Dariush 6.0, patterns, and pro moves prediction
ricoh51
We are happy to announce that Dariush 6.0 beta is out :
http://ricoh51.free.fr/indexeng.htm
News in this version :
Seek rectangular patterns in a games database (as Kombilo but faster)
Prediction of professionals moves from a base of 60 million patterns
with 55% of success (as MoyoGo but better)
Management of the multi-sgf files
Reading and writing the orobaduk files (.ngf)
Enjoy
eric
Frank de Groot
> Prediction of professionals moves from a base of 60 million patterns with
> 55% of success (as MoyoGo but better)
This is very likely not true.
Please test on never-seen games and post your results :-)
It's trivial to get 55% score on games of which the patterns are in your
database already - Moyo Go reaches 66% or more - but the real challenge is
to achieve a good pro-prediction rate on new games that have never been
"seen" by your pattern recognition system.
If what I suspect is correct and the 55% is reached only with "seen" games,
then 55% is piss-poor.
If what I suspect is false and you have reached 55% with never-seen games,
then I expect DARPA to contact you soon :-)
My sincere congratulations for the incredible achievement in Go-AI in the
latter case!
In the former case - back to the drawing board - I spent several years,
full-time in buiding an extremely advanced pattern system.
I reached 42% on 100 never-seen games but it was after an enormous expense
of time, research, trial and error and 24/7 number crunching on 10,000 USD+
machines with utterly optimized code (assembly).
I seriously doubt that you can get above 45%-50%, even with a theoretically
optimal (and insanely complex) pattern system.
Care to post some details on your measurement methods, and the pattern
system used?
(I have a PDF that explains the details, on my site)
ricoh51
> "ricoh51" <ric...@free.fr> wrote
>
>> Prediction of professionals moves from a base of 60 million patterns with
>> 55% of success (as MoyoGo but better)
>
>
> This is very likely not true.
>
> Please test on never-seen games and post your results :-)
>
> It's trivial to get 55% score on games of which the patterns are in your
> database already - Moyo Go reaches 66% or more - but the real challenge is to
> achieve a good pro-prediction rate on new games that have never been "seen"
> by your pattern recognition system.
lol, the test was made with all moves of 700 never-seen games.
Dariush displays 10 moves (with an number that is an urgency as
MoyoGo), the results are :
32,41% for the first one
46,65% for the 3 first
53,77% for the 6 first
and 55,4% for the 10
55,4% means that the pro move is in the 10 moves displayed
You can download Dariush and test it, it is free
> If what I suspect is correct and the 55% is reached only with "seen" games,
> then 55% is piss-poor.
and you? what means your 42%? Is it 42% for your move labeled "a"? or
for all your moves?
>
> If what I suspect is false and you have reached 55% with never-seen games,
> then I expect DARPA to contact you soon :-)
thank you :)
eric
Frank de Groot
> the results are :
> 32,41% for the first one
> 46,65% for the 3 first
> 53,77% for the 6 first
> and 55,4% for the 10
>
> 55,4% means that the pro move is in the 10 moves displayed
Good :-)
Moyo Go has around 85% for the first 10 moves :-)
You were a bit quick with your "better than Moyo Go" :-)
How can you claim that 55,4% is "correctly predicted" when in fact it's only
a mere 32,41%?
When I say correctly preditced percentage for MGS of 42%, of course I mean
the most highly ranked move only.
An apology/rectification seems in order..
> and you? what means your 42%? Is it 42% for your move labeled "a"? or for
> all your moves?
Move labelled "a".
>> If what I suspect is false and you have reached 55% with never-seen
>> games, then I expect DARPA to contact you soon :-)
>
> thank you :)
It's cheap manipulation/lying to name 55% as your "correctly predicted
percentage", whilst you talk about moves in the TOP TEN.
I'll give you the benefit of the doubt (who knows how old you are, etc.) but
I get a very bad taste in my mouth when I read such unfounded attacks, here.
Frank de Groot
>> 55,4% means that the pro move is in the 10 moves displayed
>
> Good :-)
>
> Moyo Go has around 85% for the first 10 moves :-)
In fact I now realize I'm selling MGS too cheap, way too cheap.
