Thursday, November 24, 2011

Olympiad finishes, 4th and 5th places for Fuego on large board sizes

First, the good news: Timo Ewalds' program Castro defended its title in the game of Havannah. This brings our total to three Gold medals. Congratulations Timo!

The Go competitions did not go so well. Fuego got 4th on 13x13 and 5th on 19x19. I believe that Fuego had chances in several games, but it showed a disturbing tendency to suddenly tenuki (play elsewhere) in the middle of a critical fight. I am not sure if it is a bug, or what else was going on. The last game of the tournament, against undefeated winner Zen, is a good example. Fuego put up a good fight in the center, and seemed to have Zen's groups in some trouble. Then suddenly at move 122 Fuego switched elsewhere, and completely collapsed as a result. Not good.

The game records for 13x13 and 19x19 are on KGS. Overall, the level of competition was very tough. Zen won all its games on 13x13 and 19x19, and is the undisputed number one. Erik van der Werf's Steenvreter made a strong comeback with two silver medals. The open source program Pachi, running on a large cluster, is a serious contender in every tournament now. The German program Gomorra is also greatly improved, and last year's 9x9 winner MyGoFriend can now play on the large board as well. It seems like several of the programs implemented new techniques described in our postdoc Aja Huang's recent PhD thesis. Ironically, these are not included in Fuego yet.

Next up: the UEC cup in early December. Fuego surprisingly won that event last year, but it will be extremely difficult to repeat that success.

Tuesday, November 22, 2011

Computer Olympiad: UAlberta programs rake in the medals

The Computer Olympiad is well under way. I am following it from far away. Some games are played and archived online. The news about others are slowly trickling in.

So far programs authored or co-authored by people in our games group got:
2 Gold (Hex and NoGo)
2 Silver (Hex and NoGo)
2 Bronze (Amazons and 9x9 Go)

In Hex, the two Alberta programs defended their Gold and Silver medals, which they have now held for three years in a row. Impressive!

MoHex (UofA): 7 wins 0 losses
Wolve (UofA): 4 wins 3 losses
Panoramex (Paris Dauphine) 8 losses

According to Ryan Hayward, leader of the Hex group, Panoramex played strongly and had several winning positions, but the endgame solver bailed out our programs in at least two cases.

In NoGo, two programs with UofA involvement also took Gold and Silver. Both programs are heavily based on the Monte Carlo search engine in our Fuego Go program, with modifications to implement the NoGo game rules and remove parts that are Go-specific. The Gold medal went to BobNoGo, by Bob Hearn and myself. Silver went to FxNoGo by my student Fan Xie. The competition had three strong programs, including the Taiwanese program CMNoGo, derived from the Go program ColdMilk. The top programs seem to be stronger than humans.

BobNoGo: 4 wins
FxNoGo: 2 wins 2 losses
CMNoGo: 4 losses

Game records for BobNoGo and FxNoGo.

In Amazons, our Arrow2 took third place (out of four programs) behind Invader and 8QP. It managed to win one game from 8QP. The program still has some weaknesses in the opening, getting its queens trapped. Arrow2 has been playing quite successfully against humans on the Little Golem site. There is a discussion thread about it on Little Golem.

In Go, the 9x9 and 13x13 tournaments have finished. The 19x19 will follow tonight (North American time). Fuego won a Bronze medal on 9x9. With a little bit of luck (well, what I really mean is without playing one or two bad moves) the Gold would have been in reach.

Top three:
Zen 10 wins 2 losses
pachi 9 wins 3 losses, won playoff 3:1
Fuego 9 wins 3 losses
Full results (all scores are off by two)
Fuego game records for 9x9

The 13x13 did not go too well. Fuego may have missed some chances, but I have not analyzed the games yet. Zen won ahead of stv and pachi.
Full results
Fuego game records for 13x13

You can watch the 19x19 on KGS tonight or look at the tournament records on the web. Also still to come: Havannah, where Timo Ewald's Castro is defending its title from 2010.


Saturday, November 12, 2011

TAAI 2011: Zen keeps winning; two more second places for Fuego

The TAAI tournament concluded last night with the 9x9 and 13x13 tournaments. In the 9x9 tournament, Fuego had technical problems with its opening book, and got into bad lines in several games. In round one against CyGo, the game with Fuego as Black turned into the infamous "Z" opening, which seems to be completely lost for Black according to our previous experience.
See the game on EidoGo. After move 9, Black's stones have a "Z" shape. Both players have two one-space extensions at opposite sides of the board. If all stones are safe, The extra black stone in the center loses its meaning, and Black cannot give the komi. In this game, White makes a mistake at move 18. Simply blocking from the outside at B6 would capture the black stones and easily win the game. Both games against Zen were simple losses straight out of the opening. In the last round, Fuego was matched against MoGo in what amounted to a playoff for second place. After being tied after two regular games, the programs played two more, and Black ended up winning all four games. So Fuego and MoGo shared second place. I will post the overall standings once they are announced.

The 13x13 tournament saw Fuego's best effort in this event. It achieved a winning position against Zen but lost its way in the end. See the game here. After Black lives at the bottom with move 77, Black is ahead. Fuego also counters White's invasion at the top left very strongly with move 81. If it captured one stone with 87 at D12, it would easily win the semeai. Inexplicably, Fuego switches elsewhere. Still, I think Fuego is slightly ahead until move 95. Blocking on the outside wins by about 1.5 points in my analysis. The inside block either loses points, or leads to a ko fight that Fuego cannot win. For 103, M9 seems better but still not enough to win. Overall, a good effort by Fuego, but a missed chance for a rare victory against Zen.

Final standings on 13x13 were the same as on 19x19:

1. Zen 3 wins
2. Fuego 2 wins
3. Mogo 1 win
4. Coldmilk 0 wins

Zen' games (all board sizes)
Fuego 9x9 games
Fuego 13x13 games
Fuego 19x19 games

Friday, November 11, 2011

TAAI tournament has started; Fuego second on 19x19

The TAAI Computer Game Tournaments are sponsored by the Taiwanese Association for Artificial Intelligence (TAAI) and organized by a Taiwanese group of computer games researchers. This year they are organizing three Go tournaments, on 9x9, 13x13 and 19x19. The 19x19 event was last night, with 4 participants.

1. Zen 3 wins
2. Fuego 2 wins
3. MoGo 1 win
4. Coldmilk 0 wins

As expected, Zen won all its games. Still, I am very happy with Fuego's performance. It put up a good fight in all three games. Even though the hardware and software only came together in the last minute, it seemed to run smoothly.

You can find the games on the KGS archives here: Zen, Fuego, MoGo, Coldmilk. Or replay the game of Zen against Fuego using the EidoGo service.

Tonight (Our time), the 9x9 and 13x13 tournaments will follow.

Monday, January 3, 2011

Combinatorial Game Theory Workshop in Banff

Next week I will attend the CGT workshop at BIRS. This is a meeting of researchers in combinatorial game theory, including a good number of people interested in Go. It is held every 2-3 years, and is a continuation of the workshops that started at MSRI in Berkeley. One of the special themes of this year's meeting is looking for connections between CGT and Monte-Carlo tree search. Olivier Teytaud will give an invited talk.

John Tromp won Man-Machine match 4:0

Still, a good showing by the computer program, The Many Faces of Go, which had John in trouble several times.

John won by being patient, going for complicated trades when behind, and waiting for the program's mistakes.

It would be fun to see a re-match in a year.

Some links:

Here is game 2 of the match. I believe that Many Faces was ahead after the successful attack in the top left, but lost its way in the top right. For example, at move 118 cutting at R11 would be good for the program.

