Is Rémi Couloms' simulation balancing widely used?
It could also be used asymmetrically in handicap games.
Instead of making strong attacking moves less likely for both sides, black
would be deprived of strong moves, until the playouts come back at 50%.
If the position improves for white, black wou
Watching his constant flow of words get interrupted by a nasty bot trick
might be some fun. But this video against weak bots is insufferable.
Bet a friend 5 $ that he can't watch the whole thing. Easy money.
Stefan
___
Computer-go mailing list
Computer-g
On Thu, Jul 3, 2014 at 9:00 AM, Darren Cook wrote:
>
> If you had a choice between a 1% 65,000-wins move and a 70% 7-wins move,
> MCTS will keep exploring the 70% move, until it either reaches 65,001
> wins, and can be chosen, or the winning percentage comes down to 1% also.
>
> BTW, that implies
Thank you for the enlightening explanation. It actually explains more than
the explainers may wish to convey. My condolences to you for having that
innovative visual opener foisted on your fine article.
Stefan
___
Computer-go mailing list
Computer-go@dva
The "artist" certainly shows a lack of appreciation and respect for go.
Whoever created it, must think that go is already "in the bag".
Stefan
___
Computer-go mailing list
Computer-go@dvandva.org
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
Nice banana car. But the picture in the article is an abomination.
What got me hooked on go, a quarter century ago, was the first look at a
real go position.
I immediately felt a rush, that told me this game trumps all I had come to
know before.
So I'm really very unwilling to forgive that shitty b
"But the Bots are deluded by the Fata Morgana of huge moyos and will never
get to such a position."
If a bot is "deluded", it will go to work with that delusion on any
position. I think it's actually more useful to study positions that
are especially susceptible to this behaviour, and then work o
Thx for the link.
Zen is portayed as a blood thirsty warmonger, and Crazystone as a
bloodless accountant.
It's easy to see how they get to that judgement on the basis of these two games.
They just don't understand the Jeckyll and Hide nature of bots.
On Wed, May 21, 2014 at 12:23 PM, Rémi Coulom
8, 2014 at 7:57 AM, uurtamo . wrote:
> Why 7?
>
> On Apr 7, 2014 10:45 PM, "Stefan Kaitschick"
> wrote:
>>
>> Yes, thanks. Far better than what the UEC cup has to offer for results.
>>
>> In the game Zen lost, n10 was just as big as c6, while also re
Yes, thanks. Far better than what the UEC cup has to offer for results.
In the game Zen lost, n10 was just as big as c6, while also removing the ko.
But Zen was about 2.5 behind, so it didn't care.
With the hair-raising tenukis, bots are turning the 13*13 games into
bloody battles ,reminiscent of
Exact komi of 7 is interesting.
Will there be a day when bots will consistently play jigo on 9*9? :-)
___
Computer-go mailing list
Computer-go@dvandva.org
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
The powerful local response heuristic can be seen as a kind of global
anti-permutation heuristic.
But the problem also exists on a local level, most prominently when it
comes to filling liberties.
Moves that are in a critical area, are close to each other, and/or
border on the same enemy chain, and
“Multiple Overlapping Tiles for Contextual Monte Carlo Tree Search”
lol, thx
___
Computer-go mailing list
Computer-go@dvandva.org
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
Has anybody ever experimented with tiling the board with a set of
possibly overlaping smaller tiles,
and updating the RAVE statistic for those tiles, whenever the same
pattern comes up for that tile in the tree?
This would probably be more reliable than AMAF, and fill faster than
RAVE, so it might
No wonder you got no takers. It's too complicated for amateurs to
comment on it comfortably.
All I can say, is that I couldnt spot any obvious mistakes in the ko fight.
These bots are actually pretty great, when it comes to shortage of libs fights.
Stefan
On Tue, Dec 10, 2013 at 1:38 PM, Rémi Co
t;
> But having two of them does not really change anything as attacker can
> choose to fight then after each other.