People do not understand the enormous significance and utility of the
software.
I will price it 120 USD, today.
boissac frederic
any one can test and compare.
I try with a game of pro.
Dariush find 70 moves and moyogo 44 (70/246=28,4% 44/246=17,8%).
you can download dariush and try.
:)
(sorry for my poor english)
"Frank de Groot" <fr...@moyogo.com> a écrit dans le message de news:
pPmdndOWpYG8OQbb...@giganews.com...
ricoh51
>> Dariush displays 10 moves (with an number that is an urgency as MoyoGo),
>> the results are :
>> 32,41% for the first one
>> 46,65% for the 3 first
>> 53,77% for the 6 first
>> and 55,4% for the 10
>>
>> 55,4% means that the pro move is in the 10 moves displayed
>
> Good :-)
>
> Moyo Go has around 85% for the first 10 moves :-)
>
> You were a bit quick with your "better than Moyo Go" :-)
if MoyoGo predicts 85% for the first 10 moves (and for all moves of the
game), then you are right, Dariush is not better, for the moment ;)
but for the game (246 moves) at the end of this post, the results are :
Dariush : 70 first moves correctly predicted : 28.4%
and 96 for the 3 first moves : 39%
MoyoGo 7.1.13.0 : 45 first moves correctly predicted : 18.3%
and 87 for the 3 first moves : 35.4%
I am conscious that only one test is not worth true statistics, but i'm
afraid to test many games by hand...
eric
----------------------------------------
(
;GM[1]FF[4]AP[DARIUSH:5.2]SZ[19]KM[6.5]RU[Japanese]HA[0]PB[³¯¸Ö²W]DT[2007-06-28]PC[LGS]BR[5p]WR[9p]RE[Adjourned]PW[µ²«°Áo]GN[µ²«°Áo-³¯¸Ö²W]
;B[pd]
;W[dc]
;B[dq]
;W[qp]
;B[ce]
;W[ed]
;B[ci]
;W[oq]
;B[ic]
;W[qf]
;B[nd]
;W[rd]
;B[qc]
;W[qi]
;B[do]
;W[iq]
;B[bc]
;W[gc]
;B[ie]
;W[bb]
;B[gq]
;W[io]
;B[no]
;W[mp]
;B[qo]
;W[po]
;B[pp]
;W[op]
;B[pn]
;W[pq]
;B[qj]
;W[pj]
;B[qk]
;W[pi]
;B[ro]
;W[pm]
;B[qm]
;W[pl]
;B[qn]
;W[ri]
;B[rq]
;W[ld]
;B[qe]
;W[re]
;B[pf]
;W[qg]
;B[ke]
;W[me]
;B[md]
;W[le]
;B[lc]
;W[bo]
;B[cp]
;W[ck]
;B[cl]
;W[dl]
;B[cm]
;W[dk]
;B[bk]
;W[bj]
;B[bl]
;W[ch]
;B[dh]
;W[di]
;B[dg]
;W[bi]
;B[cg]
;W[cj]
;B[fh]
;W[bd]
;B[hc]
;W[gb]
;B[mg]
;W[qr]
;B[rp]
;W[hr]
;B[rc]
;W[kc]
;B[kd]
;W[ne]
;B[mc]
;W[oe]
;B[kg]
;W[pe]
;B[qd]
;W[ng]
;B[nh]
;W[oh]
;B[gr]
;W[ja]
;B[kb]
;W[mo]
;B[jm]
;W[mh]
;B[lh]
;W[ni]
;B[cd]
;W[cc]
;B[gj]
;W[bg]
;B[bf]
;W[ej]
;B[ag]
;W[bh]
;B[kn]
;W[jp]
;B[jk]
;W[fg]
;B[gh]
;W[ff]
;B[eh]
;W[kj]
;B[kp]
;W[kq]
;B[li]
;W[jj]
;B[ik]
;W[lk]
;B[gd]
;W[fd]
;B[ge]
;W[df]
;B[ef]
;W[eg]
;B[rf]
;W[rg]
;B[ko]
;W[ee]
;B[cf]
;W[mm]
;B[nn]
;W[mn]
;B[rr]
;W[ij]
;B[hj]
;W[dn]
;B[dm]
;W[en]
;B[em]
;W[gn]
;B[eo]
;W[gl]
;B[fm]
;W[gm]
;B[hk]
;W[fn]
;B[cn]
;W[fp]
;B[hq]
;W[ir]
;B[ii]
;W[eq]
;B[er]
;W[dr]
;B[cr]
;W[ql]
;B[rl]
;W[pk]
;B[rk]
;W[fr]
;B[ds]
;W[fq]
;B[fs]
;W[hp]
;B[qq]
