game analysis Dots and Boxes

13 replies. Last post: 2015-02-21

Reply to this topic Return to forum

game analysis
  • purgency at 2015-02-18

    I’ll post some comments to games from championship 37 here... merely some quick notes that I write while lurking without involving too much thinking. thoughts and questions(!) are appreciated since I’m going to rush these very quickly which is why some stuff won’t be very clear and/or incorrect. You can also use this thread to request an analysis of whatever game you find interesting or do your own one (doesn’t have to be from the championship). I hope this may be helpful for the players to see their mistakes and use it to improve so I’ll criticise more than praise and focus on showing off mistakes rather than to explain the good moves.

    ——————

    1685926 Geoff vs romei http://littlegolem.net/jsp/game/game.jsp?gid=1685926
    1.h7
    I personally don’t like it as it’s weak against the 3 chain possibility due to cutting off that upper right corner. 2.g8, 2.e8 or 2.e4 might be good to aim for 3 chains and it should become quite hard for player 1 (which doesn’t mean going for 3 chains is the best thing to do).

    2.j9
    Not totally bad, prevents a possible loop threat which should favor player 1 that is done by combining k8 with h11, but I’d say there’s way stronger options for move 2 e.g. 7ics’s d3 or g8. (the strength of this move might be arguable)

    3.i8 jup

    4.g6
    not a good move, helps Geoff building two areas since we get a clear split between top left and bottom right... better would’ve been a move like i6 for an approach like this
    http://www.trmph.com/dnb/board#5,h7j9i8i6j5c8d7c6b7c4b3d3e2f11j11f9f7c10e10g10f3g4j3i2h3h11b11a2k2j1g2f1h1c2b1d1b5a6a4b9a10a8j7k8k6k4i4h5g6e6e4d5f5d11

    5.i4
    I guess this is okay. You can play a move like this if the option for one chain is no more as it should be the case here. I wouldn’t play i4 though if the position was 1.g6 2.h7 and Geoff does that, because that’s problematic when 1 and 3 chains is possible since i4 is allowing for connection of the bottom right into the center which does help getting one chain.

    6.f7
    Well, this is lost for sure with top and bottom being clearly seperated from each other and both regions being big. I think I might have tried 6.c6 here hoping for something like this http://www.trmph.com/dnb/board#5,h7j9i8g6i4c6h3c8f5f3d5d9d11f1a8b5c4e10c10a10h11h9g8f9g10f11i10e8e6d7f7a6b7b11b9j3d1d3e4g4h5i6j5k4i2b1a2k2j1k10j11j7k8k6h1
    Although as long as Geoff wouldn’t play 7.h3 there should be no danger really.
    ——————

  • The_Shark_c at 2015-02-19

    The Shark opinion:

    I am uncertain who was ahead after moves 1, 3, and 7.

    At all times during the rest of the game, Geoff was ahead.

  • purgency at 2015-02-19

    1685954 Tobias Lang vs romei

    1.i2
    I don’t like it since it doesn’t really help in building a clear structure. For starters it’s helpful to have an aim to pursue like “one chain at the top and one at the bottom” or something like that.
    2.k2
    Okay, so this move isn’t really bad... it was actually helpful in this game since it was used to build that 4-chain. In move 5 it would’ve been nice to play j3 but k2 turned that into a sacrifice so it had a pretty nice effect.Nevertheless, I’d say that k2 is suboptimal since it just lacks aggression. This was the problem with 1.i2 as well: those moves aren’t really bad in the sense that they favor your opponent (they might even turn helpful), but they kind of waste the turn since you could’ve done something else that favors yourself even more.The only effect of k2 here is that it builds an end for a potential chain in the bottom right corner, but the j2 and j4 boxes might as well turn into a 2-chain at this point so you might actually have to defend the effect of this move, see:
    http://www.trmph.com/dnb/board#5,i2k2j5k4h1
    after 3.j5 theres i4 threatening to build a 2-chain so k4 is the only way to prevent it and then 5.h1 builds a structure that is pretty good looking for player 1 since it extends that corner so it might take up the entire bottom.
    If you want to go for 3 chains then I recommend to just play something like g2 right here which is ensured to show its effect. The best strategy though is to play in a way that allows you to switch between aiming for 1 and 3 chains e.g. with 2.c8 which was used by 7ics in his latest games against the_shark.

    3.k4
    This move isn’t good I think. It looks like it could be used for some kind of ladder strategy to extend a chain along the right but I think that’s not a good way to go with i2 already played. It’s actually being misused by 4.g2 and supports the effect of 2.k2 since the 2-chain isn’t possible anymore.

