I haven't installed it yet, hence still using SCID vs PC.

Cheers,

Al.

[Lutz]

Hi Luigi,

it was very easy to compile an own stockfish8 with the sources from the repository you gave me. Many, many thanks!

Regarding the speed numbers above I can say:
Stockfish7 is running with ~560 kN/s
your 1st version of stockfish8: ~585 kN/s
your last version: ~625 kN/s  - What's your secret? :-)  I thought profile is profile and already self-optimizing?!

All values measured on DGT Pi with RPi3 after a reboot.

200 kN/s is almost as running on 1 core only. But then it rather should be 150 kN/s - strange.
Doubling the speed brings about 1 ply with modern engines like stockfish. Less than 2 times doubling (200 vs 590) --> less than 2 plies. But in ELO it's ~ 70+35 = 105 on RPi3 and short time controls. That's why I'm a fan of speed and stockfish8 instead of SF7 (other 70 ELOs = ~175 in summary). Analysing in ponder mode possible problems with time management don't matter. Engine matches are another thing of course.

Lutz

[Luigi Vogliobene]

Hi Lutz,
happy to ear you overcome our issues :)

my 1st build : sudo make stockfish ARCH=armv7

my 2nd build: sudo make profile-build ARCH=armv7 COMP=gcc config-sanity

This is the difference between the two builds, the sources are identical.

Luigi

[Luigi Vogliobene]

I ended the tournament because my RPi3 was overheating :)

Sf7 vs Sf8

2 - 7 - 8

I think there is no question about the strength of the new engine by Costalba & Co.

Luigi

[Al]

Hi all,

Yes unfortunately in a Stockfish tournament where both engines are using 4 cores, the RPi is running at 100% all the time. Even with heat sinks the RPI3 still gets extremely hot. For this reason I run tournaments between 4 core engines on a RPi2 that doesn't get so hot.

By the way, how do you measure the Kn/S of a chess engine, is there a command?

Cheers,

Al.

[Al]

Hi Luigi,

I've just added your latest Stockfish8 compile to PicoChess and SCID vs PC.

It is an extremely slow mover on PicoChess in a Fischer 5,3 time control taking 30+ seconds per move for the 1st 4 moves, then lost on time at move 22.

I have a tournament running on SCID in game 1 Stockfish8 is a pawn up after 38 moves so far ....

Cheers,

Al.

[Luigi Vogliobene]

Hi, simply watching at the engines window at Arena 1.0 arm while they play in a tournament, it shows deep and kNs among the other things.
This stockfish8 build is extremely strong but has the same time management issue, I tested also the  stockfish 7 release and it also loses in time (if y could keep playing :) ) Other engines (Texel) work better in Fischer time 30' + 15" and if consuming a lot of time after the opening (like the Stockfish engines both do) they ALWAYS make a move just a second before the flag falls.
Luigi

[Al]

Hi Luigi,

Yes I can confirm that build is at least as strong as your 1st build, as it has just beaten Stockfish7 3.5 - 0.5 too.

I agree with Lutz, it's a monster when running in Ponder mode as time issues don't matter, so I will use it to monitor tonight's 7th game in the World Chess Championship.

Cheers,

Al.

[Lutz]

Hi Al,

regarding speed measuring:
Start stockfisch directly in console mode in the appropriate directory with ./a-stockf or whatever the filename is.
Then type
"uci"
"setoption name Hash value 256"
"setoption name Threads value 4"
"go infinite" or "go movetime 60000" or "go depth 23"

Of course you can choose values as you wish.
After each ply or at the end of computing you can see best move with pv, nodes, nodes per second, time, hash full and so on. I let it run 1 minute for example.

Lutz

[Al]

Hi Lutz,

Great, that works really well

Cheers,

Al.

[Al]

Hi all,

Ok now that I've got Arena workiing efficiently thanks to Luigi, I found the following when pairing Stockfish7 v Stockfish8 on my RPi2:

1 core, 10 rounds of 5 mins each, 8 wins to Stf8, 0 wins to Stf7, 2 draws
4 cores, 10 rounds of 5 mins each, 2 wins to Stf8, 1 Stf7, 7 draws

Throughout the games, the kN/S were 100 higher for Stockfish7, i.e. 429 to 336. However I'm pretty sure my Stockfish7 is a later compile than that on PicoChess.
It appears that the faster Stockfish7 speed makes up for the stronger Stockfish8.
I'm sure the kN/S would be higher on my RPi3 but I'm using PicoChess on that for analysing the World Chess Championship.

Cheers,

Al.

