Life In 19x19

Boidhre wrote:I'm familar with many of the arguments against playing with AIs but I'm attracted to MFOG's review function where it gives its thoughts on your moves or better moves than the one you made.

I have gotten no value at all out of the above. The reasoning is just contextless "local shape move" "try to kill enemy group" or whatever, and you really already knew that.

I have both MFOG and Crazy Stone. MFOG is useful in that it is really easy to turn the engine on and off (click the player name) when you are looking at a position to get a quick evaluation and it has a cool "lookahead" feature where it graphically shows numbered variations that it is thinking about. The other thing that is good is that it gives positional evaluations in probabilities on other squares where you could've moved, so if your move and its move both show "53" or something, you can probably decide this isn't the huge game-losing blunder if that's what you're looking for. And also, if you click on a group it will tell you if it thinks it is alive or dead or dying or seki or whatever.

One thing that I don't like about MFOG is that I can't figure out how to change the time settings in an sgf file that I am reviewing, and if I played a 45 + 5/30 game and was in byo-yomi on a particular move then the computer is only willing to think for 30 seconds in order to not time out. I get round this by editing the sgf file to add more time, but it is annoying. The overall feel of MFOG is the same sort of form-over-substance retro that KGS does.

Crazy stone has recently introduced functionality into the English version to get an automatic review of the game with a little chart (like here: http://www.grappa.univ-lille3.fr/~coulo ... index.html ) and you can get a list of the moves where your moves differed from its moves and how big a mistake it thought it was. Also, when you are on a move you can ask it and it will give you a list of what it thinks are good/bad moves and how good/bad it thinks they are. MFOG kind of does this but Crazy Stone's stuff is a little slicker, I think. However, the interface for the analysis is a little clumsy (you can't sort the list or anything like that) and overall Crazy Stone feels a bit more like a game and navigation of the game feels clumsy (you have to use ctrl-arrow to go forward a move, for example), but I bought it on the basis of the evaluation feature and am really happy with it.

Oh, and somehow the Crazy Stone engine finds moves much more quickly than MFOG. On my machine Crazy Stone with 15 minutes of time beats MFOG with an hour of time. Whether that is a difference in the engines or whether Crazy Stone is hogging processor space on my machine etc I would not know.

Sorry; tldr: MFOG is a little more user-friendly with navigating and clicking on a group so it tells you if it thinks it is alive or dead is cool. Crazy Stone is slicker and can auto-analyse game records. When I play them against each other on the same machine, Crazy Stone wins but this is not actually a big concern for me.

Hello,
Just another point of view, the ranking of the top computer go programs on KGS :
http://www.gokgs.com/graphPage.jsp?user=zen19
http://www.gokgs.com/graphPage.jsp?user=ManyFaces
http://www.gokgs.com/graphPage.jsp?user=crazystone
CrazyStone seems to have a little edge on the others.
I'm a long time customer of smartgo (on windows and ipad) and I like the UI a lot.
As for studying, my best results was obtained from practicing life and death problems and getting game comments from better player, online and IRL.

Because there were so many complaints about the copy protection of Crazy Stone, Unbalance changed their system: internet connection is now required only the first time the program is started.

Rémi

Rémi wrote:Because there were so many complaints about the copy protection of Crazy Stone, Unbalance changed their system: internet connection is now required only the first time the program is started.ty

Rémi

Thanks for that info, Remi. It's always nice to know game developers pay attention to the end users' concerns. This puts Crazystone back in the running for me. The price is still high but not out of line with MFG12.

Something I keep noticing with all these Windows go apps: It doesn't seem they are improving much. I see "Top Computer Program 2008" or 2010 or whatever. Where are the upgrades?

SCWillson wrote:Something I keep noticing with all these Windows go apps: It doesn't seem they are improving much. I see "Top Computer Program 2008" or 2010 or whatever. Where are the upgrades?

Actually, they're improving very rapidly, probably the fastest since they were far into the kyu ranks. See this table: http://senseis.xmp.net/?KGSBotRatings. Perhaps what you're seeing is that the fact that there's a lot of competition, and the program that had a claim to be the best a few years ago has lost the title to someone who has already lost the title.

That makes sense.

Sadly Computer Go tournaments don't help the average Go player decide which one to buy when the winning programs are running on supercomputers and 100 CPU arrays.

SCWillson wrote:That makes sense.

Sadly Computer Go tournaments don't help the average Go player decide what to buy when the winning programs are running on supercomputers and 100 CPU arrays.

I think most people just get the free ones.

oren wrote:
I think most people just get the free ones.

No doubt. I really like all the bells and whistles that come with MFG12 and SmartGo, but Drago with GnuGo or Fuego as engines work very well and the graphics are actually better.

I think I'm going to put my planned purchase of a commercial Go app on hiatus until I get a bit more experience both with the ones I have and with playing Go. Perhaps by the end of summer I can make a more informed decision. And who knows? I might end up sticking with GnuGo.

SCWillson wrote:That makes sense.

Sadly Computer Go tournaments don't help the average Go player decide which one to buy when the winning programs are running on supercomputers and 100 CPU arrays.

Not that extreme. By and large the amount of hardware each of the contestants is running on is published. Perhaps not useful for the "average Go player" if by that you mean somebody unable to interpret that data, look up "benchmarks" and compute just how many times more powerful the tournament machines and/or bot machines are than his or her home machine.

Note that machine power and time trade off. So if (for example) the MFOG bot on KGS is running on a machine 3-4 times more powerful than your home machine that's the same as giving your home machine 3-4 times the time allotment (note: that's if the particular program begins by analyzing the machine on which it finds itself running and then sets parameters according to that and the selected time control)

SCWillson wrote:According to their website CrazyStone must make contact via Internet every time it is started in order to validate that it is a legitimate copy.

The above quote was made in this thread a while back. But after I read it I went over to the Unbalance Crazy Stone website and found the below quote that shows that they clearly only need to connect to the internet once.

"When you play Crazy Stone 2013 for the first time, your PC must be connected to the internet for activation by serial number. (This process will only be required once and we will not obtain any personal information.)"

Yes, that has been changed. You seem to have missed post #33 by Rémi.

Yes, thank you.