>
> Petri
>
>
> 2013/12/10 Stefan Kaitschick
>>
>> What happens with 2 bent fours?
>> You should be able to save one of them.
>>
>
What happens with 2 bent fours?
You should be able to save one of them.
On Mon, Dec 9, 2013 at 10:01 PM, Hiroshi Yamashita wrote:
>> In round 3 CrazyStone vs. Aya, probably Aya was expecting W to play L12 or
>> N12, if B's corner was not 100% dead in the playouts. It seems a bug in
>> Aya's playo
When you do only a single playout, you dont even need a playout, just
generate the first move candidate - voilá. :-)
Closed sources are indeed regretable at this point.
Before, it sparked a kind of bot war, and the greatest technical
advances are always made at war time.
But now that the top bots
>> Yes. 300 playouts on 19x19 initial board is like this,
>>
>> move num winrate
>> 1:d16 64, 0.437
>> 2:q16 76, 0.447
>> 3:d445, 0.444
>> 4:q4 112, 0.508
>> 5:d171, 1.000
>> 6:r161, 1.000
>> 7:q3 1, 1.000
>
q4 gets a home corner bonus :-)
_
Ayabot is playing 1k on KGS with 1200 playouts.
On Wed, Nov 13, 2013 at 3:09 AM, Dusty Leary wrote:
>
> How many probes/playouts are necessary for a 19x19 MCTS bot to beat a human
> amateur? Let's say 5k rating.
>
>
>
> ___
> Computer-go mailing list
>
One more thought: it's especially important to detect genuinely local
semeais. Because they're the ones that are hurting the bots.
Ironically, bots are happy with strange and convoluted interactions
that make it extremely difficult for humans. The reason is, that
humans are so good at isolating pro
If it doesn't have a little monitor in front of it in the user list,
it's not a bot. ("bs" was also a giveaway)
If it has a monitor and it plays 7d, there has either been another
breakthrough, or a human is using the bot protocol.
Stefan
On Tue, Sep 17, 2013 at 9:02 PM, mark schreiber
wrote:
>
The main problem is that your program has gotten too strong.
Handicaping your bot by doing only very few playouts only helps so much.
Because the largest part of your progress lies in how your bot
converges to better moves when it is given a proper number of
playouts.
Stefan
__
That certainly sounds reasonable.
Though I still don't get why bots would frivolously fritter away the
final cushion.
Even a great bean counter should know that he can't expect his own
count to be perfect.
Then again perhaps todays bots are smarter about this, and the
perplexing behaviour I'm think
I have never understood this bot behaviour, because once a position
gets very close, every move becomes critical and that should depress
the winrate atleast somewhat. The only explanation that I have for
myself, is that while the lead is still comfortable, the bot will shun
optimal moves if they re
On Thu, Jun 13, 2013 at 12:08 AM, Don Dailey wrote:
> I don't quite see the point. The goal is to find the best possible hand
> YOU can make with your 13 cards and there is no betting, so I see no point
> in using Monte Carlo here.
>
> What am I missing?
>
> Is it whether to sacrifice one of t
On Tue, Jun 11, 2013 at 8:48 AM, David Fotland wrote:
> Sorry, secret. It took me quite a while to find something that worked well.
> I think you are right that most of the benefit comes from keeping the win
> rate close to 50%.
Keeping the winrate near 50% sounds reasonable. But scoring correct
On Sun, Jun 9, 2013 at 9:23 PM, David Fotland wrote:
> Dynamic komi and some other tricks work quite well. Thanks to Ingo for
> pushing dynamic komi until I figured out how to make it work well. Often
> the playout have some bias due to a misread in a fight, so it’s important
> for the bot to ke
@Don:
> To me the biggest issue isn't winning by 0.5 but not fighting when it's
> losing.
It can sometimes feel that way, but when a bot is timid in a losing
position that's because it can't see the loss.
> I have a feeling that most of this is not about improving the
> play of the bot although
Humans may be predisposed to the fallacy of greed.
