Crazy Stone Deep Learning -The First Edition-

Это не 7 дан, даже на сильной машине, далеко не 7 дан. Просто какая-то их система показала там 7 дан. Я могу придумать систему, гду я буду 99 дан. Вот они и придумали. Программа и близко не играет как человек. Ходы машинные. Любит играть на центр, как и всё семейство програм монте-карло. Проигрывая пол очка - проигрывает всё. Выигрывая пару десятков очков может отдать вам всё преимущество, а потом и последние пол очка, делает ошибки, а выигрышь в пол очка считает надёжным, т.е. не умет выигрывать и не умет проигрывать. И как только у программы проскочит оценка, что у вас может быть лучше, она сразу отдаст вам ещё несколько своих групп, доводя игру до абсурда. Не ищите логики и обучения. Программа вкинет вам все ко-удары, растратит все свои возможности в позиции (адзи) и только потом возьмёт ко. Просто обязательно навкидывает камни по первым линиям. Просто так, потому что может вкинуть. При этом программа играет сильно. Но если для вас это сильная программа, то и несколько другим программ будут играть так же, вы просто не увидете разницы. Если вы начинающий, то с десяток программ сможет вас обыграть надёжно до конца ваших дней, нет никакой разницы. Про самыю сильную программу враньё, Zenith играет сильнее. Устаревшая база дебютов. Ничего нового, по сравнению с предыдущими версиями я не заметил. Думает дольше, поэтому играет чуть сильнее, вот и вся разница. Zenith тратит на партию 15 минут по сравнению с 50 минутами сумашедшего камня. Вывод такой. Описание не соответствует товару и вводит в заблуждение, именно поэтому такой негативный отзыв.

I bought the game even with that ridiculous price tag thinking that at least it would be like the mobile version. It is way inferior to the mobile version where you can have graded games. This is a joke, right?

I love it. This is a nice board game and great AI inside.

꽤 강하군요. 타이젬 8단인데 두기 버겁습니다.

看了棋魂，二话不说买了这个游戏，谢谢棋魂给我梦，谢谢疯石给我路。

Ok, so before everything else, a little about me: I'm not really strong at go, I played for two years and reached 12~13k on KGS, then I had to stop for two years to focus on studying. When I got back into go 2 months ago, I decided to train hard in order to "get my old 'strength' back". And that included playing against this bot once it was out. This bot beats me everytime (of course), but it doesn't feel at all like I'm playing a 7d. My first game against it (unlimited time, 7d strength) was lost by 40.5 points in an even game (in case you're interested, here's the link: https://gokibitz.com/kifu/VyqbrOUmb ). This is a bot that was supposed to give me 17 stones at the time, and it only beat me by 40.5 points. In case you're wondering, no, I'm not 3d, I'm around 8k on OGS. So, either it isn't 7d /at all/, or it played slack moves once it knew it was ahead (like AlphaGo, maybe). I personnally think it's a mix of the two. On your everyday computer, I don't think it reaches 7d, unless you force it to do 10M playouts thanks to the analyze feature, but that would mean a looooot of time for each move. The interface is also extremely old school. While it isn't ugly, it lacks game setting customization: you can only choose between predetermined main times and byo-yomis, the komi can't exceed 9.5 (so no reverse-komi handicap games), and you can only either choose between a set time for the bot to use or an unlimited time and setting a rank, so no option to "do at least X playouts per move). Funnily enough, if you put it on the "unlimited time, 7d rank", it almost never thinks about moves for more than a few seconds, but when you set it with a limited time per move, it thinks a lot more and usually plays "better" (as far as I can observe at my level, anyway). All in all, it is very expensive, and only people who are really serious (and can afford the price) about learning go will even want to consider buying this. However, I would still recommend this to those because of the analyze feature I mentionned earlier, you can make it analyze a position for as long as you want, and this way you can review games you played, see what moves it considers bad, and what it proposes instead. Playing games against it has also allowed me to progress a lot (because, at my level, it still is way better than me). If you're patient, I'm sure a free strong enough engine with similar features will be out within 1~2 years, and by the time, (hopefully) CrazyStone Deep Learning version 2 with more customizable game settings and an even stronger engine. I personnally hope V2 will be free for those who purshased the V1. In the meantime, I need to practice more :) EDIT: I've run a series of games on OGS to test its strength, it settled at 2~3d with 10 seconds per move and high priority (on a i5-3210M @ 2.5GHz), so it is decently strong even on an average computer.

There's a user-defined tag for this product that reads "casual." I can't really say that this product is for the most casual observer of Go/Igo/Weiqi/Baduk. However, what I can say is that if you're an amateur player looking for an AI Go sparring partner then here's your option on Steam. In fact, it appears to be one of the most compelling options at the moment for playing this game on Steam (the other being Tabletop Simulator if you have a friend). Verdict: If you want to Go it solo on Steam (all easy puns intended) and practice against one of the more competent AI's on the market, here is your choice. But for $80 let's hope that future updates or "The Second Edition" will be included. I'll recommend it for the purpose of hopefully inspiring more products such as this to hit the Steam Store. With even more hope, I'd like to see something like KGS or Tygem on here as well.

Like it a lot, strength of PC player slightly overrated but this is not a big issue. Recommendation for any go player who plays the game on PC to pass the time and get a good opponent.

Human like play. Positional calculation is a major plus. But the program has a weakness with ladders.

很好的围棋游戏，ai实力还行，相对最强围棋来说好了不止一点，英文也容易看懂，特此推荐