Home » Forum Home » Analysis

Topic: It is time to review Karlw's Opening Theory.
Replies: 31   Views: 117,573   Pages: 3   Last Post: Apr 5, 2009, 7:29 AM by: zoeyk

Search Forum

Back to Topic List Topics: [ Previous | Next ]
Replies: 31   Views: 117,573   Pages: 3   [ 1 2 3 | Next ]
zoeyk

Posts: 2,220
Registered: Mar 4, 2007
From: San Francisco
Age: 45
Home page
It is time to review Karlw's Opening Theory.
Posted: Nov 26, 2008, 10:57 PM

It is time, to review Karlw's Opening Theory.
I feel that much of this is valid,..but it is not pefect.
there seems to be errors. Let us all work together to make the ultimate Opening Theory for Pente.

All are welcome to help. and i ask and invite that Karlw, Up2ng, Watsu, Richardiii, Nosovs, Jozso, Melanie, Arczi, Kolia, Sjustice, Brf, Blaxxx, Gnoza, Spifster, Spavacz, Bonhart, Jayhawklov, S3v3n, Partica, Dweebo, Dmitriking, Snut, Bigboy, Niobium, Vilumisiek, Mersenne, Lepews, Karencita, Lisa_a, Awakening, Stepanov, Mike321, Mmammel, Kmbailey, Squirllking, and Homebrew,.. All contribute to this descussion.

(the names were in no order, placed at random)

if ive forgotten any one such as players from the BK website, my appologies in advance.

Perhaps to start the descusion off i will talk about the N8 opening...

White did N11 as a second for a long time. Nosovs Richardiii, and Karlw to name some of the main players helped to make this masters response popular. then one day at BK Nosovs pulled a descovery move, which Kolia has replayed Vs Richardiii here. a flaw exploit of Whites third at K11.
Now all of a sudden most of the High Skilled players have switched to using N12 as whites second move. I'm not sure how every one realized to do this at the same time...perhaps there was a Pente meeting that i wasnt invited to. if so next time ide like to be in the loop thanks
or perhaps once i publicized the following game in the P2 hall of fame games thread that every one went searching for a new second,...and of course every one went to see Nosovs new second,...Nosovs is such a trend setter lol. Any how it seems like a very solid White second.

heres the game of the White N11 White K11...



Now,....in Karlw's Pente Theory he says that for the N8 that A SJ (special jump) (N11 as example) is the best way to answer.

Group B: M10, N10, M8, N8
For the SJ: The Capital T (K10-N11-N9 shape), the Lowercase T (K10-N11-M12 shape).
Group B -> 3-Point or Special Jump.


but now it seems not,...
this is one of the many suspect things that brings me to talk about Opening Theory. We just dont seem to have a perfected Theory yet, but i feel we are close.

The following is Karlw's Pente Theory; Lets get to work!




Karlw's;
"Pente Opening Theory"



Turns out that while distance from the keystone is the best way to classify openings, it is not the only way, so we must do a little rearranging based on certain traits that moves in a group have in common.

Group A: K9, L9, M9, N9

If white answers all four of these openings with N10 (and he should), the positions that follow are very similar (two identical pairs, actually), and lead to a very distinct level of play between two strong players. These openings are based on destroying white from the inside, trying to interfere with white's plans by building an open 3 early or blocking all his initiative.


Group B: M10, N10, M8, N8

These four openings might not look like they have much in common distancewise, but after black's 2nd and 3rd moves a common thread arises: In these games, unlike games in group A, black usually plays not to divert the keystone (in two of the openings it's not even possible,) but to build up his own momentum while at the same time defending white's. This is my favorite group right here.


Group C: O9, O8, N7, O7

These are the far openings, which are mostly based on the threat of building an open three which white must make a pair to block, then threatening that pair. I call this tactic "diversion," and it is a very important principle of openings, because white is unable to build momentum without the keystone. Now i must define momentum.






Okay let's define some terms!