But bots have there own fallacies. Happily winning by 0.5, when a
higher win at almost the same perceived risk is available, is a kind
of full knowledge fallacy. A bot can lose an easily won game by
merrily giving away everything but the last half
Congratulations.
The attachment at black 16 is extravagant, but otherwise it's very
difficult to identify this as a bot game.
It looks like a cautious strong amateur taking 3 stones.
Stefan
___
Computer-go mailing list
Computer-go@dvandva.org
http://dvan
points. Seki is the exception as it can change the score for a point from
> black (or white) to unoccupied, but ko certainly does not.
>
> On Mon, Jun 3, 2013 at 6:08 AM, Stefan Kaitschick
> wrote:
>>
>>
>> Hm, ok, 2 points is also important. :-)
>> Even under chinese
> Assuming Chinese rules:
> The only way to increase the margin by one point is to create or remove
> an odd seki from the board. Winning a ko instead of losing it will
> increase the winning margin by two points.
>
> Nick
>
Hm, ok, 2 points is also important. :-)
Even under chinese rules, a ko c
>
> My original question was meant for non-resigning bots. The observation:
> Also a losing bot starts to play lazy towards the end. And the question
> is:
>
> To which results do laziness of the winning bot AND laziness
> of the losing bot together lead typically?
>
> Ingo.
A losing bot is never
Have you given the strongest single player the combined resources of
the opposing team, or did each team member get to spend as many
resources as the single player?
___
Computer-go mailing list
Computer-go@dvandva.org
http://dvandva.org/cgi-bin/mailman/li
A while back Remi Coulom made a great proposal for measuring
"criticality", the importance of a local situation in contributing to
winning or losing.
Go programming has become less of a public affair, so maybe some
programs are utilizing it in some way.
But it's actually not easy to say what to do
Ofcourse a longer thinking time can yield a worse move.
That's like a weaker player playing a better move than a stronger
player would have. It only means that the position is too complicated
for both of them, and the weaker player got lucky.
___
Computer
Very nice game for Zen. Move 40 almost won the game just by itself.
White destroyed b territory so quickly, that Zen decided to show some
of its power against whites weak stones.
The silly p3 wasnt really a problem, imo that was just a bot-typical
"let's go home" move.
Time for Handi 3 games.
Stef
That's my impression too. It feels like Zen is slacking his way to the
4 stone wins.
3 stones will be a lot more interesting imo, as we will probably get
to see some of the famous zen attacks. It's probably not a good idea
to try to recruit well known players for the first 3 stone games.
Takemiya m
Thank you for the link.
There's nothing new under the sun. :-)
On Tue, Mar 6, 2012 at 10:11 PM, Rémi Coulom wrote:
> Accelerated UCT does this:
>
> https://www.conftool.net/acg13/index.php/Hashimoto-Accelerated_UCT_and_Its_Application_to_Two-Player_Games-111.pdf?page=downloadPaper&filename=Hashi
On Tue, Mar 6, 2012 at 3:16 PM, Álvaro Begué wrote:
> I think the solution is the introduction of some flavor of minimax
> into the tree search. For instance, once a node has been visited more
> than a certain number of times, the score that we'll back out from it
> is just the score of the best
>
> Lazarus had that problem too and before that my simple bot had it really
> badly because it did not have the benefit of a tree search. I don't
> know if discounting the move is a reasonable solution or not, perhaps it
> is. But I think that if you can solve it in the playouts you solve it
Bots love to throw in useless ko threat type moves now and then.
Sometimes its just a wasted threat, at other times it loses a useful
liberty too.
Why do even strong bots show this "stupid" behaviour?
Ko threat type moves have one thing in common: a big payoff if the opponent
answers incorrectly.
On Sun, Feb 26, 2012 at 8:42 AM, Hiroshi Yamashita wrote:
> Japanese 9d pro Yoshio ISHIDA plyaed one 13x13 game with Zen yesterday.
> Ishida won. Rule was no handicap stones, no komi, 25 minutes + byoyomi.
> This game was lived on Nico Nico Do-ga.
