My phrasing is not so good, but simply speaking, I discovered Go and started playing Go because at that time, I did not own a fast computer, only second hand laptop. At that time, if I had possessed a fast computer, I would probably have kept playing mainstream games, like Word of Warcraft, Diablo3, CounterStrike...

I kept purchasing second hand computers these last 10 years, a sort of ecological activism (I don't play games, except Go, and I use Linux, so no need of modern computer). So I am interested in strong bots that can run on slow computer.

Now, with all those strong bots available, and also because it would help me develop GRP, I am considering purchasing a recent/powerful computer

_________________
I am the author of GoReviewPartner, a small software aimed at assisting reviewing a game of Go. Give it a try!

6 games match (H3) between LZ13(c0cb60) and gtp4zen (V6) 10" per move

(Sabaki 0.33.4, win10, i9-7920X, 2x1080Ti)

command line for gtp4zen :
-z 6 -t 12 -T 10 -s 200000

command line for LZ13 :
--gtp -w c0cb60.txt --noponder -r 10 -t 4 --gpu 0 --gpu 1
time_settings 0 10 1

Result : 3-3 for this match , zen won its 3 games by (early) resignation.

(in the nine H3 games in total, LZ wins 6-3)
In conclusion, on my PC, LZ is very probably 3 stones stronger than Zen6.

Take a look at the Z3 game

And now, I'll let my computer (and my flat) cool down, because between CPU-fan and GPU-fans, it's very hot !


PS. In the controversy between high-end and low-end computers , I would make a difference between
a) actually playing go : a low-end computer can be very well suited if you're not Ke Jie.
b) for computer go, to test the Elo-strength of a program, determine if it's pro strength, or such things, I think you need good CPU/GPU

The games :
L1
L2
L3
Z1
Z2
Z3 !

-s [ --strength ] arg Set the playing strength (default 10000)
In the website i provided, there were discussion about -s default is too low, some suggested 150000, i use 200000, i may find the value somewhere else. as to -s max, no where to find. another problem might be -T some said it is 20 max even if you set more than 20s is still 20s or less (this i have checked manually, -T 30 is around 30s, so -T can be set more than 20) people complained about gtp4zen might not allow zen6 or zen7 to reach their max strength due to it's argument setting.
Just tested Zen6 30s vs. LZ 30s, LZ won. zen6 looks like no match to LZ.

I have the same experience as you do, i never play games not even chess except go. but 10 yrs earlier than you did; except i do not buy 2nd hand pc's. i use linux since its inception or even before its inception (redhat.) A lot of people discussed about porting these AI go to linux (i use usb-stick linux, so i could use one pc for two OS's.)

i just experienced LZ still has a ladder problem. weight file is the latest, dated 4/12 or 4/13.

well said and i agree. my interest is how to utilize AI go to improve go players' ranks. exactly speaking, how to attack the deficiency (i.e. time consuming) of human learning process by machine learning examples. But up to this point, i only could use them to improve my opening.

_________________
http://tchan001.wordpress.com
A blog on Asian go books, go sightings, and interesting tidbits
Go is such a beautiful game.

In the new version of gtp4zen (0.3.4) added 4 new parameters. Obviously, without them, gtp4zen played differently than the original zen.
https://github.com/breakwa11/gtp4zen/releases
It is possible that the new version also has differences from the original.

gtp4zen 0.3.5
https://github.com/breakwa11/gtp4zen/releases
fix handicap games

Thank you for the link !

Just one remark : to make it work, you must have the Zen.dll (for v6 or v7) and you must put it in the same folder as gtp4zen.exe

And a question : this relase is by breakwa11, he is the author of the excellent ongoing tournament at https://github.com/breakwa11/GoAIRatings#go-ai-ratings , where the Elo of LZ#xxx, ElfVx, minigo#xxx, zenVx, etc are continuously updated.

His reference point is Zenv7 1D =1800Elo
Do you know how to set gtp4zen at 1D ?

EDIT : I've just seen this :
zen7 1D ----> -t1 -s1800 -n1 -o1.3 -p0.35


I only know these options.

-h [ --help ] Show all allowed options.
-z [ --zenverion ] arg Version of dll, must be 6 or 7. (default 7)
-t [ --threads ] arg Set the number of threads to use. (default
CPU_CORES)
-T [ --maxtime ] arg Set the max time for one move. (default 10 seconds)
-s [ --strength ] arg Set the max playout. (default 10000)
-i [ --ithink ] arg thinking interval, only set 100 when play cgos.
(default 100)
-n [ --ilevel0 ] arg factor0. (default 1)
-o [ --ilevel1 ] arg factor1. (default 1)
-p [ --ilevel2 ] arg factor2. (default 1)
-r [ --resign ] arg resign. (default 10)
-l [ --logfile ] arg Enable logging and set the log file. (default none)
-L [ --logfilenametime ] Add timestamp after log filename. (default off)
-d [ --debug ] Enable debug output to gtp shell. (default off)

zen6 and zen7 settings from 8k to 9D+
https://github.com/breakwa11/GoAIRating ... /AIcmds.md (at the bottom of the page)

for the -T option to work, you need to remove gtp4zen.lua:
"If gtp4zen.lua exists in the folder where gtp4zen.exe is located, the command-line parameter -T is automatically invalidated and the time control of gtp4zen.lua is used.
This script is used to fine-tune the time control. If you don't understand his appointment, you can Delete gtp4zen.lua, the default file exists and takes effect."