;W[pr]
;B[sd]
;W[sf]
;B[ji]
;W[gp]
;B[gs]
;W[kl]
;B[in]
;W[mj]
;B[hn]
;W[nm]
;B[pp]
;W[mf]
;B[lg]
;W[qp]
;B[on]
;W[od]
;B[oc]
;W[se]
;B[sc]
;W[km]
;B[mi]
;W[nh]
;B[fk]
;W[hs]
;B[es]
;W[gg]
;B[hg]
;W[rs]
;B[pp]
;W[oo]
;B[om]
;W[ol]
;B[lq]
;W[lp]
;B[kr]
;W[jq]
;B[mq]
;W[nr]
;B[ho]
;W[lr]
;B[ei]
;W[ek]
;B[fj]
;W[ki]
;B[kh]
;W[lf]
;B[kf]
;W[hb]
;B[ib]
;W[ia]
;B[ka]
;W[ha]
;B[ln]
;W[lm]
;B[sj]
;W[rj]
;B[rm]
;W[si]
;B[sk]
;W[gf]
;B[hf]
;W[jb]
;B[jc]
;W[ak]
;B[al]
;W[aj]
;B[ah]
;W[dd]
;B[ad]
;W[ac]
;B[ae]
;W[sr]
;B[hm]
;W[il]
;B[jl]
;W[hl]
;B[qp]
;W[qs]
;B[sq]
;W[ss]
)
Frank de Groot
> I am conscious that only one test is not worth true statistics, but i'm
> afraid to test many games by hand...
But not afraid to compare apples with pears and publicly draw the wrong
conclusions out of them..
I have little more time to waste on the "War Of The Go Software Vendors" -
the international Go playing community has proven to be grossly unworthe of
my efforts - but all I can say that I test on 100 games that are played by
9p's and that the games are not allowed to have great differences in score
(meaning, they should end with a .5 result)
I got 42% for the first ranking move.
Good luck with your software, for the rest.
Remember to only use pro games that your software has never "seen" ;-)
-
"Frank de Groot" <fr...@moyogo.com> wrote:
> I have little more time to waste on the "War Of The Go Software Vendors" -
> the international Go playing community has proven to be grossly unworthe
> of my efforts - but all I can say that I test on 100 games that are played by
> 9p's and that the games are not allowed to have great differences in score
> (meaning, they should end with a .5 result)
Unworthy, as in not buying your product or downloading the freebie?
Unworthy, as in not holding you to your word that MoyoGo or its cousin
would be "playing Go" as you have promised? Unworthy as in not offering
you the opportunity to discuss your product in various flame-war threads?
Or unworthy in not offering you some suggestions for improvement, such
as a statistic which measures the likelihood of completing a pattern?
For example, when the next move is played very far away then the
previous move had completed a pattern, (but maybe it was a _gote_).
> I got 42% for the first ranking move.
For the entire game, or merely for the first 23 or first 100 moves ?
- regards
- jb
-------------------------------------------------------------------
...some of the techniques that can be used to solve Sudoku puzzles
http://www.sadmansoftware.com/sudoku/techniques.htm
-------------------------------------------------------------------
Frank de Groot
> moves ?
For the entire game.