>
> Ishida said Black wasted ko threats before ko
Oops, that was the fourth try - the one with the 29.5 point loss.
On Sun, Jan 29, 2012 at 4:41 AM, Stefan Kaitschick <
stefan.kaitsch...@hamburg.de> wrote:
> Awesome attempt!
> 285 at R8 and you would have crushed it in the first try.
>
> Stefan
>
>
> On Sun, Jan 29,
Awesome attempt!
285 at R8 and you would have crushed it in the first try.
Stefan
On Sun, Jan 29, 2012 at 2:47 AM, Aja Huang wrote:
> You should let me try. :)
>
> Aja
>
>
> *From:* David Fotland
> *Sent:* Saturday, January 28, 2012 6:45 PM
> *To:* computer-go@dvandva.org
> *Subject:* Re: [
>
> I think, sophisticated "dynamic komi" will not be enough to achieve the
> level.
> You will need some sort of opponent modelling. But future will show.
>
Even if challengers get the source, modeling the opponent will be tough.
Because modeling in the playout policy has to be much more economic
>
> Of course, for experiments it would be neat to have a binary of this
> version, but maybe that would make the challenge too easy (or prone to
> overtuning).
>
My guess is, that opponent modeling will have to be part of the package.
For a generic weak opponent, maybe one sided simulation balanc
This game is very nicely commented by David Ormerod at gogameguru:
http://gogameguru.com/man-machine-match-final-results-game-commentary/
He did an amazing job of explaining the complicated tactical encounters.
It's a lot better than what I could have come up with.
Stefan
problem to lose against a bot ever since deep
blue humbled the world champion.
Stefan
On Thu, Jan 19, 2012 at 8:11 PM, Don Dailey wrote:
>
>
> On Thu, Jan 19, 2012 at 1:02 PM, Stefan Kaitschick
> wrote:
>>
>> I'm with John on this one.
>> Professional pride wi
I forgot to add, that I made an easily verifiable prediction.
If the pros don't make several attempts at 5 before grudgingly going to 4,
you can say "heh, Stefan, remember that prediction of yours?"
On Thu, Jan 19, 2012 at 8:17 PM, Stefan Kaitschick
wrote:
> Hi Ingo,
>
Hi Ingo,
ofcourse i wouldn't dream of accepting those conditions.
I would be accepting 30 to 1 odds for H3 and 70 to 1 odds for H2.
And the money I would be risking would make the unhappy outcomes a lot
more likely.
I would say this though: even if the bonus for winning was the same at
all handic
I'm with John on this one.
Professional pride will prevail.
300$ won't make a pros knees go weak.
Much worse would be the slim chance to lose a low handicap game.
Zen will have to beat a series of pros with 5 stones, before any of
them will consider going down to 4.
And 2 or 3 stones don't really n
>> For the 19x19 exhibition game the pro will be payed according to handicap.
>> He (or she) gets 50 Euro simply for playing. And he has the choice at which
>> handicap. When winning the game the pro gets additional money:
>> +50 Euro, when handicap was 2,
>> +100 Euro, when handicap was 3,
>> +200
On Tue, Jan 17, 2012 at 3:05 PM, Don Dailey wrote:
> A few years ago this would have been considered a very unlikely result by
> most of us!So Congratulations to Zen, and a big thanks to John Tromp
> for being such a gentleman and making this possible and winning his
> (previous) bet and bei
Very nice. Nothing to quibble about.
Example 2 is especially nice - the kind that's hard to find, but easy to
understand.
Stefan
On Tue, Jan 17, 2012 at 2:35 PM, Jouni Valkonen wrote:
>
> Here are some very good example about the usefulness of CrazyAnalysis:
>
> http://www.grappa.univ-lille3.fr/
Congrats to Zen this time.
Thanks again, Ingo.
"Later this group went over Jordan."
For some reason crossing the Jordan is not one of the hundreds of
euphemisms that do exist in the english language.
How about "it entered the garden of serenity" :-)
The MFs winrate graph is pretty good at telling
Congratulations to John Tromp.