PW[Many Faces of Go]PB[John Tromp]WR[2d]BR[2d]DT[2010-12-28]CP[BGA]EV[Man/Machine match]RO[2]PC[ISH, Great Portland Street, LOndon, England]WT["Machine"]BT["Man"]US[N.S.Wedd]C[BGAmatches [-\]: play will start soon, both players are ready
BGAmatches [-\]: Black is Many Faces of Go, on the same system as in round 1
underplay [5d\]: John Tromp is a UK player?
gogonuts [5d\]: how is the color determined?
BGAmatches [-\]: coin toss for round 1, thern alternate
gogonuts [5d\]: wasnt mf b the first game?
;B[dp]C[BGAmatches [-\]: yes, and now it is white
;B[pp]C[gogonuts [5d\]: thx, you said b is mf above
BGAmatches [-\]: yes, my mistake
;W[cg]C[BGAmatches [-\]: W is Many Faces
;W[gd]C[gogonuts [5d\]: classic answer this time by b
;B[dh]C[gogonuts [5d\]: the drawback is that mf also knew a classic answer to the classic answer :-)
;W[ci]C[klausP [4d\]: In the first game tromp choose e13 instead of e14 in this joseki
klausP [4d\]: mybe he looked it up
underplay [5d\]: c12 b12 d13 for simplicity
klausP [4d\]: here e10 is possible too
gogonuts [5d\]: i think he actually confused it with g13 instead of e18 in the first game
EpicPhail [2k\]: g16 at f16 normal?
gogonuts [5d\]: both normal
underplay [5d\]: both possible
klausP [4d\]: with the 44 stones b could choose e10 Immo
underplay [5d\]: how about c9 - kill the bot early :p
klausP [4d\]: the human playing with the influence would be interesting
gogonuts [5d\]: :-)
klausP [4d\]: c9 is joseki?
underplay [5d\]: doubt it :p
underplay [5d\]: but seems fun
EpicPhail [2k\]: c9 looks promising
underplay [5d\]: joseki is so restrictive - harder to kill
;B[jp]C[gogonuts [5d\]: b has reason to see himself as the stronger player
gogonuts [5d\]: so he will want to play safely
underplay [5d\]: i would c8 now as w
gogonuts [5d\]: safest was c12,b12,d13
;W[cn]C[gogonuts [5d\]: not a good aproach
gogonuts [5d\]: app
underplay [5d\]: might be ok
underplay [5d\]: hard to cut it
;B[fq]C[underplay [5d\]: if b is not careful w will get more than he deserves
gogonuts [5d\]: r6 was better for wthan c6
underplay [5d\]: b4 c3 d8
underplay [5d\]: if this, w is happy
underplay [5d\]: imho
gogonuts [5d\]: w is in danger of ending up low on the left
underplay [5d\]: but ofc b can not answer b4 and then w would be sad :(
;W[mp]C[gogonuts [5d\]: overplay
underplay [5d\]: wow, bot move
underplay [5d\]: now c9
klausP [4d\]: here we go
gogonuts [5d\]: r6 good enough for b
gogonuts [5d\]: w should have played r6 himself
klausP [4d\]: I think b choose f3, because the corner is less vulerable, If b invades at c9 and c6 seeks live in the corner
;B[pn]C[gogonuts [5d\]: f3 was ok
klausP [4d\]: but b does not need to hurry
;W[pq]C[klausP [4d\]: Q6 is fine too
EpicPhail [2k\]: q6 seems to force w into the corner where b can kill it
gogonuts [5d\]: hm
gogonuts [5d\]: which one?
gogonuts [5d\]: both attrctive :-)
underplay [5d\]: r3 seems enough
EpicPhail [2k\]: p3 and hope for corner kill
gogonuts [5d\]: yes, i would ponder p3, then play r3 :-)
klausP [4d\]: MFoG had some live and death problems in game 1
underplay [5d\]: with k4 in place r3 should make w too tight
gogonuts [5d\]: let it create its own problems
gogonuts [5d\]: dont be to smart about it
klausP [4d\]: by the way: tenuki is still real big
EpicPhail [2k\]: yes maybe John has exploited that MFOG has Problems with the q4 q6 shape
klausP [4d\]: lots of big points remaining
gogonuts [5d\]: please no :-)
;B[qq]C[klausP [4d\]: so b can play r3, and end up in sente
EpicPhail [2k\]: because he played this shape last game where it was clearly not so good
gogonuts [5d\]: will mf even find p4?
underplay [5d\]: unlikely
;W[op]C[AdrianMonk [4d\]: r3 is weak move, seems underplay
gogonuts [5d\]: yes
BGAmatches [-\]: :)
Sandmann [5d\]: in your face
klausP [4d\]: p4 is in the Joseki database
underplay [5d\]: :(
klausP [4d\]: so easy to find
underplay [5d\]: Sandmann so cruel :(
Sandmann [5d\]: xD
gogonuts [5d\]: q5,p2,r2 - nothing fancy please
underplay [5d\]: does manyfaces have joseki built into it?
;B[qp]C[BGAmatches [-\]: yes
gogonuts [5d\]: its not a slave to the patterns, but the joseki moves get a bonus in the search
underplay [5d\]: i see
gogonuts [5d\]: b wants sente himself
;W[di]C[gogonuts [5d\]: but w will take sente now
underplay [5d\]: big move
gogonuts [5d\]: yupp
underplay [5d\]: hane first think later
Sandmann [5d\]: i'd have played this without c6 and n4 exchanges
gogonuts [5d\]: how to handle? - double hane?
underplay [5d\]: double hane looks fun
gogonuts [5d\]: difficult point
;B[ei]C[gogonuts [5d\]: connect is better prob
gogonuts [5d\]: after w e10
;W[ej]C[Sandmann [5d\]: you mean e12?
gogonuts [5d\]: yes
Sandmann [5d\]: hm then why exchange e11
BGAmatches [-\]: B has used 12 minutes, W 6
gogonuts [5d\]: so w doesnt play it :-)
klausP [4d\]: I like the white moves here, they solve the c9 problem and the Price is ok
Sandmann [5d\]: :O
EpicPhail [2k\]: is o2 or o3 the vital point?
gogonuts [5d\]: connection seems ok to me because w has a hard time responding at f10
Sandmann [5d\]: ah i see
Sandmann [5d\]: i think f10 is ok
gogonuts [5d\]: for b now?
Sandmann [5d\]: for w
gogonuts [5d\]: ok
Sandmann [5d\]: i dont know what to play for b
;B[eh]C[WeshCousin [3d?\]: D9?
gogonuts [5d\]: d9 is such a human move :-)
WeshCousin [3d?\]: sure it is :p
;W[fj]C[cody101 [3d\]: lol
gogonuts [5d\]: yeah, yeah
klausP [4d\]: whereas f10 is such a bot move
gogonuts [5d\]: i hope b manages to punish f10
klausP [4d\]: w leaves a lot of aji here
klausP [4d\]: not so good
gogonuts [5d\]: direct cut?
gogonuts [5d\]: or c12 first?
klausP [4d\]: b can push at c12 and cut at d10 after that
gogonuts [5d\]: that seems like common sense
Sandmann [5d\]: i think the fight is better for w:D
klausP [4d\]: why?
Sandmann [5d\]: feeling
gogonuts [5d\]: c12 is prob the right move, because w c12 fixes everything
Sandmann [5d\]: e14 and b's cutting stones would both be weak
klausP [4d\]: I think the fight is nice on eather side, against MFoG, hehe
AdrianMonk [4d\]: h13
gogonuts [5d\]: i think b must go with klaus
BGAmatches [-\]: B has used 17 mins, W 8
;W[bh]C[gogonuts [5d\]: d10 pls
klausP [4d\]: for annyone who ist interested in how to beat MFoG with cruel moves, I opened up a game in the german corner
klausP [4d\]: extrsolid vs MFoG
;B[dj]C[gogonuts [5d\]: here we go
;W[bj]C[gogonuts [5d\]: tricky
;B[dk]C[Sandmann [5d\]: w is cool
gogonuts [5d\]: i suggest d9,f8,d7
gogonuts [5d\]: so is b
cody101 [3d\]: E8?
gogonuts [5d\]: oops, i meant e8
gogonuts [5d\]: w e8, b d7
cody101 [3d\]: yeah but theres C3 later for w
;W[cl]C[gogonuts [5d\]: nice
gogonuts [5d\]: well handled by mf
gogonuts [5d\]: abandon for now
gogonuts [5d\]: g11
klausP [4d\]: stong play by the bot