(The pro-prediction for the first dozens of moves is much higher than 42%)
About "unworthy" - I mean that people who do not defend "their" few talented
& productive software developers against slander, harrasment and crime do
not deserve the efforts of those software developers, and those software
developers will cease to develop Go software and commence to pursue less
stressful and more profitable endeavors.
ricoh51
> "ricoh51" <ric...@free.fr> wrote
>
>> I am conscious that only one test is not worth true statistics, but i'm
>> afraid to test many games by hand...
>
> But not afraid to compare apples with pears and publicly draw the wrong
> conclusions out of them..
Apples and pears??
>
> I have little more time to waste on the "War Of The Go Software Vendors" -
> the international Go playing community has proven to be grossly unworthe of
> my efforts - but all I can say that I test on 100 games that are played by
> 9p's and that the games are not allowed to have great differences in score
> (meaning, they should end with a .5 result)
>
> I got 42% for the first ranking move.
I doubt, I downloaded today 4 pro games from orobaduk, that are not in
our database, the results are below :
game 1 : Cho - Won : MoyoGo predicts 51/266 and Dariush 87/266
game 2 : Osawa - Chinen : MoyoGo predicts 75/235 and Dariush 92/235
game 3 : Xie - Qiu : MoyoGo predicts 39/143 and Dariush 43/143
game 4 : Cho - Ohn : MoyoGo predicts 60/284 and Dariush 81/284
with the previous game posted here, this gives a total of 1174 pro
moves, MoyoGo predicts 270 and Dariush 390
Conclusion for the first move:
MoyoGo : 23% and not 42%
Dariush 33%
Dariush is better than MoyoGo!! :)
All games are here :
http://ricoh51.free.fr/go/download/Testgames Dariush MoyoGo.rar
Anyone can test :)
>
> Good luck with your software, for the rest.
thank's, you too
> Remember to only use pro games that your software has never "seen" ;-)
yes, you too, 1174 pro moves are realistic
eric
Frank de Groot
> Apples and pears??
Yes.
My 42% is based on predicting the best move, yours on being in the top-10.
> Dariush 33%
Oh. So it's 33% and not 55%, as you previously claimed?
ricoh51
>
>> Apples and pears??
>
> Yes.
>
> My 42% is based on predicting the best move, yours on being in the top-10.
lol, no, 33% for the best move, and moyogo : 23%
I'm sorry for you (i do not think that you cheat) but you are mistaken,
i proved it
>
>> Dariush 33%
>
> Oh. So it's 33% and not 55%, as you previously claimed?
55% for the top 10, 33% for the best move
try Dariush and test any recent games that you want (that are not in
your database)
eric
Frank de Groot
> database)
Could be interesting to do, yes, but the problem is that I do not have the
auto-measuring code any more for Moyo Go, neither does Dariush have such an
automatic system, so it would take many days to do this.
-
> few talented & productive software developers against slander,
> harrasment and crime do not deserve the efforts of those software
> developers, and those software developers will cease to develop
> Go software and commence to pursue less stressful and more
> profitable endeavors.
Yes, development of Go software is a "labor of love" (though
we would not wish to have it any other way). If the crucible were
less hot then the kitchen might be too cold, and quality would suffer.
Not unlike the discussion where players wish to learn Go from other
kyus like themselves rather than seeking out high dans and/or pros.
- regards
- jb
-----------------------------------------------------------
Levitation On YouTube
http://www.rense.com/general77/levb.htm
-----------------------------------------------------------
-
Renli <usagi....@gmail.com> wrote:
> Fuck you frank. So when ya givin' up? :)
Frank's asshole is looking more attractive than the cunt of Renli's hag.
- regards
- jb
--------------------------------------------------------------
Another Nut In A Lawnchair With Balloons Soars High
http://www.breitbart.tv/?p=2860
--------------------------------------------------------------
Frank de Groot
> Fuck you frank. So when ya givin' up? :)
Have you seen any updates lately?
I'm running other businesess now.
I have a shop in Oslo where I fix PC's, laptops etc.
Good business. I make in a month what I used to make with Moyo Go in a year.
I am still selling MGS of course, due to the demand.