Forcing move: A move which, if unanswered, leads to certain victory for the mover. Examples include: Extension to 4, Stretch 4, Split 4, Stretch 5, Split 5, Threat of 5th capture.

Initiative: The person who has control of the game has the initiative. White begins with the initiative because black must play defense if white knows how to build momentum. When a player has the initiative, it is in his best interest to keep the initiative by playing forcing moves and making trias. Plays by him are either sente, meaning they retain the initiative, or gote, meaning they give up the initiative.

There are two kinds of initiative, simple initiative and self-sustaining initiative. Simple initiative is exemplified by the open 3. Playing an open 3 gives you initiative, because although it's not a forcing move, the other player has to answer after he runs out of forcing moves otherwise he will lose in 2 moves (Open 4 -> victory). But after he blocks that open 3, you can extend or stretch, then...nothing. Simple initiative runs out.

Self-sustaining initiative, on the other hand, if played properly, never runs out, and 99% of the time leads to a win. It is such an important pente concept that I give it a new name:

Momentum: Momentum is probably the dominant concept of pente openings. It is a 3-stone shape that, although not giving the player any immediate initiative, creates so much momentum that, if not properly blocked by the other player, leads to certain victory.

Momentum is important because it is what forces black to play defense in the opening. If black ignores white, momentum will make him pay, because even though white has the 2nd move restriction, he can still build momentum in his first 3 moves. There are countless ways for white to build momentum, but the best way is to begin with either a 2J (K10 to N10), 3J (K10 to O10), BJ (K10 to N12), or Special Jump (SJ) (K10 to N11). Possible 3rd moves:

For the 2J: The L, the Hat(K10-N10-L12 shape).
For the 3J: The V.
For the BJ: The L, the Broken Wing.
For the SJ: The Capital T (K10-N11-N9 shape), the Lowercase T (K10-N11-M12 shape).

Now, considering white's potential to create momentum and black's ability to defend it, I would rank these 6 shapes as follows: L > Hat > Cap T > Wing > V > Low T. The 3J also has the ability to make the post 3, so it's still a good opening. The 2J can make a stretch 3 right away for immediate initiaitive, that along with the top 2 momentum shapes makes it the strongest 2nd move for white.

2J > BJ > 3J = SJ is my ranking for white's 2nd move.

This is something black needs to take into consideritation: The best openings for black are the ones that are strongest against the 2-point jump. This is why Group A isn't quite as strong as Group B and C, altough it is still a strong group of openings because black has thousands more options for defense in any position than the B and C openings.

I believe that of the 4 possible 2nd moves, one is better than the others for each group. These are my findings:

Group A -> 2-Point Jump.
Group B -> 3-Point or Special Jump.
Group C -> Broken Wing Jump.

This doesn't mean that for any group C opening, all BJ's are better than all 2J's 3J's and SJ's, just that it seems to me that the BJ is best suited for defending the Group C openings. I will next explain why I prefer these openings for these groups (it is rather complicated).




>I will next explain why I prefer these openings for these groups??

but yet to be further explained,..another sign that the Theory is unfinished...we have some work to do here.

Scire hostis animum - Intelligere ludum - Nosce te ipsum - Prima moventur conciliat - Nolite errare

zoeyk

Posts: 2,220
Registered: Mar 4, 2007
From: San Francisco
Age: 45
Home page
Re: It is time, to review Karlw's Opening Theory. All are invited.
Posted: Nov 26, 2008, 11:53 PM

If we do create a perfected Pente Opening Theory here,..
I will add it to the Wikipedia Pente Page @ http://en.wikipedia.org/wiki/Pente .

Scire hostis animum - Intelligere ludum - Nosce te ipsum - Prima moventur conciliat - Nolite errare
skanadron

Posts: 17
Registered: Sep 17, 2008
From: lancaster CA
Age: 17
Re: It is time, to review Karlw's Opening Theory.
Posted: Nov 27, 2008, 12:20 AM

I am definitely no expert but to me it seems that with the example you used with black using the N8 opening white's problems came from the third move at K11, not the second at N11. From what I can see all the K11 stone did was waste whites 5th move, making him place K12 or lose a pair. Perhaps the N11 opening would still be stronger than N12 if a different third was found, or perhaps N12 really is stronger regardless of the third move.