klausP [4d\]: nice shape
gogonuts [5d\]: yes
BGAmatches [-\]: B has used 21 mins, W 10
;B[lc]C[gogonuts [5d\]: too early tenuki
MXHero [3d\]: i really dislike this
Sandmann [5d\]: g12 goo
cody101 [3d\]: black has crazy ideas
;W[gi]C[gogonuts [5d\]: good, bs point
WeshCousin [3d?\]: b seems fine and ahead
gogonuts [5d\]: w came back
klausP [4d\]: m17 is localy nice, but I like g11 too
cody101 [3d\]: yeah i like G11
klausP [4d\]: e14 not settled and c9 looks wasted
MXHero [3d\]: yeah
;W[pc]C[gogonuts [5d\]: ladder against p16 works, so b can play q15 instead of p16 if he wants
EpicPhail [2k\]: how to punish n4 shape later?
MXHero [3d\]: p16 is just such a nice move though
gogonuts [5d\]: :-)
klausP [4d\]: Interesting to see, that MFoG does not play like Zen at all
;B[od]C[klausP [4d\]: maybe due to josekie Database
gogonuts [5d\]: different character :-)
klausP [4d\]: yes, completely different
;W[pf]C[klausP [4d\]: r12 now
MXHero [3d\]: hmm this looks slightly wrong
gogonuts [5d\]: q14 feels a little to active
klausP [4d\]: r12 is a nice checking move I think
gogonuts [5d\]: makes playing m15 easy for b
WeshCousin [3d?\]: R12 seems fine
klausP [4d\]: it aims at r15
gogonuts [5d\]: b r12 too early
klausP [4d\]: how should b attack the p4 stones, If he wants to do that?
;B[ke]C[gogonuts [5d\]: the solid r4 makes the continuation a little difficult
EpicPhail [2k\]: i vote for o2
gogonuts [5d\]: epic
klausP [4d\]: epic?
gogonuts [5d\]: fail :-)
;W[kq]C[EpicPhail [2k\]: ^^
;B[jq]C[MXHero [3d\]: r12 is coming up, stay tuned! :)
EpicPhail [2k\]: h17 looks interesting, too
gogonuts [5d\]: L3 is heavy
WeshCousin [3d?\]: so heavy...
klausP [4d\]: h17 may turn out to be an overplay
gogonuts [5d\]: w needs to take gote at r11
gogonuts [5d\]: so defining the bottom isnt good
;W[kp]C[Sandmann [5d\]: w should k14
gogonuts [5d\]: steal r12
EpicPhail [2k\]: j16 + e8 looks interesting
klausP [4d\]: L3 and L4 look like a kyuis aproach to settle the stones
gogonuts [5d\]: b k5 wr11 unsatisfactory for b
;B[oq]C[gogonuts [5d\]: so steal it
gogonuts [5d\]: no
MXHero [3d\]: eech, this just gives w some shape
gogonuts [5d\]: and k5 or L5 makes w strong
;W[jo]C[gogonuts [5d\]: critical situation
;B[io]C[MXHero [3d\]: b can't possibly respond to the hane
MXHero [3d\]: now it looks like he made a mistake
BGAmatches [-\]: W has used 29 mins, B 16
gogonuts [5d\]: which he did :-)
;W[nq]C[MXHero [3d\]: yes, but it's bad to admit it on the board! :)
gogonuts [5d\]: dont forget that this is labled the shodan challenge
gogonuts [5d\]: the level of play is pretty high
klausP [4d\]: I dont like what b is doing here
klausP [4d\]: but for a shodan challenge you are right
;B[pr]C[gogonuts [5d\]: b is a really strong 2d
MXHero [3d\]: oh yeah, b is good
;W[in]C[MXHero [3d\]: i just want him to play r12 so badly :)
;W[ko]C[gogonuts [5d\]: atari too early imo
;B[ho]C[MXHero [3d\]: r11 now?
;B[qh]C[EpicPhail [2k\]: nice ladder breaker
gogonuts [5d\]: k7 is great shape, but making r12 a ladder breaker is questionable
gogonuts [5d\]: r11 was a good suggestion
;W[qg]C[EpicPhail [2k\]: q12
;B[ph]C[BGAmatches [-\]: MFoG gives itself 62%
gogonuts [5d\]: p14 and L6 miai
klausP [4d\]: nice to know
MXHero [3d\]: that's interesting
gogonuts [5d\]: w was ahead a few moves ago
klausP [4d\]: 62% is a lot
EpicPhail [2k\]: F6 looks nice for w
BGAmatches [-\]: B has used 32 minutes, W 22
gogonuts [5d\]: this looks okay for b again
Sandmann [5d\]: k14 now
;W[og]C[klausP [4d\]: b4 is a big point to be take in account soon
gogonuts [5d\]: k6 can move, so it should
gogonuts [5d\]: L15 is now a little thin, m15 would have been safer
gogonuts [5d\]: but b doesnt have all day
EpicPhail [2k\]: d9 can be rescued?
MXHero [3d\]: yeah l15 made me nervous when it was played
gogonuts [5d\]: resueing d9 is a long range hope
gogonuts [5d\]: the kind of hope that some almost never becomes reality :-)
;B[kn]C[klausP [4d\]: hm
gogonuts [5d\]: but moving k6 gives d9 more meaning
yoyoma [1k?\]: I was hoping to see the relentless bloodlust of Zen :(
yoyoma [1k?\]: hehe
;B[km]C[Victim [4d\]: then the name zen isnt really appropriate
Sandmann [5d\]: furikawari m17 for l3
yoyoma [1k?\]: It's to trick unsuspecting humans
gogonuts [5d\]: just looking at the position, this could easily be human vs human
gogonuts [5d\]: m6 was depressing, wasnt it
MXHero [3d\]: yes :(
;W[hm]C[gogonuts [5d\]: some half baked squeeze would have been better, i guess
underplay [-\]: w plans to kill it all
underplay [-\]: mwahaha
Victim [4d\]: i like this play more than the sqeeze, that would have made b too think against g11
Victim [4d\]: thick
gogonuts [5d\]: attach at g7?
MXHero [3d\]: i liked w l6 instead of p13 hehe, moves the problem elsewhere
klausP [4d\]: this is rather difficult for both sides
Sandmann [5d\]: only k8
;B[gm]C[klausP [4d\]: I think that protecting Q12 with Q9 gave b a lead
yoyoma [1k?\]: gogo you are aiming at D9? I'm more concerned about the center groups
klausP [4d\]: even if it is a halfbaked move
yoyoma [1k?\]: H13/Q12/L7 all weakish?
glimz [3d\]: bots are much weaker rankwise to ppl who are used to play bots
gogonuts [5d\]: im in bargaining mode :-)
EpicPhail [2k\]: m7 +n6 are also very big
gogonuts [5d\]: g7 feels like a vital point to me
;W[gl]C[gogonuts [5d\]: f7, not bad
gogonuts [5d\]: not h6 pls
;B[fm]C[EpicPhail [2k\]: critical moment of the game
gogonuts [5d\]: if this goes badly for b i am to blame :-)
BGAmatches [-\]: B has used 37 mins,W 26
gogonuts [5d\]: thx
MXHero [3d\]: w k8 now?
EpicPhail [2k\]: m7 so big
;B[lm]C[gogonuts [5d\]: w less than happy
gogonuts [5d\]: rats, still no ladder :-)
MXHero [3d\]: haha
EpicPhail [2k\]: timesuji
klausP [4d\]: yay, blaming gogo would be fun - but b is fine here I think
;B[ir]C[MXHero [3d\]: b4 is becoming very important
EpicPhail [2k\]: j8 seems vital
;B[mm]C[gogonuts [5d\]: w building a measly mini center?
MXHero [3d\]: well this does prepare for a cut at k14 or something
;W[mn]C[gogonuts [5d\]: the w percentages must be coming down
GoGnom [5d?\]: why so keen on k14?
gogonuts [5d\]: begging above, begging below
MXHero [3d\]: i'm not, but i'd rather think of that as the goal than a center moyo :)
BGAmatches [-\]: B has used 38 mins, W 29
EpicPhail [2k\]: n8
GoGnom [5d?\]: i think k14 too agressive
gogonuts [5d\]: o7
MXHero [3d\]: i agree
;B[ml]C[EpicPhail [2k\]: k8 to fix the shape?
;W[mk]C[gogonuts [5d\]: i dont like n8
gogonuts [5d\]: b wanted to add pressure on ws center
glimz [3d\]: i also thought it's bad but didn't come up with anything else.. what would be better?
;B[nk]C[gogonuts [5d\]: but its the other way around
gogonuts [5d\]: simply o7
MXHero [3d\]: just o7 did seem much more stable, yeah
