Tag Archives: Strategy Guide

Permute: Move by Move

Some time ago I introduced you all to my game, Permute — an abstract strategy game inspired by the Rubik’s Cube.  Since then, the game was implemented on Christian Freeling’s MindSports — an essential site for any fan of the genre — and in Stephen Tavener’s wonderful Ai Ai software.  Hopefully some of you may have tried the game since then, and have discovered that the game has a unique feel in play and supports some unusual strategies and tactics.

Since developing the idea I’ve played a great deal of Permute, mostly against Ai Ai’s stiff opposition, but also some games against human opponents.   In this post I’ll walk you through a game I played against Simon R on MindSports, and explain some basic tactical and strategic considerations to keep in mind while playing Permute.  This sample game contains a few instructive moments that should be enlightening for any new player.

Here’s an animation of the full game, courtesy of Ai Ai:

permute12-game2-simonR-final

Full game — Simon was Orange, I was Yellow. Final score — Orange 16, Yellow 43.

The game is also viewable on MindSports, if you want to step through it move by move yourself.

Permute Move Notation

In order to record games of Permute and replicate them later, we need some form of notation to keep track of what moves have been played.  Ed van Zon of MindSports came up with a simple notation system that is very easy to learn:

Permute 12x12 -- notation sample-01

This small sample board illustrates how the notation works.  In this case, Yellow has made a move, rotating a 2×2 face clockwise.  We can translate this move into notation easily:

  1. We start by noting the direction of the twist, with a capital C for a clockwise twist, or A for anti-clockwise.
  2. Then we note the square on the board that is in the bottom-left corner of the 2×2 face we twisted, in this case d3 (indicated by the green square in the diagram).  So our move so far is Cd3
  3. Next we add in the square marking the top-right of the twisted face, in this case e4 (the red square in the diagram).  Our move is now Cd3e4
  4. Finally, we add a hyphen for readability, then add the square we chose to bandage, which is d4 in this case (the blue square in the diagram).  So our complete move can be written as Cd3e4-d4.

That’s it!  If we note down every move in this way, in proper sequence, we can easily reconstruct a game later for analysis.  In Ai Ai the notation works basically the same, except moves are written with the word Clockwise for clockwise twists and Widdershins for anti-clockwise twists.

Preliminaries

This post is very long, so I will not recap the rules here; if you are totally new to the game, please read my previous introductory post and follow-up about the Ai Ai implementation so that you can learn how the game plays, and get yourself set up to play it.

This particular game was played on a 12×12 board, which is the standard size available at MindSports.  A board that size contains 144 pieces, which seems like a large amount, but each move in Permute directly moves a set of four pieces and also locks down a portion of the board afterward, so games on 12×12 are surprisingly fast and brutal.  Typically a full game will last between 20-25 moves per player (note I’m counting a twist + bandage as a single ‘move’).

The board size has a strong impact on play in Permute, as it does in other territory games like Go.  Playing Permute on 12×12 is similar to playing Go on a 9×9 board — fighting starts immediately, and tactics are a major focus of the game.  Controlling the centre of the board early is very important, as this offers great opportunities both for expanding your own groups and blocking your opponent’s growth.

For a deeper, more strategic battle, I recommend playing on a 16×16 board.  At this size, games will last between 40-50 moves per player, about the same length as a typical game of Chess.  Opening play on this board size will tend to focus more around the corners and edges of the board, with the centre becoming more important once territories have been settled around the sides.  If you want the territorial aspect of the game to really dominate, then you can play on a 20×20 or even 24×24 board; on 24×24 a game will last about as long as a game of 19×19 Go (generally somewhat shorter).  Ai Ai supports Permute play on all these sizes.

The Ai Ai implementation also allows numerous variants of the game to be played.  The most notable of these is the option for a 1[1]2[2]* move protocol.  This means the first player on their first turn twists and bandages once, and from then on players alternate twisting two faces, then bandaging a stone of their colour in the second face they twisted.  I highly recommend trying this variant, which allows some very clever manoeuvring, and also lengthens the game on smaller boards.  For serious strategic play however I still advocate 16×16/20×20 single-twist rules as the best option; the two-twist variant makes useful tactics substantially harder to see, because two twists can cause so many changes to the board state in a single turn.