[Uwe Badermann]

Hello Al,

I'm not absolutely sure how to construe the results for the different no. of cores used (1 vers 4 is quite a big difference),

but since the rating of SF hasn't improved dramatically (at least the desktop version I know a bit), it's ok, that

SF8 doesn't win against SF7 at 100%. In this way the 4core result looks plausible to me.

For the nodes calculated:

Unfortunately this is not a real indication for processing speed or even strength.

Because, if the search algorithm changes,the engine calculates more or less nodes than the predecessor did.

Though the result is an increase in strength.

Example (fictional):

Take scholar's mate white to make his 4th move Qxf7#

If one engine calculates Qf4, f5, f6, f7 it took 4 nodes to mate.

If the other engine starts calculation at Qf7 (intention downwards f6, f5,f4) it took her (is "her" the correct preposition?!) just 1 node.

Just comparing 1 node versus 4 nodes - subject you haven't got any additional information, would make you believe,

its not to well performing, wouldn't it?

If you know, what the engine is crunshing, like here, it's easy to judge.

But normally you don't know....

Just a remark

And BTW you and Luigi you're really awesome guys!

It's really a pleasure to see how you bring you expertise into this project.

Take my hat off!!

Cheers

Uwe

[Al]

Hi Uwe,

Thanks for that. I was aware that wasn't the best way to disguinguise the best Engine compile, so I'm now running a new Tournament on Arena.

The Engines are my best Stockfish7 compile against DJ's Stockfish 8 compile & Luigi's 3 Stockfish 8 compiles. I'm just running a small 2 round/20 game blitz first to see if it works, then I'll increase the rounds and time control.

Cheers,

Al.

[Luigi Vogliobene]

Hi Uwe,
thanks for that :)
This is a fantastic project, thanks to RPi3 and Picochess I have the chance of learn a bit of Linux from such a skilled and kindly guys.
Thanks to you ALL
Luigi

[Al]

Hi all,

I had to stop my Arena Stockfish tournament as I'm not sure what's happening?

I set up a 5 Engine tournament of 4 rounds blitz of 5 mins per game. The 1st 2 engines played each other, 1 of them won, then the next 2 engines played. Bearing in mind they should only play each other 4 times, the same 2 engines played each other 6 times, drawing all and started a 7th? Each time with the same colours, so unless an engine has to win before the next round, I have no clue what's going on?

Luigi, any ideas?

In the mean time I've aborted the Arena tournament & restarted the same tournament on SCID, where I know what's going on .....

Cheers,

Al.

[DJ Dekker]

Hi Al,

I'm experiencing similar trouble while running tournaments in Arena 1.0. For the time being Scid vs PC definitely remains the better choice.

Cheers,

DJ

Op maandag 21 november 2016 19:06:56 UTC+1 schreef Al:

[Shivkumar Shivaji]

I really wish Arena published their source code for the linux version :) For now, you can also try your luck in the Arena forum to raise their attention.

Its also a good idea to turn ponder off in tournaments to be extra safe.

--
You received this message because you are subscribed to the Google Groups "PicoChess" group.
To unsubscribe from this group and stop receiving emails from it, send an email to picochess+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Luigi Vogliobene]

I experience similar problem too
Luigi

[Al]

Hi all,

My tournament on SCID for Stockfish 7 and all 4 builds of Stockfish 8 at 5 min game over 4 rounds, total 40 games has finished.

Stk8 Luigi3 9.5/16
Stf8 Luigi2 8.5/16
Stf7 8/16
Stf8 Luig1 7/16
Stf8 DJ. 7/16

There were 9 losses on time 7 were clear wins for the Engine that won on time. Out of the other 2 games, Luigi3 v Luig1 & Luigi1 v Luigi3 were both even, so the table would still be correct.

So as I expected, Luigi's last compile looks the strongest, but Stockfish 7 is surprisingly strong.

I'm now running a 10 min game tournament for the Stockfish 8 builds only.

Cheers,

Al.

[Al]

Hi again,

The 10 mins each game tournament has ended, this time there were only 3 losses on time in the 24 games between the 4 Stockfish 8 builds:

Stf8LV2 7/12
Stf8LV1 6.5/12
Stf8LV3 5.5/11
Stf8DJ 5/12

So a different set of results, however I won't be happy until there are no losses on time.

I've started a new tournament of the same 4 engines over 4 rounds, same 24 games but with 15 mins a game for each Engine.

Cheers,

Al.

[Uwe Badermann]

Good evening Al,

buonasera Luigi,

Why do I have the picture in my mind, that a kindergarden for RPis exist? 