;W[lj]C[EpicPhail [2k\]: the result looks good for b - b gets sente while w has to fix his shape
glimz [3d\]: B in need of fixing
MXHero [3d\]: b does not have sente
;B[nj]C[gogonuts [5d\]: m10 is dubious to me
Sandmann [5d\]: k14 finally
GoGnom [5d?\]: yes
BGAmatches [-\]: B has used 44 mins, W 30
Sandmann [5d\]: very hot spot
MXHero [3d\]: b needs to protect again, but w does too; maybe it's ok
gogonuts [5d\]: o10 is a nice move to get
gogonuts [5d\]: but i think w should have played o10 himself
;W[id]C[gogonuts [5d\]: o14
gogonuts [5d\]: better than trying to connect
EpicPhail [2k\]: L18 yose first?
gogonuts [5d\]: too hot at the top
gogonuts [5d\]: 2 potential weak b groups
EpicPhail [2k\]: crude F16 helps fixing his group?
gogonuts [5d\]: w can answer at g15 now
;B[nr]C[gogonuts [5d\]: yikes
gogonuts [5d\]: w m2 unwelcome
GoGnom [5d?\]: bot might just not realize importance of top
glimz [3d\]: j16 wasn't that bad a bot move... hard to find a good reply
gogonuts [5d\]: i like o14
gogonuts [5d\]: has s13 next
;W[ob]C[gogonuts [5d\]: too boring
gogonuts [5d\]: o18 good for b
Dcugel [6d\]: not n2?
EpicPhail [2k\]: bot in yose mode
BGAmatches [-\]: the game is paused now, John has gone for a coffee
gogonuts [5d\]: thx
gogonuts [5d\]: toilett too?
EpicPhail [2k\]: paused means what exactly?
glimz [3d\]: maybe he has computer in the toilet ^^
BGAmatches [-\]: John is back
EpicPhail [2k\]: the bot can calculate bot the clock is stopped?
;B[nb]C[gogonuts [5d\]: looks like its a game between gentlemen
MXHero [3d\]: oh coffee was nearby!
BGAmatches [-\]: no, the bot has no idea that he went for coffee
gogonuts [5d\]: not playing q18 would be very bot-like :-)
;W[if]C[glimz [3d\]: not playing q18 also pro-like
MXHero [3d\]: hehe
EpicPhail [2k\]: here we go
MXHero [3d\]: well the top is strong now so just save the left
gogonuts [5d\]: p18 for o18 is a very poor exchange
gogonuts [5d\]: because b now has q18 at need
gogonuts [5d\]: but somehow w stumbled into the fact that b has 2 weak groups :-)
glimz [3d\]: by the looks of it.. W might get q18 in sente
MXHero [3d\]: seems like b stumbled into that fact as well :)
gogonuts [5d\]: try local measures first
gogonuts [5d\]: c14
EpicPhail [2k\]: g12?
gogonuts [5d\]: lol
gogonuts [5d\]: maybe its the coffee that has his attention
glimz [3d\]: this game W looked good (maybe even better) in the very beginning. then B seemed to turn everything around... and now W might manage a good comeback
MXHero [3d\]: isn't n4 killable?
gogonuts [5d\]: what if?
gogonuts [5d\]: and not sure
EpicPhail [2k\]: e14 might be killable too
gogonuts [5d\]: o6 is w sente
;B[cf]C[MXHero [3d\]: well ok touche, it doesn't matter yet :)
gogonuts [5d\]: found c14
Swifft [3d\]: n3 looks nice
;W[dg]C[gogonuts [5d\]: o2 is just the kind of move that gets humans in trouble against bots
EpicPhail [2k\]: b lacks shape?
MXHero [3d\]: o2 is the kind of move that gets humans in trouble against other humans! hehe
glimz [3d\]: n3 results ko i think
gogonuts [5d\]: lol
gogonuts [5d\]: e13 a must
EpicPhail [2k\]: e13 b14 and there is no shape left
klausP [4d\]: no
gogonuts [5d\]: the top right group is almost off the hook because of p18
klausP [4d\]: b13 possible
gogonuts [5d\]: b13 for b14 doesnt help
;B[bf]C[MXHero [3d\]: oh noo
BGAmatches [-\]: B has 55 minutes left (of 110), W has 75
gogonuts [5d\]: too expensive for my taste
klausP [4d\]: b getting into trouble
MXHero [3d\]: b got spooked
;W[eg]C[gogonuts [5d\]: was in trouble already
gogonuts [5d\]: but i dont think this helped
gogonuts [5d\]: only cost points
gogonuts [5d\]: payday
EpicPhail [2k\]: b18 looks alive
gogonuts [5d\]: sad
livingdead [4d\]: is this game 1?
gogonuts [5d\]: yes
BGAmatches [-\]: 2
glimz [3d\]: it's game 2
gogonuts [5d\]: 2
EpicPhail [2k\]: game 2
gogonuts [5d\]: 2 2 2 2
MXHero [3d\]: who won the first?
BGAmatches [-\]: John won game 1
MXHero [3d\]: or, what entity won the first? :P ok thanks
Swifft [3d\]: John won one
EpicPhail [2k\]: BGA any info on the %
;B[mq]C[klausP [4d\]: wow
yoyoma [1k?\]: woah
MXHero [3d\]: wow trade actually looks good now, after that mistake
gogonuts [5d\]: no
MXHero [3d\]: still no?
yoyoma [1k?\]: the black group is much bigger
gogonuts [5d\]: w killing the top is bigger and more certain
glimz [3d\]: hmm maybe no ko, just dead?
;W[bc]C[EpicPhail [2k\]: game over
iTengen [2k\]: what is the general idea of this matchup? does the computer have a chance?
;B[bd]C[MXHero [3d\]: tengen, watch what is happening :)
klausP [4d\]: oh
Sandmann [5d\]: what's the bet in this match?
glimz [3d\]: it's a $1K bet
gogonuts [5d\]: this match was set up 10 years ago
MXHero [3d\]: oh really?
Sandmann [5d\]: and how are the settings?
gogonuts [5d\]: they were speculating on the future
livingdead [4d\]: does john have a kgs rating? they say he's 2d is that is europe or
gogonuts [5d\]: it turned out to be a close call, which is fortunate for us kibitzes
glimz [3d\]: John was strong European 2d as per his 2005 EGD entry
BGAmatches [-\]: B has 51 minutes left, W has 73
Sandmann [5d\]: is it a best of x match?
yoyoma [1k?\]: Info here:
glimz [3d\]: maybe he got stronger meanwhile
;W[ba]C[EpicPhail [2k\]: i think w won this game
BGAmatches [-\]: best of 7
yoyoma [1k?\]: There is a poll there, 62% John 20% computer 17% close
Sandmann [5d\]: awesome
EpicPhail [2k\]: now b needs magic
yoyoma [1k?\]: Oh also had a short recap of the games
yoyoma [1k?\]: nice
gogonuts [5d\]: black magic :-)
livingdead [4d\]: i think its still close. depends what happens in the center, and b4 is huge too
livingdead [4d\]: but probably w a bit ahead
;B[fd]C[EpicPhail [2k\]: can b live??
gogonuts [5d\]: turning point, move in at L13 or try to struggle?
gogonuts [5d\]: struggle :-)
EpicPhail [2k\]: he chose my crude move
gogonuts [5d\]: lets see if mf plays my crude reply :-)
klausP [4d\]: L13 is a way to stay in a close game
MXHero [3d\]: if the game is close, the bot wins by endgame i'd guess
gogonuts [5d\]: f15 also possible
;W[fe]C[EpicPhail [2k\]: b just dead
gogonuts [5d\]: fits best into the current position
klausP [4d\]: the bots are not so good at live and death, but I thik mfog will kill here
;B[ff]C[Sandmann [5d\]: i think 50/50
gogonuts [5d\]: :-)
EpicPhail [2k\]: i bet he kills
MXHero [3d\]: yeah, seki :)
;W[gf]C[gogonuts [5d\]: q18, hope the bottom is dead and see what w play in the center
gogonuts [5d\]: but w L13 is huge
EpicPhail [2k\]: why not L18 first?
MXHero [3d\]: because it's not sente
;B[co]C[gogonuts [5d\]: those stones have to be dead as they stand, or b lost
gogonuts [5d\]: no way
livingdead [4d\]: question: does the computer think while the human is playing?
EpicPhail [2k\]: yes, why not
gogonuts [5d\]: c5 is much smaller than q18