I am rather new to pente, and still rather weak with openings, so I would not be able to come up with a third move that would be better for N11, it just seems to me as though one must exist. Perhaps N9 or M12 forming ether the Capital or lowercase T from Karlw's list possible third moves.

up2ng

Posts: 542
Registered: May 9, 2002
From: Northeast USA
Re: It is time, to review Karlw's Opening Theory.
Posted: Nov 27, 2008, 3:30 AM

This is a very interesting read -- I actually came across this (again) a week or two ago on another website that zoeyk had created where this text is read out loud in audio format, which is kind of fun.

When I listened to it then and read it now, it dawned on my how different my strategies are than those listed here. While I don't have a fully systematized approach to the game as this is, I'll make a few comments anyway.

First, I believe this was created by karlw quite some time ago when he was a good player, but not as good as he is now -- so this might be out of date even by the author's own account.

Next, I believe there might be a typo here -- the lowercase T is described here as K10, N11, M12. I think this was probably lost in translation and I have a feeling karlw was referring to the shape created by K10, N11, L13 -- correct me if I'm wrong.





Now, with the game posted here, I agree that white's third move of K11 was not strongest and that white is probably still in the lead after the 2nd move. Of the shapes listed in this thread, there are two broken wings (N13 or K8), one capital T (K12), and two lowercase Ts (M8, L13). Unfortunately, there seems to be a black 3rd move to counter each of these shapes. So, white should look for other options.

Remember, shapes are just one category of tools in the toolbox for white to start its attack. There are many other important concepts at play as well such as lining up a cap trade, setting up keystone threats, knowing when to play a building move away from the action and when to take an inside position.

For third moves, there are a few basic concepts. First, it rarely makes sense for white to spread itself out too far too early in the game. For the game shown, it would be highly unusual and almost always incorrect for white to choose a 3rd move that is outside the rectangle created by the K and N lines and the 8 and 13 lines. In other words, white's first 3 stones should usually fit inside of a 4x4 grid, unless there is a compelling reason to try something else. Next, at least 2 of white's first 3 stones should form a "potential" -- meaning they should line up so that they can form a 3 on the next move. Potentials can be XX, X_X, or X__X.

So, moving back to the game shown, we can eliminate M13, L12, M9 and L8 since these don't create any potentials. Next, it's usually better if not all 3 stones are 3 spaces apart from each other -- this eliminates K13.

The remaining moves are: N12, M12, M11, L11, K11, N10, M10, L10, M9, K9.

From here you just have to have a sense for which ones are weak in the given situation and eliminate those and evaluate the others one by one by looking ahead at how your opponent is likely to respond. There are many concepts involved in this evaluation that is difficult to write down concisely. White had something specific in mind by choosing K11 -- it actually is not a bad idea, it just didn't work out in this case.

After looking for a bit, I think that a strong choice would be 3. L9.


K10,N8,N11,P10,L9


This does a few things. First, it sets up a cap trade that is favorable to white.


K10,N8,N11,P10,L9,M8,L8,J11,O8


OR


K10,N8,N11,P10,L9,M8,L8,L10,O8,L7


Second, if the pair is attacked but the trade is declined, white has continuation (and victory) with a broken wing shape.


K10,N8,N11,P10,L9,M8,L8,O8,L10,J11,L9,L11,M10


It also creates two potentials (a XX and a X_X) which line up nicely with each other to provide plenty of initiative to win if black ignores it completely and sets up its own building move away from the action. Black does not appear to have options for a building move which also fully counters the lines set up by these potentials.