He stayed cool in tough times and is now really in it to win it.
Zen got an excellent fuseki and still lost.
Thanks for those winrate graphs, Ingo.
61 and 67 were both bad moves, both answered correctly by John.
Instead of d15, Zen should have probed at c16 - a huge d
In the info for Zen19N it still says "Mac Pro 8 core".
Is the info outdated, or will the hardware be upgraded for the match?
Stefan
On Fri, Jan 13, 2012 at 2:19 AM, Darren Cook wrote:
> > Just returned from a visit to KGS where Zen seems to have struggled a
> > bit in recent games. Maybe they'r
do you know what hardware Zen will be using?
I would give Zen19 an 80% chance and Zen19D a 95% chance of winning.
Unless you get saved by a rare bug, to beat Zen you need to outplay Zen in
the fuseki. But for that it usually takes at least 4d strength.
some tips for the underdog:
contest zen for
I think your beeing too logical.
There is no logic to vanity.
(not his bot btw)
Stefan
On Mon, Jan 9, 2012 at 7:27 PM, Don Dailey wrote:
>
>
> On Mon, Jan 9, 2012 at 1:16 PM, Stefan Kaitschick <
> stefan.kaitsch...@hamburg.de> wrote:
>
>> That was not bothater.
&g
The correct answer is 0, because shifting by just 1 point drops the rate to
> 10%
> But is this really a dynamic komi problem?
> I mean, is there really a graceful way to misevaluate semeais?
> apropos semeais: would it be feasible to expand RAVE to include the
> information of the previous move?
>
gain more territory. To cure this problem, besides winning rate we might
> have to use the information of not only “average score” but also “average
> score of certain points”.
>
> Aja
>
>
> *From:* Stefan Kaitschick
> *Sent:* Monday, January 09, 2012 4:38 AM
> *To:* com
That was not bothater.
A strong 6d razed many kgs 5ds(me included) with his account.
This is a case of importing rating points to the bot.
I really don't see a cure for this.
Wrong handicaps could be corrected with a simple function added to the
protocol.(Or the bot could look up the kgs archive be
On Mon, Jan 9, 2012 at 2:01 PM, terry mcintyre wrote:
> Just kicking out an idea: when you offer a handicap to a player, the
> presumption is that he or she or it is significantly weaker than yourself.
>
> Can that weakness be modeled in the playouts? A 5d player tends to know
> subtle things abou
By the way, are we sure it is underestimation of the edges and corners ?
Rather than overestimation of the centre ?
>
> I know those are equivalent for play itself, but the answer suggests
> different tries for solution. In the first case, we want to make the bot
> more aware that he will keep its
scoring function:
The most successful scoring function sofar is the win/lose function.
Sigmoid functions and other schemes have been tried, but none have
surpassed or even equaled the simple step function.
dynamic komi:
dynamic komi is widely used by bots in handicap games.
An initial artificial
Am 18.07.2011 14:58, schrieb terry mcintyre:
Perhaps the admin was simply trying to apply the tournament rules
uniformly?
Terry McIntyre
Even as a german, I blanch in envy at such administrative perfection.
Stefan
___
Computer-go mailing list
Comput
Am 15.07.2011 14:03, schrieb Petr Baudis:
> For black I would wave my hands differently:
> Since the opponent must be stronger, it's a good heuristic to assume
> a problem with your own analysis if you think your improving.
I'm not sure that is satisfactory explanation. Even after 10s of
Am 14.07.2011 17:52, schrieb Don Dailey:
In theory you HAVE to cede ground as you have a lost game.Even
though I do not play much go and I'm not strong, I know that you have
to give up ground in some places to gain ground in others and that is
what separates the men from the boys.
So I ca
Am 14.07.2011 17:39, schrieb Petr Baudis:
On Thu, Jul 14, 2011 at 02:07:50PM +0200, Stefan Kaitschick wrote:
Just a hand waving explanation for the ratchet: you simply cannot
afford to cede ground to black when giving a handicap.