Swifft [3d\]: no thinking
BGAmatches [-\]: does it "ponder"? - good question
BGAmatches [-\]: I suspect not
tigro [4d\]: whatever in two year he will have a processor two times quicker
;W[bo]C[EpicPhail [2k\]: moore´s law will not continue forever
livingdead [4d\]: seems like that would be a big help to write software that can run its algorithm on oppponents time too
Swifft [3d\]: two times quicker, doesn't mean two times better
EpicPhail [2k\]: but maybe with quantum comps there will be a jump
gogonuts [5d\]: play its moves when its the opponents turn :-)
klausP [4d\]: In think L13 is important for b to stay in the game
livingdead [4d\]: me too
;B[bp]C[MXHero [3d\]: calling quantum computers a jump is an understatement :)
gogonuts [5d\]: c5 didnt make getting to L13 easier
glimz [3d\]: from the original email: "you may run any program on any hardware you can bring to the playing site. (i realize this will be a 64Gb, 4Ghz machine or some such:) "
klausP [4d\]: the exchanges at f16 and f14 are a minus for black
glimz [3d\]: imagine how these numbers looked in 1997
iTengen [2k\]: lol
glimz [3d\]: :D
gogonuts [5d\]: sticks and stones can break my bones, but mips will never hurt me
BGAmatches [-\]: B has 45 mins left, W has 109
;W[qb]C[klausP [4d\]: mips?
;B[oa]C[livingdead [4d\]: L13!!!
glimz [3d\]: instructions per sec
gogonuts [5d\]: mf loves that stupid hanging connection
fotland [?\]: there is a memory problem preventing MF from using all of its time
gogonuts [5d\]: in this case q18 was obviously better
glimz [3d\]: machine has too little memory, david?
MXHero [3d\]: unless w wants ko now! lol
fotland [?\]: there is a playout bug that makes it like these bad hanging connections
;W[ig]C[gogonuts [5d\]: lol
klausP [4d\]: oh
livingdead [4d\]: i fear mf will throw this one away
klausP [4d\]: mfog mus be rather optimistik
gogonuts [5d\]: closing shop it seems
klausP [4d\]: L18
klausP [4d\]: some good endgame may save b
klausP [4d\]: are the bots good at engame or bad?
BGAmatches [-\]: good imho
livingdead [4d\]: i saw zen19 play the other day and it lost a TON of points in the endgame
glimz [3d\]: good, as the vars are fewer
MXHero [3d\]: frighteningly good
MXHero [3d\]: oh really? weird
;B[dn]C[gogonuts [5d\]: J13 is bot-ish for "i've won"
livingdead [4d\]: that's tru
Swifft [3d\]: depends , if they judge everything right, very good
klausP [4d\]: well, loosing tons of points and winning by 0,5 or loosing tons of points and loosing the game?
livingdead [4d\]: lost the game. it was against a 3d i think
BGAmatches [-\]: they give away tons of points in the endgame when they know it doesn't affect the result
klausP [4d\]: go Tromp!!!
;W[kb]C[EpicPhail [2k\]: L18 double senteish
;B[pb]C[gogonuts [5d\]: does tromp mean trumps in dutch?
;W[lb]C[gogonuts [5d\]: p18 capture is keeping bs hope alive
EpicPhail [2k\]: time?
;B[qe]C[BGAmatches [-\]: B has 41 mins, W has 65
gogonuts [5d\]: its strange, but the position looks a lot more bloodthirsty than how the game progressed
klausP [4d\]: oh
;W[re]C[tigro [4d\]: thers a ko about m6 no? with o6 p7 n2?
EpicPhail [2k\]: all in?
MXHero [3d\]: r15 was a big aji keshi move
gogonuts [5d\]: as long as its big :-)
MXHero [3d\]: :)))
MXHero [3d\]: and sente too! hehe
moi [2d\]: who can talk ? i wonder (sorry it's not about the game)
EpicPhail [2k\]: who has counted?
;W[rf]C[tigro [4d\]: o6 p7 n2 is ko for white
glimz [3d\]: is it?
MXHero [3d\]: hmm o6 p7 n2 m3?
EpicPhail [2k\]: this has backfired s12 needs fixing?
klausP [4d\]: but S14 is good for s12
tigro [4d\]: m2
MXHero [3d\]: then capture
tigro [4d\]: m4
BGAmatches [-\]: B has 38 minutes left, W has 63
EpicPhail [2k\]: 38 minutes are too less time
MXHero [3d\]: connect
tigro [4d\]: n1
MXHero [3d\]: doesn't appear to be an eye on the bottom
MXHero [3d\]: l1 in response
klausP [4d\]: can b make something out of the b7 aji?
horstmaria [3d\]: n2 m3
;B[pg]C[papohasu [1d\]: why playing something around N2? white looks already dead
;W[qf]C[MXHero [3d\]: well it's very close to alive on the bottom
EpicPhail [2k\]: MFOG %?
;W[rh]C[Sandmann [5d\]: i wonder if mfog is an anagramm
glimz [3d\]: the acronym?
GoGnom [5d?\]: omfg^^
;B[ri]C[EpicPhail [2k\]: ko
;W[nf]C[klausP [4d\]: ko is still a problem for bots
;B[oe]C[EpicPhail [2k\]: zen played a long ko very nicely
horstmaria [3d\]: bad atari
turk [6d\]: w has r11 cut
horstmaria [3d\]: b has s17
turk [6d\]: both r15 and s11 are bad moves
turk [6d\]: but b may actually gain from them...
;W[mc]C[turk [6d\]: o14 is very bad for w
EpicPhail [2k\]: bot goes for the kill
glimz [3d\]: why is o14 bad for W?
turk [6d\]: o13 captures
glimz [3d\]: what aji does it keshi?
livingdead [4d\]: 014 makes s17 work
MXHero [3d\]: becuse now s17 works
MXHero [3d\]: because*
klausP [4d\]: cause Q16 atari might be needed
BGAmatches [-\]: MFoG thinks 68%
turk [6d\]: w may lose the cut at r11
goIngo [-\]: My ManyFaces gives 72 % for White here
papohasu [1d\]: black is gonna need o13 at some point
EpicPhail [2k\]: turk who leads now?
;B[sh]C[turk [6d\]: hard to say
turk [6d\]: seems like both tried to lose :-)
MXHero [3d\]: :))
glimz [3d\]: b seems to have had greater success
BGAmatches [-\]: B has 33 minutes left, W has 59
BGAmatches [-\]: they each started with 110
MXHero [3d\]: is it absolute time?
BGAmatches [-\]: yes
gogonuts [5d\]: s17 is bs ray of hope
turk [6d\]: if w cuts at r11 with his o14, the game was oer
turk [6d\]: over
EpicPhail [2k\]: 33 minutes are just not enough time for this position and some 50++ moves to play
;W[rg]C[klausP [4d\]: oh
klausP [4d\]: can b kill now?
EpicPhail [2k\]: s17?
goIngo [-\]: strange s13
livingdead [4d\]: w aiming to cut i think
goIngo [-\]: my manyfaces does not like it
MXHero [3d\]: wait, did r11 work?
turk [6d\]: yes
GoGnom [5d?\]: how? not just s10?
turk [6d\]: s10, r10
livingdead [4d\]: r11 totally works. look at 08 and then p8 - both sente
turk [6d\]: note the cut at o8
livingdead [4d\]: word
EpicPhail [2k\]: p11?
livingdead [4d\]: i think the correct sequence here is something like bQ10, wS17, bO13
gogonuts [5d\]: uneviably difficult position in a sudden death match
MXHero [3d\]: maybe just p12 for b now?
MXHero [3d\]: if r11 is a problem?
igoyametai [4d\]: s17 kills imo
turk [6d\]: trade, not kill outright
;B[oi]C[goIngo [-\]: after Black s17 Many Faces saw 64 % for White
BGAmatches [-\]: B has 29 mins, W has 35
goIngo [-\]: Now White at 71 % again
EpicPhail [2k\]: hot blitz phase
goIngo [-\]: candidate o12
;W[ng]C[klausP [4d\]: hm
livingdead [4d\]: whoa!
klausP [4d\]: tough choice
goIngo [-\]: o13 with same %
klausP [4d\]: b nust kill now
Werfeus [4d\]: uiotuj9pityh767yuj4u74yyu'rtu7y* RTNMJ
turk [6d\]: now b should win, but...