K10,N8,N11,P10,L9,O9,M7,O12,N7,Q11,R12,L7,O7,M8,N9


OR


K10,N8,N11,P10,L9,O9,M7,P8,N10,M8,Q7,N12,L8


OR


K10,N8,N11,P10,L9,M10,M8,N7,N9,O10,J11,H12,N6,L7,N10


I play this type of move more often than anyone I play against regularly -- both as player 1 and player 2, but for different reasons in each case. It takes a long time and a lot of experience before effectively adding the XX potential to your opening arsenal, especially early in the game and seemingly unprovoked. And especially as player 1, since you'd really rather not have to play a move like this as player 1. But sometimes, it can be the strongest move. This is an example of playing to take the inside position, as I alluded to before. You're sacrificing some structural strength early in order to solidify your position, and to get your opponent further out of position in their responses. Initiative in these types of games comes a bit later.

Ok, let me step back now and give my take on openings in general (as Player 1). First, a comment was made previously that a lot of the best players were playing 2. N11 in response to N8 and now have recently switched to 2. N12. Well, you won't find me among that group. Why, you ask? It's simple. I had always claimed that 2. N11 was a mistake ever since it first started becoming popular. Now, I also believe 2. N12 is also a mistake. That's right, I believe that the very best players in the world are making a mistake on their very first stone!

Now, before you all cry foul at this bold statement, let me back up even further. Some basic facts about stone games. Fact -- unless the game board is extremely small, Player 1 should always win with unrestricted rules. This is true for Pente, Gomoku, Connect6, Go or any other stone game with unrestricted rules. There are mathematical reasons for this which are undisputable. Now, once move restrictions are added, if they are not restrictive enough, player 1 should always win -- if they are too restrictive (or if a swap option is involved) then player 2 should always win. There is no inbetween and there are no ties (unless the game board is very small).

Under Pro-Pente rules -- the most common ruleset for serious Pente players, Player one must play it's second move at least a 2-gapper away from it's first stone. My opinion is that this is still so far skewed in favor of Player 1 that Player 1 should always win relatively easily. Just watch any two Pente masters play a serious game of D-Pente and you will see how wildly in favor of Player 2 the openings appear to be, just to try and make it more balanced.

With that being said, I believe that it is pretty difficult to choose a losing 2nd move as Player 1. For any black 1st move, there might be as many as a dozen winning second moves available to white -- yes that's right. However, among all these winning moves, some are certainly vastly superior to others.

So, going back to the Pente Opening Theory discussed here, I observe that of the 11 or so reasonable 1st moves as black -- white responds "off-axis" on 7 out of these 11 situations! Here is where I differ most severely with other expert players. Although I appreciate all the effort to classify black's first moves into Group A, Group B and Group C and all the different responses to the various situations, my 2nd move works this way:

WHITE'S SECOND MOVE IS ON THE VERTICAL OR HORIZONTAL AXIS, 2-GAP AWAY FROM THE CENTER STONE.

Every time.

That's right -- for every possible first move as black, white's second move should be one of 4 possible locations on the board -- N10, K7, K13 or G10. That's it, it's that simple really. Of these 4, there is generally 1 (or 2 if the position is symmetical) that is superior.

In fact, a very highly respected (retired) Pente Master was so sure about this that a new variation of the game was created -- G-Pente -- which bans these moves (and on-axis 3-gappers) from white's possible second move list. In addition, my opinion is that 2-gapper on-axis 2nd moves are vastly superior to 3-gapper on-axis moves in every situation, and in some situations 3-gapper on-axis moves are actually losing moves. They are much weaker than nearly everyone believes and should always be avoided.

There are a few reasons for playing on-axis 2-gapper moves. First it creates a potential, which is an important building block in Pente. In particular, it creates a vertical or horizontal potential. Remember (or learn) that vertical / horizontal trias are stronger than diagonal trias. From an on-axis 2-gapper 2nd move, there are 4 ways to create a L-shape with the 3rd move -- the strongest opening shape possible. The reason it's so strong is that it can create 2 different vertical / horizontal trias with the 4th move -- no other shape can do this. If the one-gapper part of the L shape is placed near an opponent's stone, it can create a keystone threat, a hammer-like position or a number of other strong possibilities.

