Are there any plans to incorporate Kata Go into this app? The additions of variable komi, score estimation, and improved handicap play would be awesome! Thanks.

_________________
- Brady
Want to see videos of low-dan mistakes and what to learn from them? Brady's Blunders

wineandgolover san,

Thank you for your continuous interest in my app!

I am also interested in porting KataGo into "A Master of Go" and trying it.
But I think that I need time.
An uncertain point is whether Core ML 3 can support KataGO NN or not.
Current Core ML 2 does not support KataGo NN.


For my understanding of KataGo, I have mocked up a Lizze with KataGo on Web.

https://github.com/new3Rs/react-lizzie

You can play with it until I will port KataGo into "A Master of Go".

Enjoy!
(Caveat: "React Lizzie with WASM KataGo" is just a mock-up. It is not a stable app.)

Hi y_ich,

You definitely have my ongoing support. At this year's Go Congress, I had people approaching me to see what Brady-sensei thought of their games. Of course this meant "Can I see what Master of Go thinks?"

I am interested and saddened that Leela Zero works in ML2 or 3, and that KataGo doesn't. Is it just a different language, or a completely different implementation approach? I wish you luck in figuring it out and finding a solution.

I am delighted to see that you are continuing to improve your app. I will be patient. Thanks again.

_________________
- Brady
Want to see videos of low-dan mistakes and what to learn from them? Brady's Blunders

wineandgolover san,

Thank you for promoting my app in the Go Congress^^

KataGo uses an operation of Neural Network which Core ML 2 does not support.
And KataGo needs more information(ladders, liberties, etc.) for NN.

I will study them as well as KataGo's advantages themselves.

Thank you.

Even though I don’t really play go these days anymore, it’s still very nice to see an app like A Master of Go being developed and updated as frequently as it has been.

Please don’t mind my following suggestions even though I have not purchased the app.

While I am not sure how Apple’s Apps Store handles searches, it may be easier for people to find the app if you include terms such as the Japanese, Korean and Chinese names of the game of Go as well as their English equivalents such as igo, baduk, and weiqi in the app description.

Can the app function in landscape instead of portrait orientation? If not, it would be nice to have this function because not a lot of iPad covers or stands support portrait orientation.

For some reason, based on the preview screenshots, I find the aesthetics of the app may be able to be improved. I mean it is already using slate and shell stone, grained board and tatami images, but maybe you could try putting together similar images with different colour temperature/balance? Perhaps the variety of colours like the purple buttons and the black background chart at the top could have a more coherent colour scheme with the board and stones? Is there a way to hide some of the buttons and chart during a game or review?

I wonder if having Go stones and/or board in the app icon would make it more relatable to the game of Go.

In any case, it’s really nice to see your efforts in this app. Please keep it up!

pragmaticleas san,

Thank you for your interest in my app!

Nice suggestion about English description. I put multi-byte character words only for localized descriptions.
I will put them in English description.

The app support landscape mode too though no special features are added.

I agree that the aesthetics is not good since I am not a designer^^;
The theme color is purple, which is the color of Tao, but no adjustments between it and images are applied.
I will think about it and the app icon in the next major version.

Thank you for your advices!

Yuji, could you please explain the use of "coins" in the app? I generally avoid buying apps that require an in-app purchase. The changelog for your app says that the coins are used for "scoring", because it is a server-side function. How important is "scoring" to the effective use of this app? Would I need or want this scoring function in order to study games?

trgo san,

Thank you for your interest in the app!

Basic use case of scoring is to know how much stones you “lost” at the of end of game with AI.
(When you won, it usually means that AI resigned before end game.)
Scoring is only available in study mode.
Since you enter study mode right after the end of game, you can check the score immediately.

0% of winrate does not mean a big defeat, so some people care the score.

If you do not care, you do not need to pay any extra money.

I hope that this comment will be helpful for you.

Thank you.

Thank you for your prompt reply, Yuji-san. It sounds like I won't need the coins.

If anyone has a chance to test it with the new iPhone, I'm curious how well the app performs on it. I assume somewhat better than on the last edition, but have no idea whether it'll be a big difference.

_________________
Occupy Babel!

just tested on a brand new Iphone 11 pro Max. Must say it was terrific for the A13 chips are very strong. I use the Elf engine and also the Elf V1 weight, average nps around 150-200/s and eps around 130-150/s. I supposed it could do better but need some time to adjust. I also have a problem what is meaning of "Ladder thinking expending max moves", what is the proper number setting for this? Another problem is the "batch size number", there are 1-8 options, default setting is 1, should I set to the max?

elliotdong san,

Thank you for using my app!

About "Max. times of time-extensions for ladders",
when a primary variation of a move which the AI selects ends with some Atari position, the AI extends reading.
The AI needs at least a few hundreds playouts to read simple ladders.
Since A13 is pretty fast, you do not need to care the setting if you give the AI more than 3 seconds per move.
This setting is for devices older than iPhone XS.