In any case, 12×12 with single twists is the ideal introduction to the game, and that will be our focus in this article.  In a future article, I will cover strategic tips and techniques for play on the larger boards, aimed at advanced players.

Permute 12×12 — Simon R vs Eric S

The opening begins!  Simon was playing Orange, meaning he gets to move first.  He strikes out for the centre immediately, which is a sensible play on the 12×12 board.  After just that single move, both of us now have two 3-groups each — each move in Permute substantially changes the local group composition, particularly in the early game, where few pieces have been bandaged.

I respond with an anti-clockwise twist at e4f5, aiming to contain any ideas of Orange extending to the western or southern edges.  In the process I connect my two 3-groups into a single 6-group.

Orange responds by growing their own 3-group by twisting f8g9.  Expansion is never a bad thing, but this move is a bit passive; the resulting group is still smaller than my largest, and this move does not really hamper my expansion.  In response, I twist h4i5, connecting two more pieces to the 6-group by rotating a piece into the connecting corner on h5.  From there, I can extend by adding to the piece attached to the north of h5 — I like to call these one-square extensions from bandaged friendly pieces anchors, as they are good bases for expansion — or I can connect two more pieces on the eastern edge of the group in two different ways, thanks to the 3-group covering i3-i4-j4.

Orange spots the threat of more pieces connecting to my group, so he instead connects his own pieces beneath and around my group by twisting h3i4.  This is a good move, in that it serves two purposes — it instantly gives him a nice-sized group, and it actively blocks my group from expanding to the south.  I was prepared for this, though, thanks to the anchor on h6, so I connect three more pieces with a twist at g7h8.  Note that my group is quite secure; each extension has bandaged pieces one space apart, which cannot be separated by any opposing twists.  When building your groups, one-space extensions are the safest way to connect new pieces.

We’re very early in the game, but the centre is already taking shape.  Here Orange decides to secure some more pieces for their southern group by twisting e3f4.  Again this move is a bit passive; Orange has the opportunity to sandwich my centre group between their upper and lower groups, and my hopes of extension rest on me being able to protect the corner of my group at e5, or the northern extension at g8-h8.  Better would have been Ae5d6-e5 — that would have severed my corner at e5, blocking expansion to the west, while further securing Orange’s biggest group.

Having been spared a squeeze on e5 from Orange, I twist c4d5 and bandage d5 to secure that vulnerable corner and gain space on the west side, with ambitions of connecting to the western edge and cutting off the bottom of the board, so that Orange cannot connect their upper and lower groups on that side.

Orange takes action to prevent my group heading west, but in actuality this move only delays me very briefly.  My group is also still expandable to the east and north, thanks to the anchors on h8 and i7.  This move does give Orange a large group in a commanding position over the southern reaches of the board, but at this stage of the game, exerting influence and securing areas for future growth is more important.  A nice move here would have been Cg8h9-h8, which would have blocked off any northern expansion for my group.

I respond by moving forward with my plan to connect to the western edge, which restricts Orange’s growth opportunities and gives me an anchor on b4.  At this point I also notice Orange’s vulnerable corner on i8, which is an important weakness now that Orange has invested several moves in growing this southern group.

We will take a close look at Orange’s next move, which is a crucial moment in the game.  Orange elected to twist i5j6, extending their group closer to the eastern edge while interfering with my group’s potential expansion on that side.  Unfortunately this move does not protect the vulnerable corner at i4, which gives me a huge attacking opportunity.

The middle diagram shows one potential alternative — securing the corner at i5 increases Orange’s hold on the bottom of the board, but it also further cramps their space for extension, confining their largest group mostly to the bottom.  A better alternative would have been the block at h8 I recommended previously — this move is still effective now, forcing a response from Yellow and adding some tactical complications.  However, at this point Ai Ai estimates Orange’s winning chances at just 36% even if they had chosen to block at h8.

Seeing that Orange has not protected the corner at i5, I launch my attack and twist out Orange’s piece on i5 (marked with an X in the diagram).  In one stroke this slices Orange’s largest group in half.  Bandaging at j5 ensures Orange cannot easily repair this damage, and claims further space toward the lower-right corner, making it very hard for Orange to even attempt to connect around this new blockage at i5.