If the opponent plays an opening that would counter an L-shape with a block or a keystone attack, there are an additional 4 ways to create a Hat shape, probably the 2nd strongest opening shape. If none of these appear strong, there are another 4 ways to create an inside line / cap trade position (i.e. "The Wedge").

Well this post is getting way too longwinded to go into how the 3rd or 4th moves should be evaluated, but at least I've put my two cents into White's 2nd move -- take it or leave it.

A quick rundown of strong 2nd moves in response to black's first move:

L10 --> K13
L9 --> N10
M10 --> G10
M9 --> N10
M8 --> K13
N10 --> G10
N9 --> N10
N8 --> K13
N7 --> K13
O10 --> G10
O9 --> K7
O8 --> G10
O7 --> K13

I guess that's it for now until we discuss 3rd and 4th moves. It is interesting how different people will have such different opening theories, but that's what makes the game fun to play -- if everyone played it the same way, it would be boring. Have fun!


Message was edited by: up2ng at Nov 27, 2008 11:31 AM
(Board images added)

karlw

Posts: 968
Registered: Mar 7, 2006
From: Eugene, Oregon
Age: 32
Re: It is time, to review Karlw's Opening Theory.
Posted: Nov 27, 2008, 7:58 AM

First, I believe this was created by karlw quite some time ago when he was a good player, but not as good as he is now -- so this might be out of date even by the author's own account.

Very true.

Next, I believe there might be a typo here -- the lowercase T is described here as K10, N11, M12. I think this was probably lost in translation and I have a feeling karlw was referring to the shape created by K10, N11, L13 -- correct me if I'm wrong.

Also true.

You write excellent pente analyses, dean. As for your bold assertion that the best answer to any black 1st is a 2-point jump, I have to say that that is the direction my game is heading. The only I exception I might make is sometimes the "special jump" is very strong because it is the closest to an axis jump and it gives white an equally diverse set of options for attack. As an answer to M10, O9, and maybe a few other long-distance attacks I think you should consider doing a special jump.

It must feel great to be able to say "I told you so" regarding N11. Maybe you are right, maybe it is flawed and a move like G10 or K13 is stronger. But I think your point that for any black 1st there are probably a dozen winning responses is even more important: more and more I'm thinking that really, black's 1st and white's 2nd don't matter all that much (provided they are reasonable); what really matters is the next few moves, where white solidifies his attack and black begins to parry and counterpunch. That is my 2 cents.

L10 --> K13
L9 --> N10
M10 --> G9*
M9 --> N10
M8 --> K13
N10 --> G9*
N9 --> N10
N8 --> K13
N7 --> K13
O10 --> G9*
O9 --> J7*
O8 --> G10
O7 --> N12*

*: different from up2ng

It's no use going back to yesterday, because I was a different person then.
karlw

Posts: 968
Registered: Mar 7, 2006
From: Eugene, Oregon
Age: 32
Re: It is time, to review Karlw's Opening Theory.
Posted: Nov 27, 2008, 8:22 AM

Next we should probably discuss 2...M8, probably my single greatest contribution to pente theory. Right now I am a fan of 3 M12. Any one want to register a dissenting opinion?

It's no use going back to yesterday, because I was a different person then.
zoeyk

Posts: 2,220
Registered: Mar 4, 2007
From: San Francisco
Age: 45
Home page
Re: It is time, to review Karlw's Opening Theory.
Posted: Nov 27, 2008, 2:58 PM

2...M8 3 M12
can you give that in a line ie; 1)K10 - ??? 2)??? - ??? 3)??? - ??? ext...
or perhaps post it as a game board to visually view,..
dweebo made a nice tutorial of how to do that.
thanks

Scire hostis animum - Intelligere ludum - Nosce te ipsum - Prima moventur conciliat - Nolite errare
zoeyk