I bought a Saab 900 Turbo, three days ago, from Q1 & Q2 profits.
Q3 & Q4 will pay for trimming :-)
-
Renli <usagi....@gmail.com> wrote:
Mef
> "Renli" <usagi.mei...@gmail.com> wrote
>
> > Fuck you frank. So when ya givin' up? :)
>
> Have you seen any updates lately?
> I'm running other businesess now.
> I have a shop in Oslo where I fix PC's, laptops etc.
> Good business. I make in a month what I used to make with Moyo Go in a year.
>
This is good to hear, though not surprising. I imagine there's a much
large market for fixing computers than there is for go software.
Unfortunately, such is [the go playing] life.
Cheers,
Mef
Frank de Groot
> This is good to hear, though not surprising. I imagine there's a much
> large market for fixing computers than there is for go software.
Oh no, absolutely not.
The yearly market for Go software is 5 million USD/year - mainly in SE Asia.
I will soon have employees, here in Oslo, but 5 mio/y isn't the cat's piss
either.
> Unfortunately, such is [the go playing] life.
Oh no, absolutely not.
For half a decade, I was firmly comitted to making the best Go software,
until a handful people's criminal attacks became too much.
The main part being the fact that nobody in the Go community came to my
defence.
No harm done - you guys will be without innovative software for another
generation - my bad for taking you - as a community - seriously.
-
> software, until a handful people's criminal attacks became too much.
> The main part being the fact that nobody in the Go community came
> to my defence.
I'm not aware that anybody took Frank to court, even though Frank
had invited them to do so. Frank's ads were delisted from a website,
surely the perogative of any website owner to do so once receiving
veiled threats from Frank. Frank's legal expertise also exceeds that of
his lawyer (means that Frank does not require a legal defense fund).
Frank blames only a handful, rather small considering the total size.
Frank's software product received good reviews and mixed reviews.
Frank softpedaled his own stuff by steering very wide from a prospect
of "playing Go." If Frank has really suffered "criminal attacks" then EU
laws ought to allow Frank for complete recovery of damages.
As for "Go community" this is surely a common delusion. Nobody
has contributed funds into a community treasury to be posting to r.g.g.
There's no Mayor, no (official) moderator, no board of directors, no
elections, no governance, no regular meetings by community groups.
Some posters appear to be on mental autopilot, resembling of bots, or
accuse others of resembling bots. Frank has repeatedly informed us
that he is "not a player" of Go, nor will his program soon play Go. In
such circumstances not much there for any community to connect with.
SE Asia is a risky market because Singapore obeys no copyright law.
- regards
- jb
---------------------------------------------------------------------
A Review of the Universe - Structures, Evolutions, Observations, and Theories
http://universe-review.ca/index.htm
---------------------------------------------------------------------
Frank de Groot
> That's because you're a big fat jerk who pissed off *too many* people
> in the go community.
I could care less how your ego's work.
Like I said, I moved on to more profitable work, for people who appreciate
what I do for them.
Instead of working for people with inferiority/superiority complexes, I now
work for a nice average of society.
A breath of fresh air!
> attitude or something - I don't know. But when you start threatening
> to kill people
I never threatened to kill any specific person.
I said that I would defend myself against intruders into my bedroom, I said
that I might kill someone copying my software etc. but I never threatened
any person. If I have, sue me. I live in one of the most easy places to sue
someone successfully for that. Let the persons "threatened" sue!
> and in general act like a know it all asshole
Sorry but I AM a know-it-all asshole.
> A few days ago during a conversation on a certain go server your
> software came up and I swear to god 3 or 4 people were saying how you
> were mentally unstable or that you threatened to kill them.
That only shows what a bunch of lying scum people are who are envious of my
success.
Hardly any news.
Oh - BTW I sold a copy, an hour or so after you posted this.
Looks statistically significant.
> Seriously dude - do you even know what the effective difference is
> between kombilo and your program?
>
> Kombilo is free
And my software does about a hundred more things than Kombilo does.
Those programs are in different leagues. Kombilo is hacked together in a
fortnight, Moyo Go Studio took half a decade and was reverse-engineered by
Microsoft.