papohasu [1d\]: s18?
EpicPhail [2k\]: huge center
livingdead [4d\]: s17 for the win!!! right/
igoyametai [4d\]: s17 s18 r17 t17 q16 s16 q17 t13 r19 s10 r11 t11 r9
turk [6d\]: right
goIngo [-\]: ManyFaces does not see his group in danger
klausP [4d\]: Werfeus, are U ok?
goIngo [-\]: or "her" group
gogonuts [5d\]: mf says s17 is humbug :-)
turk [6d\]: but the wonderful world of nidans is full of surprises
klausP [4d\]: lol
EpicPhail [2k\]: turk s17 kills?
glimz [3d\]: the nidans say ni
turk [6d\]: yes, eric
klausP [4d\]: how to kill turc?
turk [6d\]: s17
EpicPhail [2k\]: open a demo ^^
klausP [4d\]: s18
goIngo [-\]: manyface has one problem in t inming routine
turk [6d\]: r17
klausP [4d\]: s16
Werfeus [4d\]: sudden expression of what i think of lag. about the same as d10 in this game.
goIngo [-\]: when it is down to 5 min it starts superblitz
turk [6d\]: take
klausP [4d\]: t17
horstmaria [3d\]: r17 t17 q16 s16
turk [6d\]: connect
horstmaria [3d\]: q17 1 eye
;B[rb]C[papohasu [1d\]: there
klausP [4d\]: r18
papohasu [1d\]: s18
gogonuts [5d\]: hm
MXHero [3d\]: oh nooo
gogonuts [5d\]: s17 worked fine imo
turk [6d\]: now big ko ^^
papohasu [1d\]: doesn't it kill by ko?
papohasu [1d\]: ok
gogonuts [5d\]: only 1 eye if w squeezed
igoyametai [4d\]: this is ko
MXHero [3d\]: yeah s17 totally worked :(
goIngo [-\]: white % falling now
turk [6d\]: see what i mean aobut the world of nidan ^^
goIngo [-\]: from 70 to 65
BGAmatches [-\]: B 27 mins, W 57
goIngo [-\]: 64
;B[qa]C[livingdead [4d\]: if w connect might he win the race / seki?
bleen [8k\]: who won
LazyDragon [3k\]: black of course
bleen [8k\]: nice
hugifrb [20k?\]: 2:0 for human?
bleen [8k\]: yes
;W[qc]C[EpicPhail [2k\]: both have some reasonable ko threats
gogonuts [5d\]: straight kill was too boring :-)
gogonuts [5d\]: palcement
gogonuts [5d\]: almost
goIngo [-\]: expectiong t16
goIngo [-\]: by MnayF
livingdead [4d\]: b has to play t16, but then w s19 and t17 are sente, and then the race.....
EpicPhail [2k\]: I dont want to be the human with this little time here right now
gogonuts [5d\]: and it should :-)
turk [6d\]: w has s10 to extend liberty
gogonuts [5d\]: but b might lose
igoyametai [4d\]: t17 for ko
turk [6d\]: but it's a big if that MF would see it
gogonuts [5d\]: hard to approach
papohasu [1d\]: hard ko for clack t17
klausP [4d\]: t16 just kills
papohasu [1d\]: black*
gogonuts [5d\]: s17 would really have been preferable :-)
goIngo [-\]: mf sees line t16 wt17
underplay [-\]: b cant win race, n16 now perhaps
turk [6d\]: t16, s19
Werfeus [4d\]: t16 and b dies
livingdead [4d\]: i think bt16 loses for b. he has to play ko
BGAmatches [-\]: B has 25 minutes, W has 56
klausP [4d\]: yes, t116 kill black
goIngo [-\]: t16 wt17 s19 n11
goIngo [-\]: manyfaces main line
livingdead [4d\]: no w throw in at s19 first
klausP [4d\]: this IS the shodan challenge after all
EpicPhail [2k\]: what about t17 now?
livingdead [4d\]: i think b has to play t17
turk [6d\]: both trying to be shodan?
klausP [4d\]: jup
goIngo [-\]: after Black t17 74% for white
klausP [4d\]: :)
goIngo [-\]: t17 wt16
turk [6d\]: now i understand their play
klausP [4d\]: to bad b missed s17
igoyametai [4d\]: n18 for life
goIngo [-\]: bkacj running into time trouble
goIngo [-\]: black
gogonuts [5d\]: yes, try n16
igoyametai [4d\]: n18 n16 o17 n19 n15
turk [6d\]: n18 lives, but loses the game
EpicPhail [2k\]: too less time
;B[me]C[turk [6d\]: unless w tries to kill...
goIngo [-\]: mf 68 % for white
goIngo [-\]: white m13 ?!
BGAmatches [-\]: B has 22 minutes left, W has 55
EpicPhail [2k\]: w+t
;W[mh]C[MXHero [3d\]: now kill
goIngo [-\]: mf expects n18 here
goIngo [-\]: my manyf
EpicPhail [2k\]: i would play n16 as b
gogonuts [5d\]: the smell of seki is in the air
;B[mb]C[goIngo [-\]: my monitor smells neutral ;-)
goIngo [-\]: even from 1 cm distance
;W[ld]C[underplay [-\]: KILL!
goIngo [-\]: with 63 % for white
MXHero [3d\]: ok now the kill has to be possible, isn't it?
klausP [4d\]: mfoG seems unsuspicious
underplay [-\]: kill easy now
goIngo [-\]: mf expecting t16 now with 40% for black
WeshCousin [4d?\]: O17
;B[md]C[underplay [-\]: b very conservative
gogonuts [5d\]: i liked o17 better too
igoyametai [4d\]: s19 t19 t17
goIngo [-\]: now white s19 ?!
MXHero [3d\]: yeah the kill totally works now :(
BGAmatches [-\]: B has 20 mins, W 53
;W[ra]C[underplay [-\]: s19 not good, negative points
;B[sa]C[klausP [4d\]: maybe b is thinking, that if the bot does not protect, it will never smell the kill
MXHero [3d\]: it removes a b eye
goIngo [-\]: L17 for white ?!
goIngo [-\]: with 63 %
underplay [-\]: no it doesnt?
underplay [-\]: how to get eye?
gogonuts [5d\]: L17?
goIngo [-\]: manyf proposal
MXHero [3d\]: if w just t17, b s19 is sente against the group, andthen b can t18
;W[ma]C[klausP [4d\]: oh
gogonuts [5d\]: wow
underplay [-\]: but t16 was the move
klausP [4d\]: t16 kill now
gogonuts [5d\]: o17
goIngo [-\]: t16 40% for black
klausP [4d\]: bt16 w t17 b Q16
gogonuts [5d\]: t16 is gote
MXHero [3d\]: underplay, then it's a ko
gogonuts [5d\]: so first o17
;B[sd]C[underplay [-\]: there is no ko?
goIngo [-\]: black times?
gogonuts [5d\]: greed
klausP [4d\]: remember: shodan challenge
gogonuts [5d\]: o17 was alive
goIngo [-\]: t17 40% white
igoyametai [4d\]: W gets seki at best
BGAmatches [-\]: Tromp: "something's going to die, and that'll be it. I don't know what"
;W[se]C[MXHero [3d\]: wise words :)
;B[sg]C[gogonuts [5d\]: if w gets L17 and seki, then t16 was useless
horstmaria [3d\]: thats easy
gogonuts [5d\]: oops
igoyametai [4d\]: abort abort :)
goIngo [-\]: mf smells something now, down to 57 %
turk [6d\]: funny how the game turned
goIngo [-\]: 56
klausP [4d\]: 57% lol
gogonuts [5d\]: t15 wasnt salvation
turk [6d\]: or it is all to be expected...
goIngo [-\]: down from 62
goIngo [-\]: 55 %
klausP [4d\]: captain, we are sinking!
;B[ip]C[goIngo [-\]: L17 proposed by my mf
WeshCousin [4d?\]: w+time
papohasu [1d\]: with which percentage does your many faces resign?
glimz [3d\]: there is enough water for all passengers
GoGnom [5d?\]: waht are you sinking about?
igoyametai [4d\]: W s10 was too deep in the search tree
goIngo [-\]: j3 horizon effect
gogonuts [5d\]: J3 is a very good sign for humanity :-)
goIngo [-\]: yes
MXHero [3d\]: what is the horizon effect?
goIngo [-\]: thumbs up for john
goIngo [-\]: going into side battles
;W[is]C[goIngo [-\]: when you dont see good continuation
klausP [4d\]: if I died there, I must be able to live somwhere alse, grm
klausP [4d\]: like a human
tigro [4d\]: it would have been hard witht17 instead t15
klausP [4d\]: could be me
;B[hr]C[glimz [3d\]: bots don't think tha way ^^
goIngo [-\]: they do in some sense
iTengen [2k\]: isn't that dead?