Posts: 2,220
Registered: Mar 4, 2007
From: San Francisco
Age: 45
Home page
Re: It is time, to review Karlw's Opening Theory.
Posted: Nov 27, 2008, 8:01 PM

i was very impressed by your post up2ng.
on axis 2 jump for everything,...it can be done, yes
is it minimal struggle? every time?....
im not sure bout that..
the seconds should be minimal struggle,..for a givin first move of black, can white have multiple equally minimal struggling moves for its second both on axis and off?
on axis 2 jump for everything seems simplistic, and im not saying that its not as easy as that, but lets be honest,..is there a off axis that can equal it, perhaps even one up the on axis move?
im mainly speaking in terms of responding to those black first moves which are mid to long range K10 keystone attack threats.

and thank you for adding the Game board images. very nice.

Scire hostis animum - Intelligere ludum - Nosce te ipsum - Prima moventur conciliat - Nolite errare
watsu

Posts: 1,427
Registered: Dec 16, 2001
Home page
Re: It is time, to review Karlw's Opening Theory.
Posted: Nov 28, 2008, 2:48 AM

I think I basically agree with Up2ng on the axis 2 jump and I'll say a bit about why...
Considering Pente as a solvable win for white in x moves or fewer game, a computer program/database could theoretically be constructed which would win against any black line. This (IMO) would be easiest to do by considering the minimum number of cases. Due to its versatility in finding a good reply to any black second (10-12 good possible options when wedge thirds and split trias are added to Ls and hats) a two jump on axis is clearly a favorite to minimize number of game lines which have to be explored- at most 4 possibilities for any given black first move- with symmetry often fewer.

This is not to say that factors such as minimal struggle or choosing a suboptimal response in order to take the game out of the studied realm aren't going to be factors in play against human opponents, but purely from a (computer assisted) game theory standpoint, I think proving a winning line for any black first would take first precedence- fine tuning by finding minimal number of moves would happen after white has been shown to have a winning reply for any black first, which as I've stated above I believe would be easiest by defaulting to a two jump on axis.


Message was edited by: watsu at Nov 27, 2008 8:48 PM


Retired from TB Pente, but still playing live games & exploring variants like D, poof and boat
zoeyk

Posts: 2,220
Registered: Mar 4, 2007
From: San Francisco
Age: 45
Home page
Re: It is time, to review Karlw's Opening Theory.
Posted: Nov 28, 2008, 9:36 AM

i just noticed your guy's black first move at O10,...
is that a typo? i dont think ive ever played that move before,..does it really hold enough value to be in the list?


and when comparing M10 --> G9* to N10 --> G9*
I think white will struggle slightly more in the N10 opening, and the M10 opening will be slightly easier/faster to beat.
but i could be wrong..

Scire hostis animum - Intelligere ludum - Nosce te ipsum - Prima moventur conciliat - Nolite errare
karlw

Posts: 968
Registered: Mar 7, 2006
From: Eugene, Oregon
Age: 32
Re: It is time, to review Karlw's Opening Theory.
Posted: Nov 28, 2008, 10:57 AM

O10 is basically the best kept secret in black openings. let's keep it that way.

It's no use going back to yesterday, because I was a different person then.
zoeyk

Posts: 2,220
Registered: Mar 4, 2007
From: San Francisco
Age: 45
Home page
Re: It is time, to review Karlw's Opening Theory.
Posted: Nov 29, 2008, 2:47 AM

you know the 3 blind men and the elephant story?
well,..im looking for the snake the tree and so on, but im also looking for the whole elephant too in regards to pente theory. a honest whole from all perspectives.

for every black first, white has a broad array of white winning seconds,...and those i believe have values in terms of struggle/amount of moves black can push the line.
and amount of creativity and confusion for white to find that winning path. in a theory i would like to state for a given black first all the winning white seconds, and all the losing white seconds regardless of struggle.
then ide like to pin a value to each of the winning white seconds in response to the black first.
and do this 12-13 times over for all the black firsts.
this would be the opening bible, that regardless of a individual player's style, it could not be disputed or un-usable, because it would represent all styles.

but the value,.. yes the value, thats a tuff one. what human is good enough to make that determination? who can see that far and accurately? it would take a team of masters to accomplish this task. i doubt even nosovs could chew that big of a bit on his own, although i could be wrong.

im willing to try if others are willing too.