> Whats the benefit of having 400,000 games in your opening book vs.
> 40,000?
No idea. If you think Moyo Go is an "opening book", then I got news for you:
It's a pro-level Joseki tutor in the context of your own games, amongst
other things.
And those 400,000 extra games contain several tens of thousands games by
anonymous Korean pro's.
And my research shows that (their) modern Joseki - state of the art, cutting
edge - is only to be learned from those games, not from the GoGoD games.
Like I said, I will raise Moyo Go's price from 30 to 120 USD, to more
accurately reflect its true value.
120 USD is still very cheap, but at least it will not confuse people so
easily any more.
In fact - I think more realistic would be to be substantially above BiGo -
Let's say 395 USD.
> You don't understand go players or what they want.. although it's true
> you're a great graphic artist, your program does look nice, but I
> still deleted my copy.
You never owned a copy.
If you did, you have just admitted to software piracy.
I am not a "great graphic artist".
I *suck* at graphics.
My strong points are programming what Go players want and need, not
graphics.
> I would rather have a few hundred commented games from my favourite
> pro than 450,000 uncommented games.. big clue frank.. go players are
> not go software programmers. We don't need random data.
Since when are pro games "random data", Mr. Shit Talking Idiot?
http://www.soundclick.com/bands/songInfo.cfm?bandID=24002&songID=243091
Frank de Groot
> That isn't news frank. Thats a paraphrase of what I said.
Since when is Joseki Fuseki?
You lied when alleging that 400,000 mnore games merely yield 2 deeper moves
in Fuseki.
You deliberately hid the fact that it adds the latest Joseki and greatly
expands on continuations on Joseki.
> games by anonymous Korean pros? Oh please frank. That's the most
> ridiculous thing I've ever heard.
Shows more about you than about me.
It's your problem that you don't know who played on CyberKiwon.
> You don't know what go players need frank :) Your program is
> functionally no different from any of it's competitors.
Learn spelling, dude. It's = "it is".
I have a hard time believing that such a Go genius as yourself (with a "shit
talking idiot" song written about him!..) doesn't even know elementary
spelling of his native language.
> If it's a
> little faster or if it has a few more games great - those are really
> great selling points frank, and hey that just means your product is a
> little better than it's competitors from a functional standpoint. But
> it is in no way revolutionary frank. And people are not going to
> respect you just because you did something 5% or 10% better. No, if
> you want to be a know it all asshole you have to be at least 60%
> better than the competition.
There is no software that does what MGS does.
Not Bigo, not Dariush, not SmartGo and not Kombilo.
The features simply are not there.
Hence the number of copies I am still selling.
> I mean heck - I am a chinese 4 dan now, and I just gave you some
> really useful constructive criticism. Did you listen? No, you didn't,
> you called me a shit talking idiot. This is why you aren't doing so
> well with the program, Frank.
I have been steadily selling the program for the past 6 monthts w/o having
worked on it.
I am sure the principles of economics are lost to you, but that makes MGS a
real cash cow at the moment.
ricoh51
> There is no software that does what MGS does.
> Not Bigo, not Dariush, not SmartGo and not Kombilo.
> The features simply are not there.
Right Frank, Dariush is not MGS, Dariush(33%) is better than MGS(23%),
I proved it.
Good holidays
eric
-
Renli <usagi....@gmail.com> wrote:
> How do you know they're pros if they're anonymous?
They're players with pro rank.
> And frankly, unless they played other pros.. I am likely not
> interested in having a mid dan's moves poison my opening book.
That's not how instructional software operates. It would not
recommend mid dan moves. It applies amateur games to obtain
positional variations to better encompass a templated playing space.
> ... do some more work on it and stop being an idiot on RGG.
> I already told you what you SHOULD be doing. You don't have
> to listen to me, but very few other people are going to give you
> the kind of advice I have.
I have given him similar advice. But such advice is free, isn't it?