    4.g2
    That’s a good one

    5.j5
    That’s a bad one since it continues to build the chain there. Something like
    http://www.trmph.com/dnb/board#5,i2k2k4g2f3g4i4i6k6d1c2g6h9f7f9d7c6h11i8j9b5j11i10g10k10d11c8a8e10f11b3a10b11c4a4a2b1e2d3f1e4e6d5f5j1j3j5k8j7h1
    might work, although the position already isn’t the easiest anymore and I’m sure player 2 has pretty good chances if he plays well.

    6.c10
    That’s kind of random again. If you wanted to repeat what happened in that bottom right corner and build a chain then that’s not how you do it even though the top left turned into a 3-chain later.
    If you want to increase the number of chains then you can continue an approach like the one from 4.g2 by playing something like 6.e2 or 6.j7. If you want to cut of that top left corner I recommend using something like e10 maybe that would look like this:
    http://www.trmph.com/dnb/board#5,i2k2k4g2j5e10d9b7b9b5
    Your c10 can be abused by d11.

    7.b9
    I think this is probably okay to play´although I prefer d11 which however can still be played later and b9 actually threats with a8. As a downside it increases the strength of b7 since c10+c8+b7 can’t form a 2-chain anymore and instead forms another endpoint at the top left for a chain.

    8.b7
    Yeah, that’s good.

    9.j9
    Nah, that’s random again. d11 should be played here maybe c6 to extend that corner... you definitely need to extend some more if you want to get 2 chains here and 4 chains probably isn’t even worth considering.

    10.e8
    I prefer e10 here since d11 seems to be the best move for player 1 and should be prevented. It’s not bad though.


    Okay, I’ll stop the analysis here. Different from what’s being done I think player 1 should still aim for 2 chains even now and the h2 sacrifice is most probably necessary to do that. I can’t tell who is winning here.

  • Loony at 2015-02-19

    Good points purgency. I personally feel like player 2 is already winning after move 10. Let’s see if Shark can confirm this. :)

  • Sean_Hettenbach at 2015-02-19

    Well I’ll state this, I don’t like games where there is a lack of control on the center of the board. The game is just way to open for comfort in my opinion. I like to cut the board into sections as soon as possible and begin planning what I want the board to look like. Even moving into one of the 4 center moves after move 2, the board still has some funky possibilities.

    (3.f5) This looks good to me as first player trying to keep some order on the board and a sense of control, a fight for the center has at least been established. Going further, first player will now be the one deciding what the bottom portion looks like according to a second player response in the surrounding area. for example:

    http://www.trmph.com/dnb/board#5,i2k2f5h5f3h3e2 http://www.trmph.com/dnb/board#5,i2k2f5g4h3 http://www.trmph.com/dnb/board#5,i2k2f5e4h3h5 http://www.trmph.com/dnb/board#5,i2k2f5i4g2h5j5

    Now I’m not really good at predicting opening moves yet, so these are not really responses I’d expect. I’m not to familiar with opening theory, but I’m getting better with it. Except my point here is only responses pertaining the bottom portion. Player 1 can respond to any threats to the bottom region with a pretty good idea of what the board will look like as a result of his decision.

    (3.e6) Well, not as pretty looking as f5, but still, it’s a move in the center. Having said that, there is just to many possibilities of combinations deciding what that bottom right portion of the board is going to look like. Where as before, there were only a handful of moves between f5 and k2&i2, now there are almost double the amount of move in between moves 1&2 and 3. The only good thing is the fact that it is a center move.

    (3.f7) Pretty much the same thing as e6, just different direction and flow of moves. The top portion and left side of the board have a lot of “what ifs” and just looks like a headache to me analysis wise, so I won’t bother for the moment haha.

    (3.g6) Finally the fun one! This move looks promising to me. It’s a center move, and it can combine with i2&k2 pretty nicely. Examples:

    http://www.trmph.com/dnb/board#5,i2k2g6i6j5k4h3g4f5g8g10 http://www.trmph.com/dnb/board#5,i2k2g6h5j5k4h1

    Heres one exampleas to where I’m not to sure what a good response might be: http://www.trmph.com/dnb/board#5,i2k2g6h1

    Anyways, I still have a lot learn about the opening game, so these might be bad moves altogether, I do not know haha! When ever I play an unorthodox first move, I usually follow with a center move to give the first move some structure. A base to build the center on if you will. When I see games played around the edge first, I get a little uneasy about the un-clarity of what the center will look like, and all the possibilities surrounding the edges. Anyways that’s my little input! I’d like to know if the moves I presented made any sense, and had some truth to them. Thanks!