And you both are the nursery teachers letting the RPi's play with each other
...a room full of RPi's on the floor,...crawling,...shouting,....

:-)

Now, becomming serious.

Maybe this information might be helpful or of interest:

I'm guilty! 

I must admit, that I have fumbled the Hash.

Set it to 512MB and SF7 to level 7 with 4 cores at 90+30 to train for my team ch. on Sunday.

(on image 0.75, no time to update, sorry) 

The time pressure for SF7 extends dramatically!

At the moment I'm writing this, we are in move 30 and SF7 has 7mins left on the clock.

(meanwhile the game ended in move 46. due to time flag)

When I recall (at least that was my experience) 256MB at level 6 same time control without time trouble for SF7,

I get the feeling that there must be an optimum for each

HASH / level / Time control combination

Maybe the administration/search in the hash is not so good at certain dimensions,

or SF7 simply cannot deal with increments?

If the question for the optimum combination is relevant, than I'm asking myself, 

if you can somehow integrate this into your RPi's tournaments, to maybe

disclose the devil in SF#. 

Or better to say, to finally have a matrix, which in terms of loosing on time,

shows the best combination for the setup of the engine.

I am not sure if that is the right strategy to approach this problem.

Just an idea.

Have a nice evening

Uwe

[Al]

Hi Uwe,

On my RPi3, PicoChess Hash settings are set at 192, that value came from you I believe.

On my RPi2 Tournament machine, SCID vs PC has suggested Hash Values for each Engine, Stockfish is 16 so I've set them as that. Arena has no suggestions, so they are set at 192 as PicoChes, but Arena is buggy so I'm awaiting the next level.

I too have played around with settings on Stockfish 8 on PicoChess, my level 6 has Threads at 4, Hash at 192, & SlowMover at 1, this does speed up the level in 5 min chess, but not enough to prevent Time losses.

I'm trying to find out the best Stockfish 8 build via tournaments. I've just stopped the 15 mins per game tournament over 4 rounds with 4 engines, so 24 games as I saw 2 losses on time in the 1st 4 games. I've now started a 5 sec per move tournament (no time constraints) to see how long that takes, then will adjust to 10/15/20 secs per move when I can work out how long each tournament will run.

Cheers,

Al.

[Al]

Hi all,

My 5secs per move 4 Engine Stockfish 8 tournament over 4 rounds & 24 games took just under 3 hrs. There were only 2 wins, the same 2 engines winning and losing against each other so the result was a 6/12 draw with all 4 engines. Obviously the time per move was too small, but no losses on time as there wasn't a time limit.

So I've started the same tournament at 15secs per move, expecting a result in around 9 hrs time. Then the top 3 engines will face 20 games against each other resulting in a 60game tournament expected to last around 22 hrs if at 15 secs per move, but I'll adjust the move time based on the the current tournament, so that the new tournament will run for around 24 hrs.

Cheers,

Al.

[Al]

Hi all, the latest 4 round, 24 game at 15 sec per move 4 Engine tournament is nearing the end. With 1 game to go the current table is:

Stockfish8 Luigi2 6.5/11
Stockfish8 Luigi3 5.5/11
Stockfish8 Luigi1 5.5/12
Stockfish8 DJ. 5.5/12

So as you can see whatever happens in the last game, it's a close run thing again. However as Stockfish8 DJ has consistently come last or equal last, I'm dropping it for the final tournament. I wanted a 60 game tournament lasting around 24 hrs, but as the current 24 game tournament is still running after 13 hrs, where 5 games have been over 100 moves (2 of those over 170 moves), I'm going to drop the tournament to 48 games, where each Engine will have 16 games against the others, resulting in a score out of 32 each.

The winner will be my choice of Stockfish8 Engine for PicoChess. Of course if a latter compile comes forward, I will play that against the winner here.

Cheers,

Al.

[Al]

Hi all,

The last game in the penultimate tournament was a draw so the table was:

Stockfish8 Luigi2 7/12
Stockfish8 Luigi3 6/12

Stockfish8 Luigi1 5.5/12
Stockfish8 DJ. 5.5/12

The last tournament is still ongoing 30/48 games are complete, I'm expecting a result late this evening ....

Cheers,

Al.