igoyametai [4d\]: having a computer beting humans at 19x19 GO would be actually great
glimz [3d\]: only a question of time
goIngo [-\]: down to 54%
klausP [4d\]: bots still dont "think", the ponder, as mentioned earlier
BGAmatches [-\]: B has 19 mins, W has 48
MXHero [3d\]: pondering is nice too :)
goIngo [-\]: bots do not pardon they pondon ;-)
;W[le]C[klausP [4d\]: 19 mins is not a lot to finish the game
goIngo [-\]: 51 % for black now
goIngo [-\]: no good sign for bot team
;B[nc]C[MXHero [3d\]: b6 is getting very big
goIngo [-\]: white m14 now ?
gogonuts [5d\]: there is one good sign for the bot: johns time
goIngo [-\]: right
EpicPhail [2k\]: b just needs to play honte now
BGAmatches [-\]: he knows :)
;W[kd]C[klausP [4d\]: engame should be easy now
;B[ne]C[EpicPhail [2k\]: o15 enoug
EpicPhail [2k\]: yes
livingdead [4d\]: that's game
goIngo [-\]: black playing safely now
goIngo [-\]: 48 % for white
goIngo [-\]: accordibng to mf
EpicPhail [2k\]: next move is o4
livingdead [4d\]: n2?
klausP [4d\]: if b adds a move round o4, bot will maybe res
;W[np]C[moi [2d\]: M3?
MXHero [3d\]: yeah
gogonuts [5d\]: good call
goIngo [-\]: o7 now expected
goIngo [-\]: by mf
snowmen [3d\]: tricky
BGAmatches [-\]: B has 17 mins, W 45
;B[lq]C[gogonuts [5d\]: m3
gogonuts [5d\]: simple
sum [5d\]: after this kill, B only needs 2 min thinking time to win
EpicPhail [2k\]: now 2 or 3 rocksolid moves and the game is unlosable
;B[mr]C[goIngo [-\]: mf % now in free fall, below 46%
;B[qj]C[WeshCousin [4d?\]: (is there like a video broadcast ?)
BGAmatches [-\]: no
WeshCousin [4d?\]: (ok)
MXHero [3d\]: b6 is so big
igoyametai [4d\]: b7 better
MXHero [3d\]: b6 for w
;W[sj]C[EpicPhail [2k\]: no need for big moves
cody101 [3d\]: lol whtie pro tesuji
sum [5d\]: on 19x19, MF playing stronger than MC programs
;B[si]C[goIngo [-\]: sorry, mf is mc-based
sum [5d\]: pure MC programs
gogonuts [5d\]: the strong programs arent "pure"
goIngo [-\]: there are no pure mc-bots
igoyametai [4d\]: MC programs work well only on short depth games
EpicPhail [2k\]: time?
;B[pi]C[BGAmatches [-\]: B 16, W 41
BGAmatches [-\]: minutes left, that is#
;W[sc]C[goIngo [-\]: 46% white
;B[ra]C[igoyametai [4d\]: 49% probability of winning?
gogonuts [5d\]: a lot more than expected
goIngo [-\]: now 523% for white
goIngo [-\]: 52 %
gogonuts [5d\]: playout wins
EpicPhail [2k\]: ok 16min enough for 100 easy moves
papohasu [1d\]: considering every dumb move john could do
;W[hq]C[igoyametai [4d\]: winning by time is possible still :)
;B[gq]C[gogonuts [5d\]: 523%, thats impressive
goIngo [-\]: typo
gogonuts [5d\]: sure
papohasu [1d\]: but if it could consider the really possible moves it would be under 20%
papohasu [1d\]: ^^
;W[hs]C[goIngo [-\]: yes, the best moments are over
;B[hp]C[EpicPhail [2k\]: w fishing in the dark
goIngo [-\]: falling to 43 %
goIngo [-\]: for white
gogonuts [5d\]: mf thinks its doing ok, but its playing nothing but kothreats?
ben0 [6d\]: k8....
BGAmatches [-\]: B 16 mins left, W 37
cody101 [3d\]: only the dark has fishes
sum [5d\]: after quite some middle game adventures, B found impressive exchange and kill strategy, wonder if he did so expecting W to weak at it
gogonuts [5d\]: it was giving itself a 1% chance per ko threat :-)
BGAmatches [-\]: so weak?
;B[nl]C[goIngo [-\]: 38 %
underplay [-\]: ouch
goIngo [-\]: and falling
igoyametai [4d\]: at least W does not play yose
gogonuts [5d\]: subtract 1% for every correcly answered ko threat :-)
sum [5d\]: yes, weak (compared to B): B turned unsecure position into safe win pos
LocoRon [1d?\]: It's in ddk mode now. Play inane moves hoping for mistakes. :(
;W[om]C[igoyametai [4d\]: if (!winning) avoidYose();
livingdead [4d\]: when is game 3
underplay [-\]: p6
gogonuts [5d\]: tomorrow at10?
;B[pm]C[yoyoma [1k?\]: Game 3 schedule: See BGAmatches info
goIngo [-\]: 11h germnan time
BGAmatches [-\]: game 3 10:00 tomorrow (GMT)
yoyoma [1k?\]: If you look at the info, you don't have to do timezone math :)
glimz [3d\]: function resign(): return false //does not help us :)
gogonuts [5d\]: q7: subtract one more percent :-)
igoyametai [4d\]: I am curios if MFG considers o4 alive
sum [5d\]: written in C?
;W[dm]C[glimz [3d\]: pseudocode
sum [5d\]: i mean MF
;B[em]C[MXHero [3d\]: b6 b6!
EpicPhail [2k\]: glimz [3d\]: function resign(): return false //does not help us :) xDDDD
igoyametai [4d\]: better wait for b6 and ko
goIngo [-\]: white stably 40- %
MXHero [3d\]: b6 is sente
;B[do]C[MXHero [3d\]: now it's not
igoyametai [4d\]: W should invent a ko and win by time
MXHero [3d\]: oh, it still is whoops
BGAmatches [-\]: B has 15 mins, W has 34
goIngo [-\]: 36 %
goIngo [-\]: 35
;B[dl]C[igoyametai [4d\]: it seems MFG programmers did not investigate the psichology of winning a lost game :)
EpicPhail [2k\]: b3 aji?
goIngo [-\]: 33%
gogonuts [5d\]: no
goIngo [-\]: 32
gogonuts [5d\]: only aji is a4 and b cant block
;B[on]C[igoyametai [4d\]: p6 with 100%
gogonuts [5d\]: p5 wasnt even sente
goIngo [-\]: 29 %
goIngo [-\]: 28
sum [5d\]: when will resign? at 0.0001%?
;W[nn]C[gogonuts [5d\]: lol
EpicPhail [2k\]: glimz [3d\]: function resign(): return false //does not help us :)
underplay [-\]: minus 30%
;B[cm]C[EpicPhail [2k\]: m4
;B[qo]C[goIngo [-\]: back to 36 %
;W[pl]C[EpicPhail [2k\]: m4!!!!
goIngo [-\]: back to 39 %
glimz [3d\]: I wonder what the bot imagines it can kill/save
gogonuts [5d\]: s8
goIngo [-\]: times ?
igoyametai [4d\]: better play till the end and count - at any time there is a positive probability that the game ends due to unexpected events (alien attack etc.) :)
BGAmatches [-\]: B has 13 mins, W has 31
goIngo [-\]: mf expects r8
;B[qk]C[gogonuts [5d\]: s8 looks safer
goIngo [-\]: 48 % !?
MXHero [3d\]: the aliens will only attack if MFoG is about to win, since that is proof that our technology is getting too powerful :)
goIngo [-\]: for white
gogonuts [5d\]: lol, live seperately
gogonuts [5d\]: seems to also work
;W[ql]C[underplay [-\]: everything works for b
igoyametai [4d\]: s7 99%
horstmaria [3d\]: s14 has only 2 libs
underplay [-\]: as long as he keeps coffee intake up
gogonuts [5d\]: lol
;B[rm]C[EpicPhail [2k\]: play t14 and end it
goIngo [-\]: 43 % for w
;W[rl]C[gogonuts [5d\]: tricky now
;B[qm]C[cuthalion [1k\]: i think the human controller should pull the plug, imo
gogonuts [5d\]: w wont play p8
gogonuts [5d\]: but q9
MXHero [3d\]: it's interesting to see bot logic though
igoyametai [4d\]: still - MFG plays amazing for a bot
;W[ol]C[goIngo [-\]: back to downtown: 37 %
sum [5d\]: much better than some years ago
gogonuts [5d\]: crisis over
MXHero [3d\]: it's very fascinating that the bot can play stuff like this and -also- play 2d+ tesujis
glimz [3d\]: he should shut down the machine gently and pretend everything is OK. he still needs it tomorrow. then he can go out and scream.