About batch size,
if you choose 8, maybe you will get 10% more eps(evaluations per second) on iOS 13 depending on weight you use.
(On iOS 12, I was sorry that there were no effects)
But honestly I am not sure it is effective for strength or not.
batch( = parallelism) may loose accuracies of playouts even if it increases playout speed.

I hope that the information will helpful for you.

Thanks for the reply, Mr. Y_ich. I still had a question that I tried many combinations of the Engines and the weights. Seemed the ElfOpenGo engine and Elf V1/V2 weights were the slowest combination of the nps and eps playouts. what I mean is that if you try the Leela master or leela 15b weights, both the nps and eps playouts will be more faster. Is it that means the Elf weights are stronger than the others? or as well as the Elf Engine is stronger too. I Just try to find an appropriate combination for the A13 chips. Thanks again.

elliotdong san,

You are right, you will get more playouts by 15b weights than by ELFs.
But it does not mean that they are stronger at the condition of time parity.

It seems that ELFs are stronger than 15b LeelaMaster but LeelaMaster plays better than ELFs for handicapped games.
And it seems that Leela Zero #242 may be a little stronger than ELFs.

All these weights are super human, so I recommend you to find your favorite one by seeing their styles rather than strenghs.

Tired of borrowing my wife's iPad mini (most recent version), I got a new iPhone 11 Pro.

Running side-by-side with her iPad it seems to be at least 20% faster on the leelaMaster gx89 version, less so on LeelaZero 242.

I’m happy enough. Will explore more.

_________________
- Brady
Want to see videos of low-dan mistakes and what to learn from them? Brady's Blunders

Hi y_ich,
1) On my iPhone 11 pro, running iOS 13.1.3, increasing batch size seems to make it a little faster, but so far, it also finds difficult-to-find better moves much quicker. I'm currently using batch size of 5.

2) Are the settings defined anywhere? What do all the options do? For example, what is low thread priority, resign threshold, Cinit, softmax temperature, and virtual loss? Can I assume that networkLeelaZero_242 the official 40-block network 242 converted to your format?

3) I would like to help you update the networks, but I can't follow the directions on your GitHub page: https://github.com/new3Rs/a_master_of_g ... /README.md

Just for starters, the sample dropbox links provided tell me I don't have access, so I can't see your examples.

I'm not a programmer. However, I am able to follow step-by-step instructions, and was able to compile / install lizzie / LZ / Katago. So if you can either help me directly, or clarify the readme, I'd be happy to build the json / converted weight files, to make available the most current 15b, 20b, and 40b Leela Zero weights. I'm using macOS.

If you'd rather leave it to a gifted programmer, or if one is reading this and wants to help, that is fine with me.

Current strongest models:

LZ15b_245
http://zero.sjeng.org/networks/0d2694c5 ... 7e9c9c3.gz

LZ20b_248
http://zero.sjeng.org/networks/9d0457b1 ... 6da8751.gz

LZ40b_249
https://zero.sjeng.org/networks/6ee288b ... 6cf0538.gz

Thanks.

_________________
- Brady
Want to see videos of low-dan mistakes and what to learn from them? Brady's Blunders

wineandgolover san,

Thank you for keeping using my app!

About No.2, I amd sorry for lack of documentation.
If you set "low thread priority" on, you will reduce speed of playouts and get longer battery duration.
"resign threshold" is the winrate at which AI resigns.
"Cinit" determines wideness of search tree. The larger Cinit is, the wider tree become.
"softmax temperature" determines broadness of policy(next move intuition). The smaller it is, the narrower candicate moves become.
"virtual loss" adjusts how much the search avoids same variations for parallelism. the larger it is, the better AI avoids. It does not mean the stronger though.

You can leave these parameters alone, or select one among "instant parameter settings"

And yes, LeelaZero_242 was converted from the official #242 weight.

About No.3,
Could you give me a log with error?
I may help you or fix the problem if it is a bug.

You seem to want to get the very strongest AI at present.
Good, me too^^

But I want to point out that the difference of Elo rating between #249 and #242 is only 155 points, which is about 70% winrate.
(It sounds big but do you really want to choose best sparring partner among　Tyson, Foreman, Ali, et al.? I am kidding.)
I will update the weight when the difference of rating become 200 point bigger.

Thanks.

y_ich-san,

do you have any plans of including KataGo that has support for handicap stones, variable komi and territory ownership estimation?

Does iOS 13's Core ML support the required operations?

Thank you!

Grünauer san,

I have succeeded to run KataGo on iOS 13 though I am not sure it supports handicap stones and variable komi well.
Do you have good experiences of KataGo on PC?

Still I need time to embed it into my app.

Please look forward.

One excuse. You may need extra money for KataGo since it works only on iOS 13.
(Otherwise, iOS 11/12 users will have to pay for a function they cannot use...)

I am running it on macOS 10.15 on a 2019 iMac; it works very well.



I don't mind paying for an in-app purchase once to get this new functionality.

Thank you.