[Al]

Hi all,

The Stockfish8 tournament has finished, this was 16 rounds at 15 secs per move, 3 engines, so 48 games:

Scid vs. PC
?, 2016.11.23 - 2016.11.24
Score Stf8LV1 Stf8LV2 Stf8LV3
----------------------------------------------------------------------------
1: Stf8LV1 17.5 / 32 XXXXXXXXXXXXXXXX =1====1====1=00= ==01==101===1=== (+7 -4 =21)
2: Stf8LV2 15.5 / 32 =0====0====0=11= XXXXXXXXXXXXXXXX ==============01 (+3 -4 =25)
3: Stf8LV3 15.0 / 32 ==10==010===0=== ==============10 XXXXXXXXXXXXXXXX (+3 -5 =24)
----------------------------------------------------------------------------
48 games: +8 -5 =35

This showed that Luigi's first compile won this tournament but we already know this was the slowest of the compiles on PicoChess. Overall tho, Luigi's 2nd compile won & was also the Engine that lost the least on time during the tournaments, at 1 only:

_______________________________________________________
Engine T1 T2 T3 T4 T5 Total
_______________________________________________________
SF8L1: 7/16 6.5/12 6/12 5.5/12 17.5/32 42.5/84
SF8L2: 8.5/16 7/12 6/12 7/12 15.5/32 44/84
SF8L3: 9.5/16 5.5/12 6/12 6/12 15/32 42/84
SF8DJ: 7/16 5/12 6/12 5.5/12 23.5/52
SF7 : 8/16 8/16

T1: Game in 5 mins, 4 rounds, 5 engines, 40 games / 7 losses on time
(L1x2, L2x1, L3x2, DJx3, SF7x1)
T2: Game in 10 mins, 4 rounds, 4 engines, 24 games / 3 losses on time
(L1x1, DJx2)
T3: Move every 5 secs, 4 rounds, 4 engines, 24 games
T4: Move every 15 secs, 4 rounds 4 engines, 24 games
T5: Move every 15 secs, 16 rounds, 3 engines, 48 games

Losses on time during T1+T2 - L1:3, L2:1, L3:2, DJ:5, SF7:1

So I'm going to use Luigi's 2nd compile as my Stockfish 8 Engine on PicoChess.

The last tournament pgn is attached.

Cheers,

Al.

[Al]

Hi all,

I've been trying to change some uci options so that PicoChess running Stockfish 8 does not lose on time in Blitz or Fischer 5,3 or 10,5 time controls.

Originally I had tried changing SlowMover which has a default of 100 & had changed it to 1, no effect. Then last night I was reading a thread whereby 10 is the lowest level it can be set at. Ok, it still didn't help. Then I read a thread whereby you can force an engine to make a make a move with 1 second to spare by changing the MoveOverhead to 1000 (ms), again this didn't help. I was trying these changes on level 6 only within a-stockf.

Then whilst checking my settings this morning, I noticed that I had SlowMover=10 and MoveOverhead=1000 (no spaces) whereby 'Skill Level = 6' had spaces.

So I changed my level 6 as follows:

[Level@06]
Skill Level = 6
Threads = 4
Hash = 192
Move Overhead = 1000
Slow Mover = 10

and Voila..... we have a fast moving monster.

I've taken out the 'Max Overhead' parameter as it's no longer needed, now I need to find the correct setting of 'Slow Mover' to get the correct balance of speed and strength.

Cheers,

Al.

[Al]

Correction ...... these changes were applied in 'a-stockf.uci' not 'a-stockf' ....... sorry
Al.

[Al]

Hi all,

Early indications show that 'Slow Mover = 30' is a good compromise.

I can now play a blitz game against Stockfish8 level 6 without waiting too long for a move, or knowing I'd eventually win on time. I'd have to add these same parameters to other levels if choosing those in blitz.

Of course, this value can be adjusted accordingly, the longer the game the higher the number (up to the default of 100).

Cheers,

Al.

[Luigi Vogliobene]

Hi Al, I agree with you, Slow Mover = 30 is a good compromise, my a-stockf.uci file was built with that exact value in it and the engine, for me, works well in Fischer 30'+15"
Thank you for your sharing :)
Luigi

[Uwe Badermann]

Great findings Al,

You're my star of today!

I searched for the notation of the uci option (uci defition attached) and the white space

seems reasonable, but not stated as clear as you would wish.

=> I propose that this should be added to the example uci-files in Picochess image, so pleople become aware of it!

Further more, in the uci definition, the input format for the option ist

Optionname_X_value_Y

Where "_" represents 1 white space

and the word "value" seems really literally to be used instead of "="?!

But maybe this is "translated" by Picochess for us?!

What concerns "slow mover" I didn't find any clear definition

http://www.talkchess.com/forum/viewtopic.php?start=0&t=42308&topic_view=flat

showing that in Shredder, this option is called "time usage percent". 

