I uploaded a new source to sf.net, I added the FLAT blockset
management (Bot player). After a quick tunning, i got:
3x3x6 FLAT: Avg=26128.2 Min=52(35588687) Max=220970 nbGame=1000
Let me know if you get problems with this new source.
BL2 Bot Player
Re:BL2 Bot Player
Last OOC result:
Tests are long, this will not ease params tunning :(
<br><br>Post edited by: jlp_38, at: 2008/02/24 00:22
Tests are long, this will not ease params tunning :(
Code: Select all
[ 49] 107128 Avg=52534.24 [44726.00,69509.50] (Min 1964 Max 289385)
[ 50] 46923 Avg=52424.22 [44726.00,69509.50] (Min 1964 Max 289385)
[ 51] 59813 Avg=52566.31 [44726.00,69509.50] (Min 1964 Max 289385)
[ 52] 32186 Avg=52181.77 [44726.00,69509.50] (Min 1964 Max 289385)
[ 53] 22712 Avg=51636.04 [44726.00,69509.50] (Min 1964 Max 289385)
[ 54] 172298 Avg=53829.89 [44726.00,69509.50] (Min 1964 Max 289385)
[ 55] 78924 Avg=54278.00 [44726.00,69509.50] (Min 1964 Max 289385)
[ 56] 36319 Avg=53962.93 [44726.00,69509.50] (Min 1964 Max 289385)
[ 57] 54960 Avg=53980.12 [44726.00,69509.50] (Min 1964 Max 289385)
[ 58] 22288 Avg=53442.97 [44726.00,69509.50] (Min 1964 Max 289385)
[ 59] 87707 Avg=54014.03 [44726.00,69509.50] (Min 1964 Max 289385)
[ 60] 32965 Avg=53668.97 [44726.00,69509.50] (Min 1964 Max 289385)
[ 61] 39264 Avg=53436.63 [44726.00,69509.50] (Min 1964 Max 289385)
Re:BL2 Bot Player
Last OOC results, I may reached ~100000 :)
<br><br>Post edited by: jlp_38, at: 2008/02/24 12:48
Code: Select all
[ 35] 104502 Avg=90771.81 [61448.33,184447.00] (Min 1369 Max 365390)
[ 36] 220460 Avg=94276.89 [61448.33,184447.00] (Min 1369 Max 365390)
[ 37] 186413 Avg=96701.53 [61448.33,184447.00] (Min 1369 Max 365390)
[ 38] 215681 Avg=99752.28 [61448.33,184447.00] (Min 1369 Max 365390)
[ 39] 195282 Avg=102140.52 [61448.33,184447.00] (Min 1369 Max 365390)
[ 40] 292924 Avg=106793.78 [61448.33,184447.00] (Min 1369 Max 365390)
[ 41] 100492 Avg=106643.74 [61448.33,184447.00] (Min 1369 Max 365390)
[ 42] 106448 Avg=106639.19 [61448.33,184447.00] (Min 1369 Max 365390)
[ 43] 219653 Avg=109207.68 [61448.33,184447.00] (Min 1369 Max 365390)
[ 44] 121487 Avg=109480.56 [61448.33,184447.00] (Min 1369 Max 365390)
[ 45] 135956 Avg=110056.11 [61448.33,184447.00] (Min 1369 Max 365390)
[ 46] 52790 Avg=108837.68 [61448.33,184447.00] (Min 1369 Max 365390)
[ 47] 92498 Avg=108497.27 [61448.33,184447.00] (Min 1369 Max 365390)
[ 48] 97923 Avg=108281.47 [61448.33,184447.00] (Min 1369 Max 365390)
[ 49] 91075 Avg=107937.34 [61448.33,184447.00] (Min 1369 Max 365390)
[ 50] 197081 Avg=109685.25 [61448.33,184447.00] (Min 1369 Max 365390)
Re:BL2 Bot Player
Last OOC results, 1 game above 1000000 cubes :)
I think the evaluation is almost ended. I adapted all
coefficients so they should be ~independent from
pit dimensions. I made several tests, it seems to work
rather well.
The parameters for FLAT and EXTENDED may still need
to be a bit tunned and a min-max search (a bit
modified though) can be added.
Here are last results:
3x3x6 FLAT: ~29000
3x3x6 BASIC: ~1800
3x3x10 EXTENDED: ~200
5x5x10 EXTENDED: >200000
I upload the final source on sf.net ASAP.
