Not on 9x9 board. Can beat real 3 dan easly.

_________________
Tsumego/Tesuji apps for iPad, iPhone & Android devices:http://www.lifein19x19.com/forum/viewtopic.php?f=18&t=7511

Wait really? One of my theories of why I was a really bad player is for the life of me I can't beat crazystone w/ a 2 stone handicap at its highest strength ..(I get the 2 stones).. This makes me not feel as bad.. I've been trying to work my up the 9x9 before moving to the larger boards.. perhaps this isn't really the the right way to do it?

-avi

Yes, it is very strong on a small board. I can hardly win one out of ten and I'm 3k. On 19x19 it's weaker than real 6 kyu (by "real" I mean European ranking, to be more precise: Polish ranking).

If you have played a lot of 9x9 games already - move to 13x13.

_________________
Tsumego/Tesuji apps for iPad, iPhone & Android devices:http://www.lifein19x19.com/forum/viewtopic.php?f=18&t=7511

What we are seeing here is the result strong non-linearity between computer power available vs playing strength for the MCTS algorithm.

The computational requirements per move for 9x9 are much less than for 19x19. This is a case where we are examining a program running on a machine where the hardware is adequate for 9x9 (into the region where it can play well and where strength increases only slowly with more crunch power) and inadequate for 19x19 (below the region where the play of this algorithm is consistent).

Remember, probability is involved. With probability, reliability falls off very rapidly below some critical sample size and increases only slowly as sample size is increased above the minimum. The sample size is proportional (linear with) the crunch power of the hardware being used.

I don't think it's stronger than igowin on 9 by 9 board

I didn't made a direct comparison. PC or iOS version of igowin? I remember PC version was quite weak in pre-montecarlo times.

_________________
Tsumego/Tesuji apps for iPad, iPhone & Android devices:http://www.lifein19x19.com/forum/viewtopic.php?f=18&t=7511

Depends on the definition of "quite weak". I believe it was around 6 kyu KGS playing 19x19. That's what was in ManyFaces 11 (and what is still in ManyFaces 12 for the playing strengths 6 kyu and weaker).

THis is a fascinating thread and explains quite a bit. I used to be quite good at defeating various go programs (older versions of mfofgo / gnu go) on the 9x9 before moving onto the larger boards (it was clear they were bad at looking at the whole board and you could just corner them into a small section). (10 years later of not playing..) this time I'm doing things a lot more systematically but it is almost disturbing watching crazystone beat me on 9x9 at its highest levels, w/out me having really a clue to how it seems to exploit me so well. I was initially thinking I've just managed to have forgotten strategy in a way I knew earlier.

I didn't made a direct comparison. PC or iOS version of igowin? I remember PC version was quite weak in pre-montecarlo times.[/quote]

Yes, igowin in PC is quite weak. I can beat it on 9 by 9 board even it has 3 stones handicap advantage. Mais, I haven't won a game with 2 stones handicap in IOS

Something is wrong here. Darkpolarbear, if you are really a 4kyu (and in this case won't matter by what standard) you would NOT be easily beating ManyFaces 12 on a PC at its higher levels. On almost any real PC, that would be about 1 dan KGS on the 19x19 board.

Are you sure about what you are running? You paid for MFOG 12 and so have it enabled and aren't simply running the free demo which is limited to about 18 Kyu KGS on 19x19? And of course confusion about the names of the programs comes in. I don't really understand what you might have meant by "igowin in PC" except as MFOG 12 (igowin is Fotland's version of MFOG for the little devices)

And bloosqr, stop responding in terms of "real" rank. You should have noticed that I was specifying the rank basis. That's how we should be talking, simply saying what we mean when we give a number. It's not that I consider KGS "real" rank but what you are doing is like insisting that Centigrade is more "real" temperature than Kelvin when what is required is to specify which is being used.

Hey apologies that wasn't me.. but it was my fault. I mangled the quote recursion when I replied to a message so it looked like me. Just went back to fix it.

Maybe Darkpolarbear is referring to this free program:


Igowin Software Download

Igowin is a 9x9 free version of Many Faces of Go. Play go against the computer with great graphics. It keeps track of your record and playing strength and has an on-line manual with history of go, where to buy books and equipment, etc. It includes the Go Tutorial from The Many Faces of Go, so it is a good program to give a beginning go player. It plays competitively from 25 Kyu through 6 Dan, so stronger players will enjoy it as well. It learns from its mistakes, so every game is different.

Mike: there's an igowin for the pc: http://senseis.xmp.net/?Igowin. I've used it before.

_________________
Occupy Babel!

You mean this:

"Igowin (for Windows) is a 9x9 go program that uses a weak version of the 1998 engine from The Many Faces of Go vers.10. It's considered to be a nice, fairly easy introduction to Go. You can get it free at David Fotland's page: [ext] http://www.smart-games.com/igowin.html. "

There were no MCTS go playing programs in 1998. ManyFaces 10 was before ManyFaces 11 and even that wasn't MCTS based (Manyfaces 12 was the first).

It is common for software sellers to do this, give away old versions free. Nothing wrong with that and old versions and/or greatly reduced price and these free version might be all that a relative beginner needs. But don't make judgements about the strengths/behaviors of the current versions of the software based upon the versions being given away for free.

What would help when people want to talk about what a program does or doesn't do is besides the name of the program always include the version information.