I have contacted Shredderchess for clearification and will follow up.

A "move overhead" of maybe 2 seconds (=2000?!) or 3 seems reasonable for me:

Without that parameter set, I played SF7@level7, 90+30, Hash 256(!) yesterday and the 

engine was really "fighting" not to lose on time. Three times the display already show "time", but

SF seems somehow made it in time, because the message vanishes 1 or 2 secs later and showed

the remaining 30sec for SF7.

(Ann: I don't remember all the details, but a few releases ago a delay of 1 or 2 seconds was implemented in PicoChess.

maybe this is relevant and maybe counterproductive in the context of losing on time?!)

Hash once again: In my opinion, SF7 plays better with >=192 and =<384. 

So somewhere in the area of 256 seems to be the optimum in terms of quality of play. 

Just my opinion - but proved by game analyze on a desktop.

Once again thanks for the work you're doing! Much appreciated!

Keep you posted

Cheers

Uwe

EXCERPT FROM ENGINE-INTERFACE

* setoption name <id> [value <x>]

this is sent to the engine when the user wants to change the internal parameters

of the engine. For the "button" type no value is needed.

One string will be sent for each parameter and this will only be sent when the engine is waiting.

The name and value of the option in <id> should not be case sensitive and can inlude spaces.

The substrings "value" and "name" should be avoided in <id> and <x> to allow unambiguous parsing,

for example do not use <name> = "draw value".

Here are some strings for the example below:

  "setoption name Nullmove value true\n"

      "setoption name Selectivity value 3\n"

  "setoption name Style value Risky\n"

  "setoption name Clear Hash\n"

  "setoption name NalimovPath value c:\chess\tb\4;c:\chess\tb\5\n"

[Al]

Thanks Luigi & Uwe,

Here's where I read about 'Slow Mover':
http://www.talkchess.com/forum/viewtopic.php?start=0&t=42308&topic_view=flat

Here's where I read about 'Move Overhead' (scroll down until about 1 page from the end):
https://www.reddit.com/r/chess/comments/3z7ubx/stockfish_7_is_out_60_elo_stronger_than_stockfish/
1000ms is 1 second, so yes Uwe, 2000 would be a 2 second buffer.

Here it what some of the other Engine parameters mean:
http://support.stockfishchess.org/kb/advanced-topics/engine-parameters

Cheers,

Al.

[Uwe Badermann]

Hi Al and Luigi,

just have received the answer from Shredderchess concerning 

"time usage percent" what' is in stockfish is "slow mover":

(Beginn - with my words)

When the engine comes to its result for a move, for example within 200 seconds, it uses

the percentage given (therefore 10 to 100 as value) to additionally 

go ahead with its calculation. So in this example and slow mover=10 for 20 seconds.

So it somehow verifies its result and therefore get's a bit stronger than 

with a lower value.

(End)

But this value should not have any impact on the overall time consumption

and therefore cannot be responsible for a loss on time!

(Since the slow mover must be limited by the Move overhead and time control in general)

Just FYI to have at least one mystery explained.

Now, enjoying WCC!

Good night fellows

Uwe

[Al]

Hi Uwe,

'This value shouldn't have an impact on the overall time consumption' but it does on the Stockfish 8 Engine, it helps us immensely.

I'm using Luigi's 2nd build and as you, now enjoying the 11th game of the WCC using Stockfish 8, level 20 with no UCI changes, in Ponder Mode ....

Cheers,

Al.

[Al]

Hi all,

Attached is my new Stockfish8 Engine compile from today's latest source.

Place both files in your /opt/picochess/engines/armv7l directory
(You may want to rename the old a-stockf & a-stockf.uci files to stockfish7 or whatever first)

By the way, Stockfish 8 has effectively won the TCEC season 9 super final against Houdini 5 as it is currently +16 -7 =69 and in front in Game 93/100.

Cheers,

Al.

[JesseJhon Andrews]

This link not working: "Here it what some of the other Engine parameters mean:
 http://support.stockfishchess.org/kb/advanced-topics/engine-parameters" #Sad Face

[Scally]

Hi,

It looks as though the whole of the Stockfish support site has been removed ...

This article might help, it also refers to the above link, ignore that:

https://www.chess.com/forum/view/general/help-with-setting-up-engine-stockfish-8-on-lucas-uci

Other UCI explanations can be found here:

http://wbec-ridderkerk.nl/html/UCIProtocol.html

https://python-chess.readthedocs.io/en/latest/uci.html#uci-commands

I hope this helps...

Al.