Code: Select all
[ 75] 61276 Avg=207755.93 (Min 2285 Max 850170)
[ 76] 188910 Avg=207511.18 (Min 2285 Max 850170)
[ 77] 36881 Avg=205323.62 (Min 2285 Max 850170)
[ 78] 624411 Avg=210628.52 (Min 2285 Max 850170)
[ 79] 36079 Avg=208446.65 (Min 2285 Max 850170)
[ 80] 1000003 Avg=218218.95 (Min 2285 Max 1000003) Max
[ 81] 28425 Avg=215904.39 (Min 2285 Max 1000003)
[ 82] 210146 Avg=215835.01 (Min 2285 Max 1000003)
[ 83] 9537 Avg=213379.08 (Min 2285 Max 1000003)
[ 84] 54242 Avg=211506.88 (Min 2285 Max 1000003)
[ 85] 138386 Avg=210656.64 (Min 2285 Max 1000003)
[ 86] 192873 Avg=210452.23 (Min 2285 Max 1000003)
[ 87] 315670 Avg=211647.89 (Min 2285 Max 1000003)
coefficients so they should be ~independent from
pit dimensions. I made several tests, it seems to work
rather well.
The parameters for FLAT and EXTENDED may still need
to be a bit tunned and a min-max search (a bit
modified though) can be added.
Here are last results:
3x3x6 FLAT: ~29000
3x3x6 BASIC: ~1800
3x3x10 EXTENDED: ~200
5x5x10 EXTENDED: >200000
I upload the final source on sf.net ASAP.
Re:BL2 Bot Player
OK,I\'ve just uploaded the final release of the bot player
on sf.net. Let me know if you find a bug or something that
can be improved :)
To Lieven:
I would like to add your name and email address to the
README file. Do you have a public email address that i
can use ?
I let your original code in the source. It is not used
but It brings me a truly help :) Thanks again for
your indeed good job. If you want, I can remove your
original source... Just let me know
on sf.net. Let me know if you find a bug or something that
can be improved :)
To Lieven:
I would like to add your name and email address to the
README file. Do you have a public email address that i
can use ?
I let your original code in the source. It is not used
but It brings me a truly help :) Thanks again for
your indeed good job. If you want, I can remove your
original source... Just let me know
Re:BL2 Bot Player
recently on www.tetrisconcept.com there were some interesting links posted, that eventually might give you some more ideas on how to improve your bot player:
\"For those interested, this article defines many field metrics such as number of holes, surface smoothness, and blocks above holes.\"
http://www.cs.cornell.edu/boom/1999sp/p ... ction.html
http://colinfahey.com/tetris/tetris.html
the corresponding thread in tc:
http://www.tetrisconcept.com/forum/viewtopic.php?t=927
\"For those interested, this article defines many field metrics such as number of holes, surface smoothness, and blocks above holes.\"
http://www.cs.cornell.edu/boom/1999sp/p ... ction.html
http://colinfahey.com/tetris/tetris.html
the corresponding thread in tc:
http://www.tetrisconcept.com/forum/viewtopic.php?t=927
Re:BL2 Bot Player
Thank you very much Herc for your links :)
They looks quite interesting, there is even
statistics study (They use binomial distribution).
That\'s too bad that i didn\'t have a look at this
pages before but OK It\'s never too late.
I think that the current evaluation function
already take in consideration most of parameters
i saw on these pages, I don\'t yet read all and
there are definitely many other interesting things.
I let you informed.
They looks quite interesting, there is even
statistics study (They use binomial distribution).
That\'s too bad that i didn\'t have a look at this
pages before but OK It\'s never too late.
I think that the current evaluation function
already take in consideration most of parameters
i saw on these pages, I don\'t yet read all and
there are definitely many other interesting things.
I let you informed.
Re:BL2 Bot Player
Congrats :)
3x3x6 FLAT: ~29000
3x3x6 BASIC: ~1800
3x3x10 EXTENDED: ~200
5x5x10 EXTENDED: >200000
I am genuinely impressed!
>I would like to add your name and email address to the
>README file. Do you have a public email address that i
>can use ?
>I let your original code in the source. It is not used
>but It brings me a truly help Thanks again for
>your indeed good job. If you want, I can remove your
>original source... Just let me know
Thanks a lot, but I rather not have my name or email address in the public! As for the bit of the code I wrote, I think that you can safely remove it now that you found these really good functions.
I am especially amazed how general your functions are, ie they work for all modes.
Lieven
3x3x6 FLAT: ~29000
3x3x6 BASIC: ~1800
3x3x10 EXTENDED: ~200
5x5x10 EXTENDED: >200000
I am genuinely impressed!
>I would like to add your name and email address to the
>README file. Do you have a public email address that i
>can use ?
>I let your original code in the source. It is not used
>but It brings me a truly help Thanks again for
>your indeed good job. If you want, I can remove your
>original source... Just let me know
Thanks a lot, but I rather not have my name or email address in the public! As for the bit of the code I wrote, I think that you can safely remove it now that you found these really good functions.
I am especially amazed how general your functions are, ie they work for all modes.
Lieven
Re:BL2 Bot Player
Thanks :)
I\'ll remove the old code. Thanks again for your genial ideas, especially edges and smoothness square (a kind of \"roughness
variance\"), which are very well adapted to the blockout game strategy.
I\'ll remove the old code. Thanks again for your genial ideas, especially edges and smoothness square (a kind of \"roughness
variance\"), which are very well adapted to the blockout game strategy.