~zoey

Scire hostis animum - Intelligere ludum - Nosce te ipsum - Prima moventur conciliat - Nolite errare
watsu

Posts: 1,427
Registered: Dec 16, 2001
Home page
Re: It is time, to review Karlw's Opening Theory.
Posted: Nov 29, 2008, 4:11 AM

Yeah, the value of lines of play is a toughie. If there were a good easy way to formulate such a thing, we could play random position duplicate Pente and open up a whole new realm of possibilities for the game. I suspect that a computer program such as MMammel's with the ability to assign numerical preference values to any given move after the opening 6 stones are placed, searching out nine full moves would be a good place to begin. Giving it the best possible opening book would be the first step there.

Retired from TB Pente, but still playing live games & exploring variants like D, poof and boat
up2ng

Posts: 542
Registered: May 9, 2002
From: Northeast USA
Re: It is time, to review Karlw's Opening Theory.
Posted: Nov 29, 2008, 5:57 AM

This has been an interesting discussion so far.

karlw: "more and more I'm thinking that really, black's 1st and white's 2nd don't matter all that much (provided they are reasonable); what really matters is the next few moves, where white solidifies his attack and black begins to parry and counterpunch."

Yes, this is exactly my thought about openings also. While it is possible to blunder white's second, it is unlikely. Most of the time white will still maintain a winning position after the 2nd move and a wide range of possible winning 2nd moves are available. However, evaluation of white's 3rd move (and next few moves thereafter) is absolutely critical, and is the first real place where white can lose the game. In fact, if black plays a strong opening, there may only be a small number of winning 3rd moves possible for white to continue on with. Plus, the evaluation process in many ways becomes more complex and therefore white is more prone to commit errors in this stage of the game.

Regarding the board position karlw mentioned, I believe he was referring to this:


K10,K9,N10,M8,M12


karlw and I recently played a turnbased game which landed on this position. I played black and studied this position for quite some time and could not find anything that would even cause white much trouble. I believe this is a winning position.

watsu: "Considering Pente as a solvable win for white in x moves or fewer game, a computer program/database could theoretically be constructed which would win against any black line. This (IMO) would be easiest to do by considering the minimum number of cases.
...
I think proving a winning line for any black first would take first precedence- fine tuning by finding minimal number of moves would happen after..."

While this is not exactly the reason why I've chosen the 2nd moves that I have, watsu and I are on the same page here about what it means to solve the game. Really all that is required is to assign one single winning move for each possible black move for white's 2nd, then 3rd, then 4th and so on until the game is won under every scenario. While it may not be "best", as long as the move is a winning move it can be part of a "solution" for the game.

Also, I don't necessarily agree that the move that results in the fastest win is the "best" move. While I do place some virtue in finding the fastest win, I wouldn't go so far as others have in the past and advocate fastest wins as a tie breaker in tournament play, for example.

zoeyk has also used phrases such as "struggle" and "amount of moves black can push the line" in evaluating white's choices. Again, this is slightly different from how I view a "best" move choice. By struggle, I interpret that to mean the liklihood of losing initiative, or having only narrow numbers of (possibly complex) options in order to maintain initiative. This is certainly important and moves that lead to increased struggle should be avoided.

