(;GM[1]FF[4]CA[UTF-8]AP[CGoban:3]ST[2]
RU[Chinese]SZ[9]KM[7.50]TM[1800]
PW[Orego]PB[valkyria]DT[2008-06-19]PC[The KGS Go Server at http://www.gokgs.com/]C[Orego [-\]: GTP Engine for Orego (white): Orego version 6.04c orego.uct.UctPlayer
valkyria [-\]: GTP Engine for valkyria (black): ValkyriaThread version 3.2.1
]RE[B+Resign]
;B[ee]BL[1799.57]C[mundungus [6k\]: Ah, there we go!
]
;W[dd]WL[1754.726]C[mundungus [6k\]: What did the trick?
zapp [?\]: mode=custom was bad
mundungus [6k\]: Interesting!
zapp [?\]: I think it means it start a challenge
mundungus [6k\]: Does Valkyria ponder?
zapp [?\]: But maybe opponent name must be there
zapp [?\]: Yes
]
;B[de]BL[1699.361]
;W[ce]WL[1754.556]C[zapp [?\]: INot much time to ponder there
]
;B[cf]BL[1678.463]C[zapp [?\]: Do you have an opening library?
mundungus [6k\]: No, do you?
zapp [?\]: Just E5 and White follow ups
zapp [?\]: Hardcoded...
]
;W[bf]WL[1708.186]C[mundungus [6k\]: I recently added a feature where, when it becomes my turn, if I've already done N (=2 million) playouts through this branch, I just play without further thinking.
mundungus [6k\]: Ooh, b4 was a bit aggressive.
zapp [?\]: How many games per second do you do?
]
;B[ed]BL[1599.649]C[mundungus [6k\]: Hmmm, I'd have to look. It's pretty good, because I've reimplemented Lew's EGO library in Java.
mundungus [6k\]: I'm also running 4 threads on a 4-core processor.
]
;W[cg]WL[1661.854]
;B[df]BL[1583.959]C[zapp [?\]: I am on Pentium M 1.4 GHz single Core
]
;W[dg]WL[1615.503]C[mundungus [6k\]: I do have a hardware advantage. (3 GHz cores.)
]
;B[be]BL[1554.135]C[mundungus [6k\]: Yup, I was wondering when black would get around to that
]
;W[eg]WL[1569.149]C[zapp [?\]: Valkyria exppected wB5 instead of wD3
mundungus [6k\]: Orego now thinks black has pulled (slightly) ahead for the first time.
]
;B[bg]BL[1510.724]
;W[ff]WL[1568.94]C[zapp [?\]: Valkyria disagrees with that (69%)
mundungus [6k\]: 69% for black? Ouch.
zapp [?\]: This early it happens it is wrong
mundungus [6k\]: Do you spend more time in the opening than in the later game?
zapp [?\]: Yes, it is a funtion of the time remaining
zapp [?\]: It is a simple if (timeleft > X second) then think for Y seconds
]
;B[gc]BL[1346.061]C[zapp [?\]: So it is handtuned
zapp [?\]: But it thinks faster or slower depending on how clear the situation at root is
mundungus [6k\]: Interesting. Orego just divides the remaining time by the number of vacant points (with some buffer)
]
;W[af]WL[1521.225]C[mundungus [6k\]: Yech!
]
;B[ae]BL[1319.979]C[zapp [?\]: That move is not so bad
mundungus [6k\]: ?
]
;W[fd]WL[1473.489]C[zapp [?\]: It is often ok to sacrifice one more stone, because the black shape become a little heavy after the capture
]
;B[fc]BL[1304.925]C[zapp [?\]: F7 Was played quick becuase it was forced for example
]
;W[gd]WL[1425.765]
;B[fe]BL[1266.047]
;W[ge]WL[1378.078]
;B[hc]BL[1249.111]C[mundungus [6k\]: So you don't do any thought if the heuristics find a forced move?
]
;W[ec]WL[1330.352]
;B[cd]BL[1233.595]C[mundungus [6k\]: Orego has black winning 63%
zapp [?\]: The forced move is defined if 95% of the games go to single move after some safe time of thinking
zapp [?\]: 82%
]
;W[gf]WL[1282.642]C[mundungus [6k\]: Well, I need to go. Can you email me the SGF when it's over?
zapp [?\]: Sure
]
;B[eb]BL[1202.834]C[mundungus [6k\]: Thanks, and barring a miraculous tournaround, congratulations.
zapp [?\]: Ok, thanks for the game and the initiative!
]
;W[hd]WL[1236.618]
;B[bh]BL[1190.436]
;W[dc]WL[1190.506]
;B[db]BL[1176.27]
;W[cc]WL[1144.5]
;B[bc]BL[1167.046]
;W[cb]WL[1098.501]
;B[bb]BL[1159.304]
;W[ca]WL[1052.511]
;B[ic]BL[1150.01]
;W[bd]WL[1006.537]
;B[ad]BL[1148.655]
;W[hb]WL[960.63]
;B[fb]BL[1145.843])