The World vs. Twixtbot analysis page 3 TWIXT PP

25 replies. Last post: 2020-02-02

Reply to this topic Return to forum

The World vs. Twixtbot analysis page 3
  • MisterCat at 2020-01-22

    Here is the direct link to the new analysis page:

    The World vs. Twixtbot page 3

    The fake game stub #2142725  contains no moves.

    To see the ACTUAL game, visit:

    Game # 2140447


  • MisterCat at 2020-01-26

    I won’t start a new thread for this.  There is a 4th (and last) page for analysis of this game.  The game should be over soon.

    Direct link: The World vs. Twixtbot analysis

    the ACTUAL game remains as above: Game # 2140447


  • MisterCat at 2020-01-28

    The team of human beings has triumphed over the machine!  I will say more, later.



  • Alan Hensel at 2020-01-28

    Someday, robots will rule the world. BUT NOT TODAY!

    Way to go, Team!

  • David J Bush ★ at 2020-01-28

    The bot made a couple of unusual moves in the opening. We defeated it by making super solid moves, nothing flashy.

    I hope a stronger version will be made available. How about spending 12 hours per move?

  • TwixtBot at 2020-01-28

    There’s a chance something stupid like running out of memory might happen, but we can certainly give it a try if people want.  It’s easy to change the “50,000” to “7,000,000” and see what happens.

  • David J Bush ★ at 2020-01-29

    Well maybe you could start a different account and call it SlowTwixt? Not everyone agrees with me. Thanks for your trouble. I hope you will still be able to use your computer!

  • technolion at 2020-01-29


  • passenger ★ at 2020-01-29

    Amazing job guys!

  • Florian Jamain ★ at 2020-01-29

    GG (-;

  • CosimoC at 2020-01-29

    Well done, boys! I feel (and I can only write “feel”) that TwixtBot is weaker in the opening. When he finds a tactical weakness, there is no hope for anyone. As a one cent advice, it can be tried to modify the 50,000 figure to, for example, 500,000 only for the first moves, using some more time for the opening. For the remanining part of the game, 50,000 is enough.

  • Alan Hensel at 2020-01-29

    By the numbers:

    28 days, 2 hours, 21 minutes.

    37 pegs.

    0 links in the first 16 pegs (43% of the game).

    20 links in the 21 pegs that followed, 14 by humans and 6 by the bot.

    20 decisions made by humans, starting with the decision of whether to let the bot go first.


    10 participants.

    Participation by percentage of comments:

    22.7% nie_wiesz (Maciej)

    21.7% MisterCat

    19.6% twixter (David Bush)

    18.0% spd_iv (Michal)

      9.0% Florian Jamain

      5.8% Alan Hensel

      2.7% bob440

    <0.5% BonyJordan

    <0.5% DrJochum

    <0.5% ypercube

  • MisterCat at 2020-01-29

    Thanks for those stats, Alan.  Interesting, but ‘percentage of comments’ might be modified to ‘percentage of GOOD comments’; then, I’ll drop right to the cellar!  LOL


  • Florian Jamain ★ at 2020-01-30

    100% agreed with David, we did not play anything “flashy” just very solid moves in the opening, it is enough to take the advantage again the Bot, in the very first moves he is not playing really well so we took our time there to secure an advantage and then just no blunder and it was a win. 

  • Carroll ★ at 2020-01-30

    Congrats @humans, @TwitBot, could you please provide figures, statistics or logs about the thinking process and when did the bot know that it was behind?

  • Maciej Celuch at 2020-01-31

    Good job guys! We’ve proved that we are more than just a numbers (at least more than “50000”) ;-)

  • DrJochum at 2020-02-01

    This experiment was great fun!
    Thank you to all who participated and especially to MisterCat for initiating it!!

  • P_Guz ★ at 2020-02-01

    Even I have not participated actively I really enjoyed just looking at this game. As David wrote, the strategy was very really solid, nevertheless some moves were far from obvious, not cliche ones. Building the left side was very impressive for me. In general humans controlled the game. After every human’s move there was a big question: did the bot find out some vulnerability or did he surprise playing some unexpected move. But he did not.

    Congrats for the win and for the style!

  • MisterCat at 2020-02-01

    Hey all – have a look-see at the official game link over in commentator!  BonyJordan has provided some very interesting data – revealing what Twixtbot thought was the best line of play at each move, and also it’s numerical analysis of the position.  At a glance, it seems that it ALWAYS thought it was behind (negative evaluations throughout), but when the evaluation got below -.85 (as explained by Jordan), it finally resigned.

    Were we REALLY winning after only one move?  Or two?  This, to me, is hard to understand, although that opinion WAS given by some of The Team – especially by Florian, who insisted that Twixtbot plays poorly in the opening, so that the key is to take advantage of weak opening play and right at the start, and then follow through with powerful moves to maintain and increase that advantage – never letting go.

    You can have a look here:

    the ACTUAL game remains as above: Game # 2140447


  • MisterCat at 2020-02-01

    DrJochum – thanks for the kind nod.

    David J Bush – it appears that you got your wish; see the  new player ‘TwixtbotSLOW’.  If this is the 7,000,000 version, may Heaven help us all!

    Twixtbot vs. Twixtbotslow


  • TwixtBot at 2020-02-02

    It’s only 2 million ; I run out of memory at 7 million.

  • Tasmanian Devil at 2020-02-02

    I’m curious about 29. g12, after which black resigns. Does it help o12 at all? Admittedly I haven’t spent much time studying the position.

  • MisterCat at 2020-02-02

    TD, hole O12  is already taken.  You meant something else.  I added a possible winning line for TBSlow at Twixt Commentator here:

    Note that TBSlow has already defeated TB in their first encounter; I expect more results of this type.


  • Tasmanian Devil at 2020-02-02

    No, I mean does it help me peg at o12, I.e.,does it serve a double purpose, in view of the fact that black promptly resigned? If l9 etc works without the exchange g12-f11, it looks redundant. 

  • Tasmanian Devil at 2020-02-02

    Does it help *the peg

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]