I would characterize a "best" move to be one that maintains a winning position while reducing complexity. This often means adding flexibility -- increasing the number of variations that would lead to victory -- and strengthening momentum / initiative, but mostly it means creating a path to victory that is easily seen and followed. This may take an extra move or two than a more direct and complex line, but increased complexity leads to increased likelihood for mistakes on future moves. For computers that can literally search far enough ahead to see a sure win, complexity doesn't matter at all since computers are unphased by it, but humans think differently than computers and complexity is very important in human strategy. These concepts are the major factors in my listed choices for white's second move.

zoeyk: "in a theory i would like to state for a given black first all the winning white seconds, and all the losing white seconds regardless of struggle.
then ide like to pin a value to each of the winning white seconds in response to the black first."

This is the first time I've really seen this idea for what it means to solve the game. It is certainly more comprehensive than what I described above. Some would say overkill, but I can see value in it -- especially when thinking in terms of D-Pente, for example. This would also be much more ambitious to achieve accurately and completely as there is just a tremendous number of permutations possible in the first several moves of Pente. Although if this scoring system were done only for all possible 2nd moves (and not subsequent moves), it might be a reasonable thing to complete.

(As an aside, if you open and review mmammel's opening book that is used by his computer AI, each opening actually is scored -- the 3 choices used are "slightly favors white", "approximately even", and "slightly favors black". Although this is a bit different from what you are proposing, it's interesting to see that a similar effort has been done before. Of course, these openings and their scores are out of date, some are known to be inaccurate, and many openings are missing or incomplete. Also, in my opinion, there is no such thing as an "even" position in Pente. There are only winning positions, and losing positions. I think winning positions could be scored further as you've described so that some are tough wins and others are easier wins.)

watsu: "a computer program such as MMammel's with the ability to assign numerical preference values to any given move..."

Yes, I was thinking of this also. MMammel's program does a decent (but not perfect) job of evaluating and scoring the strength of different moves based on how the position might end up x frames into the future, where x is the program's "level". I'm not sure exactly what the algorithms are to comparing and scoring different positions, and I'm sure this information is proprietary, but again it performs pretty well at this task. The main problem, of course, is that the computer only looks so far ahead and then is forced to compare board position snapshots, whereas a human can generally have a sense for whether a given line will lead all the way to a victory. When I first started playing, I heard some masters at that time claiming that the longest black could be expected to hold out was approximately 27 moves in a perfectly played game looking for a fastest win. This means once we have computers powerful enough to play mm_ai54 (level 54) against itself in a reasonable amount of time, the game will be solved. My personal opinion at this time is that 27 moves is generous for black and that white should expect to win a few moves faster than this. But it's still a game that's far too deep for computers to solve by brute force and so it's still up to us to make it happen!

watsu

Posts: 1,427
Registered: Dec 16, 2001
Home page
Re: It is time, to review Karlw's Opening Theory.
Posted: Nov 29, 2008, 6:32 AM

Nice post, Up. Just to clarify on my reasons for preferring the 2 jump move- it allows for the fewest number of positions to be considered for a second while allowing for the greatest number of strong/winning thirds. Which I think is related to coming up with a viable "solution" to tournament rules Pente in the near future, but I think in a game situation I usually simply pick one of the the axis moves due to its combinations of strength, flexibility, minimization (of possible lines)and ease of recall and the sense that "one of these doubtless has a winning line somewhere in its future, provided I can only find it."


On AI and scoring, I think a possibly fruitful approach would be something along these lines-
1.) get the experts to do an overhaul/update on the opening book
2.) set up a program similar to some of the prime number search programs, so that a computer in screensaver mode could analyse a single move or two in a position each time it had a few minutes to spare
3.) failing #2, at least set up a place where people could post results of how AI played AI for each given opening in the book
4.) keep an eye out for game situations where white has a VCT win in 9 or fewer moves- obviously a solved position- so that the moves between the opening book position and the VCT are the only moves which have to be looked at to see whether black played best defense.


Message was edited by: watsu at Nov 29, 2008 12:46 AM


Retired from TB Pente, but still playing live games & exploring variants like D, poof and boat
Replies: 31   Views: 117,573   Pages: 3   [ 1 2 3 | Next ]
Back to Topic List
Topics: [ Previous | Next ]


Powered by Jive Software