  • The_Shark_c at 2015-02-19

    Shark opinion about 1685954:

    Player 1 is ahead at the start.

    After 5, I do not know.

    After 6, player 1 is ahead.

    After 9, player 2 is ahead.

    After 10, player 1 is ahead.

    After 11, player 2 is ahead.

    After 16, player 1 is ahead.

    After 17, player 2 is ahead.

    After 18, player 1 is ahead.

    After 19, player 2 is ahead until the end of the game.

  • Sean_Hettenbach at 2015-02-19

    Just lol at the shark response. I like that it tells us who is winning, but what are the moves that are winning for each player at said move!?

  • Christian K at 2015-02-20

    Very interesting that it swings so often. I also would have thought that geoff was ahead after move 6 and stayed that way for the rest of the game.

  • William Fraser at 2015-02-20

    That’s probably true of Geoff.  (Although the Shark was unable to tell after move 7).
    The back-and-forth swings were in Tobias Lang vs romei.

  • Sean_Hettenbach at 2015-02-20

    Hey William, how does the shark respond to these threads? Do you post for the shark or is it coded to respond as it does? Either way, is there any possibilities of it allowing us to see its move trees that it comes up with for each of the positions? Also, how does it score a player as winning or losing?

  • William Fraser at 2015-02-20

    I post manually.
    I’ve been a bit worried about how much I should post, as the shark is a bit of an oracle.
    I think before I start putting up large quantities of data I should start a thread about “ruining the game”.

    The Shark is mostly a 10TB  (More than 35,052,891,301,382 position) database of endgame positions (where endgame means that both j3 and c10 have been played).


    In addition, there is an opening book  of positions (393,180 positions) which have been proven using the endgame database and a light-weight look-ahead engine.  One such position is the starting position, which is known to be a first-player win.
    For each position in the endgame database, it knows the theoretical score and adds it to the boxes already taken to determine who should win and by how much.
    In the opening book, the same is true, except that it uses alpha-beta with cutoff, so it does not always record the margin of victory correctly (i.e. a game in which player 1 should win by 5 boxes is the usually the treated the same as one in which player 1 should win by 3 boxes).
    Otherwise, it uses the look-ahead engine to determine upper and lower bounds.
    I do have the ability to list, for a given position, for each legal move, how many followers lead to wins, how many to losses, and how many to unknown positions.

  • Sean_Hettenbach at 2015-02-20

    Ah shucks, computers make my human brain feel in-superior. Haha, though, the physical amounts and capabilities of human surpass most computer funtions, its the knowledge base that I think us humans sought out the most. The ability to memorize and to solve. The ability to know the outcomes. But then what would the point of playing a game be? If we knew what move was best and the outcome of over 35,052,891,301,382 positions, it wouldn’t be fun to us. Exploring the un-known, and discovering for ourselves the outcome of a given situation... So William, I am very grateful that you have created such a very program that can be used to better our knowledge of the game, and understanding of it. You say you are worried about people looking at the shark as an oracle, when that’s pretty much what it is! Haha, It is pretty much the dots and boxes guru. Nothing to worry about, and I think I’m speaking on the behalf of everyone else when I say to post as much as you think can help us out with any problems we come across, or any questions we might have. Thanks again William!

  • purgency at 2015-02-21

    @Sean regarding 1685954 Tobias Lang vs romei

    3.f5, 3.e6, 3.f7 and 3.g6 all look like good moves... I think that after 1.i2 2.k2 the board is still very similar to the completely empty one since i2 and k2 dont really have a big impact on the structure. Your trmph-lines seem rather odd though. 4.i4 seems to be good after 3.f5 but 4.g4, 4.e4 and 4.h5 don’t make too much sense imo. As for the variation with 4.i4 I also disagree with 6.h5. Instead I would try 6.i6, 6.j7 or something like 6.c8Other options than 4.i4 could be 4.f9, 4.c8 or 4.i8

    A good response after 3.g6 4.h1 may be 5.f3 or 5.f7. For move 4 I would suggest 4.d9. I agree that 3.g6 benefits from i2 k2... it does resemble chickenlords chickenwing (b9 and f7 being a good combo) a little bit.

Return to forum

Reply to this topic




Include game board: [game;id:123456] or [game;id:123456;move:20] or [game;id:123456;move:20;title:some text]