BGAmatches [-\]: B has 12m, W has 29
gogonuts [5d\]: q9 pls
;B[pk]C[glimz [3d\]: is q9 even needed?
;W[sl]C[igoyametai [4d\]: q9 unecessary
EpicPhail [2k\]: s10
;B[lp]C[sum [5d\]: how many moves are for virtual pass-fight? 200?
EpicPhail [2k\]: destroy any hope
goIngo [-\]: 31 %
goIngo [-\]: for white
;W[lo]C[SqueezeIt [3d\]: will this guy play all 7 games?
ben0 [6d\]: 0% for white -___-
;B[rk]C[gogonuts [5d\]: lol
goIngo [-\]: match will end after 4-vs-x
klausP [4d\]: time?
;B[oj]C[igoyametai [4d\]: 0.0000001% actually - tromp might resign without reason
BGAmatches [-\]: B has 11 mins, W 27m
goIngo [-\]: 24 % for w
SqueezeIt [3d\]: oh w wanna win by time
SqueezeIt [3d\]: 25% :D
;B[dm]C[EpicPhail [2k\]: 11min = 66 moves
EpicPhail [2k\]: easy moves
igoyametai [4d\]: winning by time is a human strategy - should be forbidden to software
cuthalion [1k\]: w is correcting it's win % based on the fact that mr tromp didn't get a flu shot this year
gogonuts [5d\]: 25% chance of winning by time :-)
ben0 [6d\]: " oh, i missclicked on the resign button ! *
goIngo [-\]: 25% means almost lost - against other bot
;B[cj]C[glimz [3d\]: tenuki now
gogonuts [5d\]: lol
;W[bi]C[goIngo [-\]: 21 %
iTengen [2k\]: hehe
;B[mf]C[gogonuts [5d\]: next n14
BGAmatches [-\]: B 10m, W 25m
sum [5d\]: j8 also nice
goIngo [-\]: gain over
horstmaria [3d\]: k8 is left
igoyametai [4d\]: B m11 looks OK
underplay [-\]: m11 fun too
SqueezeIt [3d\]: go w, go! put up a complicated fight
;W[ls]C[underplay [-\]: gote
igoyametai [4d\]: pas
gogonuts [5d\]: almost sente
;B[lf]C[glimz [3d\]: now is W's chance!
SqueezeIt [3d\]: made a ko threat
EpicPhail [2k\]: K1 !!!!!!
glimz [3d\]: play two moves in a row quickly!
LocoRon [1d?\]: This is becoming sad to watch. Like a broken up Johnny 5.
;B[kg]C[igoyametai [4d\]: life and death is the bane of all bots
;B[hn]C[EpicPhail [2k\]: darren should resign
MXHero [3d\]: i'd think they would be so good at l&d
igoyametai [4d\]: is not fair to ask a bot btro resign
underplay [-\]: time is short so fair to play on
;W[jg]C[underplay [-\]: although no mess possible :p
;B[ie]C[MXHero [3d\]: o.O
goIngo [-\]: back to 40 %
igoyametai [4d\]: B mistakes yay
gogonuts [5d\]: fairly expensive
underplay [-\]: spoke too soon
klausP [4d\]: w should bres
LocoRon [1d?\]: That was probably intentional. He probably doesn't even care at this point.
;W[kf]C[klausP [4d\]: res
;B[jd]C[MXHero [3d\]: he should care
gogonuts [5d\]: hallucinbation
;B[ic]C[SqueezeIt [3d\]: maybe bot doesn't see
underplay [-\]: 1,2 vs 1,2,3
glimz [3d\]: this b for b b/c he loses time
gogonuts [5d\]: b just wasting his good time
;B[er]C[MXHero [3d\]: oh lord
SqueezeIt [3d\]: is it 8 minutes already?
;W[he]C[goIngo [-\]: 37 % for w
;B[ao]C[MXHero [3d\]: e2 was too automatic
LocoRon [1d?\]: I'd be pretty ticked off at this point.
goIngo [-\]: 43 %
;B[im]C[sum [5d\]: ladders...
goIngo [-\]: 46 %
BGAmatches [-\]: B 10m, W 20m
;B[jl]C[gogonuts [5d\]: lol?
EpicPhail [2k\]: over
goIngo [-\]: 49 %
goIngo [-\]: drama ?
;W[jj]C[igoyametai [4d\]: now yose needs more than 10 min to finish
EpicPhail [2k\]: haha bot wants to win on time 49% winning chance to do so
;B[gk]C[sum [5d\]: W thinks l1 eye?
MXHero [3d\]: maybe w predicts b will play l1 :)
;B[hl]C[goIngo [-\]: 52 % for w
gogonuts [5d\]: mayor w misscalculation
glimz [3d\]: b l1 fantastic tesuji :)
;B[ij]C[goIngo [-\]: 48 now
EpicPhail [2k\]: B plz play L2
;W[jk]C[goIngo [-\]: 44 % for w
;B[ac]C[gogonuts [5d\]: 26% b loses on time + 26% b plays L1
EpicPhail [2k\]: I would not play the best moves now but the most hope destroying ones
glimz [3d\]: what is the exact meaning of this percentage indicator in mfog?
MXHero [3d\]: hahaha
;B[ik]C[MXHero [3d\]: ohhh dang
gogonuts [5d\]: playout percentage
gogonuts [5d\]: not winningchance
;B[ii]C[goIngo [-\]: right
MXHero [3d\]: are they mc playouts?
gogonuts [5d\]: yes
MXHero [3d\]: gotcha
;B[ih]C[gogonuts [5d\]: so even in a terrible position you get 20%
MXHero [3d\]: yeah
;B[gj]C[glimz [3d\]: i would like a mcplayout for breakfast
BGAmatches [-\]: w 10m B 16m
;W[sk]C[goIngo [-\]: falling below 40 now
;B[na]C[MXHero [3d\]: so is the general goal to play the move with the highest playout percentage? or are there other factors too?
gogonuts [5d\]: if your way ahead against a bot and it wont resign take out the stones you captured, that stabelizes the playouts
;B[so]C[igoyametai [4d\]: the bot consider m5 alive
goIngo [-\]: 34 % w
;B[ro]C[EpicPhail [2k\]: maybe someone should tell this advice to tromp
;B[cp]C[MXHero [3d\]: hah, i guess that in many of the playouts b does indeed play l1
igoyametai [4d\]: tromp needs no advice :)
goIngo [-\]: resign sequence
;B[hk]C[igoyametai [4d\]: W gote
EpicPhail [2k\]: play L2
igoyametai [4d\]: B g12
goIngo [-\]: 32 %
;B[ap]C[SqueezeIt [3d\]: how much time left? :D
;B[lk]C[yoyoma [1k?\]: 33% chance -- K1, L2 = black win L1 = white win
MXHero [3d\]: yeah
;B[bk]C[klausP [4d\]: this is sad
gogonuts [5d\]: a9, lol
BGAmatches [-\]: MFoG resigns
glimz [3d\]: a9 the same thing that causes hanging connections"?
cuthalion [1k\]: ty
EpicPhail [2k\]: congratz
igoyametai [4d\]: is not sad - W not human
klausP [4d\]: after all
underplay [-\]: congrats human!
glimz [3d\]: thx for relaying
gogonuts [5d\]: yes, thanks
MXHero [3d\]: thank you
EpicPhail [2k\]: with a better computer tomorrow w will put up a better fight
glimz [3d\]: is the comp going to be changed?
MXHero [3d\]: it was close for a long time!
igoyametai [4d\]: how many cores today?
goIngo [-\]: congrats to John
EpicPhail [2k\]: it was fairly close before w lost S15 group
glimz [3d\]: W was leading, no?
gogonuts [5d\]: w was leading after it killed the top left
gogonuts [5d\]: o2 lost the game - for a while
EpicPhail [2k\]: I still hope for ZEN as a secret weapon
BGAmatches [-\]: tomorrow MFoG will be running on a more powerful platform
gogonuts [5d\]: ah
gogonuts [5d\]: suspense building :-)
yoyoma [1k?\]: Remotely?
EpicPhail [2k\]: yes
glimz [3d\]: cannot be remotely acc/to agreement
glimz [3d\]: ?
BGAmatches [-\]: today it ran locally, tomorrow it will rewnt power from Amazon
gogonuts [5d\]: set the multicore flag to true :-.)
BGAmatches [-\]: John has relented on that agreement
yoyoma [1k?\]: They changed the agreement many times, they have been very flexible
gogonuts [5d\]: but thats agains tthe agreement isnt it?
glimz [3d\]: ah, nice.. :)
BGAmatches [-\]: they have agreed to change the agreement
EpicPhail [2k\]: they both agreed on it
underplay [-\]: why so many agreements?
gogonuts [5d\]: because theyre gamblers :-)
underplay [-\]: seems it doesn't really matter - its just interesting to see bot play - result irrelevant