Even though the playouts do not model a weak response by black, a
Am 12.07.2011 11:54, schrieb Petr Baudis:
Hi!
On Tue, Jun 21, 2011 at 02:13:50PM +0200, Olivier Teytaud wrote:
I have posted too quickly - after all I have something which works both for
black and white and for various board sizes, using
the "rule 42" spirit. Good news :-)
Glad you got i
worse than other moves without komi, could be
sorted out.
Stefan
Von: Stefan Kaitschick
...
A good alternative might be to let each computer evaluate the position
at a different komi level.
Then you get a risk/reward profile for the current position.
As this is more about the risk/reward
Jonathan,
Hideki,
what strength player are you?
I'm IGS 4D
Just make your arguments, and others will realize your strength naturally.
in my experience:
strong players play good moves,
they dont play better moves in handicap games.
weak players play bad moves,
they dont play worse moves i
When the hardware is made up of several computers, the communications
overhead strongly reduces the utility of the combined computing power.
A good alternative might be to let each computer evaluate the position
at a different komi level.
Then you get a risk/reward profile for the current positi
I don't think that adjusting to a certain winrate at a certain point in
the game is really good.
What I would like to see, is an independend winrate vs. dyn. komi
profiling for the current position.
Only then can you really decide, which komi level currently delivers the
best bang for the buck.
ot;unsound" if you are losing anyway. I see
no problem with your idea but the devil is in the details.
On Fri, Jun 17, 2011 at 4:23 PM, Stefan Kaitschick
wrote:
Zen19S is an account on KGS with long time controls(20 + 30/5)*, running
on acluster of 6 pcs. It holds a solid 4dan rating.
I thin
Zen19S is an account on KGS with long time controls(20 + 30/5)*, running
on acluster of 6 pcs. It holds a solid 4dan rating.
I think it's handicap openings have really improved with both black and
white, and I think dyn. komi is a big part of this.
But I have seen some 6 stone games as white(the
With programs getting really strong, there is another factor to consider:
As you approach perfection, becoming a stone stronger becomes infinitely
difficult.
This is really a quirk of the go ranking system, which defines strength
as the ability to give handicap stones.
If strength were defined a
count and
read ahead if given unlimited time.
Jouni
On Jun 14, 2011 5:56 PM, "Stefan Kaitschick"
wrote:
Jouni,
you're really putting your money where your mouth is. Congrats for that.
I agree that especially zen, the only bot to hold a consistent 5d on kgs
sofar, profits alo
X-Prize#
If someone makes a gobot that wins me on best of 5 match (80 40
fischer), I will pay for the programmer €1000 X-prize. Prize will be
in effect for the next 2000 years or when the winner of the
competition has been found. Challenges can be played during European
Go Congress, but I accep
There is no "Greenwhich normal time".
There is only a "Greenwhich mean time".
The "mean" refers to the mean (average) solar time.
(solar time is what we would have without timezones - the "mean" creates
the timezone)
Please excuse by nitpicking, and thanks for the info
Stefan
Today, MyGoFrien
Killer moves works well in alpha-beta search I guess because it found
by searching the PV then when alternative moves are played to the best
they are cut of one at a time with the killer move. Thus the killer
move is applied to almost the same position in every attempt.
In a playout the boards
I suppose that is called "Adaptive Playout".
Hendrik Baier reported LGRF heuristics and other lots of failed methods.
www.ke.tu-darmstadt.de/lehre/arbeiten/master/2010/Baier_Hendrik.pdf
--
Yamato
Thanks for the link.
The author comes to a slightly different conclusion though:
"In summary,
In this game, there was a big semeai on the left side. The result
was a won
position for Aya, but both Aya and ManyFaces thought that ManyFaces
had won
(or perhaps that it was a semeai), so eventually Aya resigned before
it was
played out.
This position reminds me a similar situation in t
It's childishly simple for a human, large eye against small eye.
Programmers need to find a way to make this a simple position for bots too.
Solving it with great computational effort isn't good enough.