- regards
- jb
----------------------------------------------------------
Earthquake Sets Japan Back To 2147
http://www.theonion.com/content/news/earthquake_sets_japan_back_to_2147
----------------------------------------------------------
Frank de Groot
> proved it.
MGS has a 42% pro-prediction, so back to the drawing board for you.
(Besides, predicting good moves is only a small part of MGS's
functionality.)
subscr...@online-go.com
Gosh, I came here from an external link and have just managed to be
put off a piece of Go software for life.
Dariush however, whilst having certain slightly clunky feelings
aspects of the interface, is still my primary .sgf editor and score
analyser (Eric, your "analyse score" simply rocks, keep up the great
work).
On that point, if it can get close enough to give me a better game
(recently I won by 100 against 5 handicap stones, and I'm only ~5k
EGF), I would be more than willing to voluntarily donate money purely
to support a great engine.
PS Eric, by some crazy coincidence I played someone that claimed to be
your friend at Bracknell tournament 2007 in the UK (on a short trip
over from France at the time) - I hope he passed on my compliments as
I asked ;)
Frank de Groot
> Gosh, I came here from an external link and have just managed to be
> put off a piece of Go software for life.
I hope you die soon :-)
topazg
> On Jul 31, 12:14 am, "Frank de Groot" <fr...@moyogo.com> wrote:
>
> > <subscripti...@online-go.com> wrote
>
> > > Gosh, I came here from an external link and have just managed to be
> > > put off a piece of Go software for life.
>
> > I hope you die soon :-)
>
>
> -
Yeah, nearly persuaded me to see the error of my ways, but .. nah, on
second thoughts....
And on this note, I would like to eat my words about Dariush. I was
playing 5.2.0.1 when I pushed it around with 5 handis, and have just
scraped my way to a 20 point win against 6.0.0.0 on an even board.
Time I put my money where my mouth is.
Great work Eric :-)
ricoh51
> On Jul 23, 9:53 am, "Frank de Groot" <fr...@moyogo.com> wrote:
>>> Right Frank, Dariush is not MGS, Dariush(33%) is better than MGS(23%), I
>>> proved it.
>>
>> MGS has a 42% pro-prediction, so back to the drawing board for you.
>>
>> (Besides, predicting good moves is only a small part of MGS's
>> functionality.)
>
> Gosh, I came here from an external link and have just managed to be
> put off a piece of Go software for life.
>
> Dariush however, whilst having certain slightly clunky feelings
> aspects of the interface, is still my primary .sgf editor and score
> analyser (Eric, your "analyse score" simply rocks, keep up the great
> work).
Thank's, yes it is very useful, and now the "Pro move" and the
"Pattern" buttons can improve your game
>
> On that point, if it can get close enough to give me a better game
> (recently I won by 100 against 5 handicap stones, and I'm only ~5k
> EGF), I would be more than willing to voluntarily donate money purely
> to support a great engine.
We are still working on the engine, in particular, integrating the pro
moves to the engine, but the very good moves given by the pattern
engine are rarely at the Dariush level :(
> PS Eric, by some crazy coincidence I played someone that claimed to be
> your friend at Bracknell tournament 2007 in the UK (on a short trip
> over from France at the time) - I hope he passed on my compliments as
> I asked ;)
Yes, it was Julien (Sai51 on KGS), many thank's
eric
ricoh51
> And on this note, I would like to eat my words about Dariush. I was
> playing 5.2.0.1 when I pushed it around with 5 handis, and have just
> scraped my way to a 20 point win against 6.0.0.0 on an even board.
>
> Time I put my money where my mouth is.
>
> Great work Eric :-)
Thank you very much :)
And if you have really too much money :)
http://ricoh51.free.fr/go/doneng.htm
PS : Are you the topazg from OGS?
subscr...@online-go.com
>
I am indeed, that is my current contribution to the Go community as a
whole :)
Why, do you play on there?
Best Regards,
- Graham
ricoh51
>>
>> --http://ricoh51.free.fr/
>
> I am indeed, that is my current contribution to the Go community as a
> whole :)
>
> Why, do you play on there?
I have an account on OGS, good job ;)
eric