Stefan
Aya resigned in winnning position...
http://files.gokgs.com/games/2011/5/24/ManyFaces
Am 24.05.2011 08:51, schrieb "Ingo Althöfer":
Hurray. It seems that GMX is transfering messages from the list again:
this night I got a batch with 13 messages.
Von: Olivier Teytaud
We have tried to motivate Rémi, without success :-)
There is one possible way for the long range (used many tim
Abstract:
Monte-Carlo Tree Search is a very successful game playing algorithm.
Unfortunately it suffers from the horizon effect: some important
tactical sequences may be delayed beyond the depth of the search tree, causing
evaluation errors.
continued ...
I have a question: what exactly is the
Surely, the position between Zen and Pachi doesn't qualify :-)
As for a human-weaknesses database, my advice to any programing team
would be: just mind your own blind spots.
Stronger bots will need to reduce the current plague of refighting every
fight in every branch of the tree.
This gets eve
I'm also under the impression that Zen stopped throwing away corners.
In a bot tournament that usually means ending up with 3 corners, because
currently all bots still undervalue corner plays.
It's also true that Zen can now play a "normal" opening without
trampling on it's own positions later.
Am 26.04.2011 16:52, schrieb Brian Sheppard:
I know one guideline that is close to ironclad: take the ko last. (That is,
fill other liberties before beginning the ko.)
Even that's not ironclad. There are kos where one side should take the
ko early, but then the other side can trade libs as a ko
Hi Aja,
Thank you for the report.
It seems Zen passes Turing test.
In the tournament I haven't play with Zen, but the fact that Zen's
final score (5.333) is much higher than Erica (3.2) proves Zen played
like a human player. The human players were constituted with a certain
variety: high dan,
Am 02.03.2011 19:19, schrieb Colin Kern:
On Tue, Mar 1, 2011 at 2:29 PM, Michael Alford wrote:
On 2/10/11 4:31 PM, Michael Alford wrote:
A reminder for anyone interested:
The supercomputer Watson will appear on the Jeopardy program this coming
Feb 14-16.
Michael
Watson won on the Jeopardy
Am 05.02.2011 12:50, schrieb Alain Baeckeroot:
Le 04/02/2011 14:48, Stefan Kaitschick a écrit :
Am 04.02.2011 14:36, schrieb Robert Lupton the Good:
Actually that's the Entropy, S; Enthalpy is E - PV (energy corrected
for pressure work).
If you are interested in thermodynamic analogues
Am 04.02.2011 14:36, schrieb Robert Lupton the Good:
Actually that's the Entropy, S; Enthalpy is E - PV (energy corrected for
pressure work).
If you are interested in thermodynamic analogues, the Helmholtz free energy (E
- TS) might be an interesting concept if you can assign meanings to temp
Am 28.01.2011 12:15, schrieb Nick Wedd:
That sounds crazy, because it has such a gambling feel to it, but its
true.
Its because even a mediocre player can muddle his way through the
roll of the dice, but only a strong player can judge well what his
current chances are.
The basic idea is that yo
The problem is that you can still play the game out until there is just one
or two moves left and then resign.So for this work it has to be done at
some reasonable point in the game and who is to decide when that should be?
No, you either concede immediately, or you continue to play for twice
My experiments were are 9x9 too.
I believe what was happening with my implementations of this is that
it worked well most of the time, but not when it really mattered.
When it didn't work, it was turning a simple win into a struggle and
sometimes a loss.
Don
Why would a simple win b
on 20.01.2011 05:35, Darren Cook wrote:
Told you I was getting passionate about this point;-). Though actually
I wasn't even thinking so much about number of playouts, more about
other ideas such as first spending a few seconds on static semeai or
life-death analysis on the root position then usi
Side question: How difficult would it be to design
a program that generates such "bot-difficult semeais"
(or at least building blocks for such) automatically?
Only when bots can solve essential positions, do you need random
problems to make sure that the answers are robust.
Right now, there
1 - 100 of 143 matches
Mail list logo