At this point, Ai Ai estimates my winning chances at 88%.  Orange had invested a lot of moves in this group, and now the bottom portion has been entirely cut off from the section on the right, and is unlikely to form the basis of a winning group.  I have control over the centre and plenty of expansion options on the northern and eastern edges of my largest group.

Orange can see they are in trouble, so they respond with a twist at i8j9, in an attempt to constrict my expansion options to the north.  Orange now has to defend perfectly to have a hope of victory; when forced to play moves like this which do not extend one’s own groups, the opponent is often free to build up their advantage further.

This exchange gives us perhaps the most important message to take away from this game: protect your corners!  In Permute you will often need to bend your group around to connect pieces in various directions, and any corner piece in your group left unbandaged is a potential weakness.  One twist can potentially slice your group to ribbons and leave you struggling.

Now that Orange is on the defensive, I decide to aim for connection to the northern edge.  When ahead and in command of a large group, a useful strategic sub-goal is to try to split the board.  The group doing the splitting is then unassailable, assuming you have bandaged it well, and your opponent is heavily constricted in any attempts to connect large chunks of pieces.

Orange responds by constraining my group’s northern expansion, squeezing it from the west.  This does secure the northwestern section of the board for Orange, which could allow them to build a decent-sized group, and their current group stretching down to f6 is currently pretty secure.  However, this move still gives me the chance to connect to the north or east sides.  Ai Ai recommends moves that either disrupt the unbandaged, unsecured Yellow pieces at i6-i8, or blocking Yellow’s anchor at h10 and preventing the northern connection attempt.  In either case, though, Ai Ai pegs Orange’s winning chances now at only about 3.5%.

As it happens, I elected not to complete the northern connection, and instead disrupt Orange’s newest largest group.  My Yellow piece at f9 makes connection for Orange’s group quite awkward, so my hope is that further incursion into Orange’s northwestern territory will further decrease their chances to build a competitive group in that area.

Orange hits back by disrupting my 3-group around h11, ensuring that my bandaged piece at h9 remains isolated for the time being.  However, this leaves me with an opening to cut off three Orange pieces from potential connection via a later twist at f11g12.

At this point I can see that my central group is very strong, so I decide to continue my plan to disrupt Orange’s opportunities for expansion while securing my position.  By extending down from my anchor at b4, I further constrict Orange in the south, while adding a couple more pieces to my central group.

Orange elects to start enlarging their group in the northwest, establishing a bandaged piece at e7 that also prevents further extension of the Yellow group from e6.

I estimate at this stage that Orange’s northwestern group could not grow large enough to challenge my lead, but just in case I decide to extend my group and bandage at j8, in the process securing the vulnerable piece at i8.  Orange fights back by twisting my pieces at j9-j10 into the corner and firmly blocking them off from my group.  This also builds a nice group connecting through to the northwestern section, but those Orange pieces at f11-g11 remain vulnerable.

My first impulse here was to cut off Orange’s group on the top with Cf11g12-g12, but then I spot a move that looks bigger.  The clockwise twist at c2d3 grows my own largest group by three while simultaneously blocking out three Orange pieces permanently, which seems like a favourable exchange.

Orange responds sensibly, forgetting about the unfortunate southern group for now and further building on their central group.  Orange knows that at some point I will be forced to respond if I want to restrict Orange’s growth there.

Now that Orange’s southern group is well contained, I believe that their only chance to win is to connect their central and northern groups, then extend far down the eastern edge to pick up sufficient additional pieces.  I decide to try to connect to the east to completely cut off any ideas of expansion down that edge for Orange.

Orange sees an opportunity to place a secure bandaged corner at k7, extending their northern group a bit further.

Orange was able to extend that northern group a bit, but I could immediately follow up with Ck5l6-k6, completing my goal of connecting to the eastern edge.  Now Orange has no way to connect to the bottom groups, and the bottom groups themselves are cut off from each other thanks to the bandaged piece at j3.

Orange follows up with a prudent move, securing their vulnerable corner at i10 with a bandaged piece at j10.  Unfortunately, at this point the game is fully out of each; Ai Ai sees no future for Orange, giving them a winning percentage of a flat 0%.  There is no way for Orange to make a larger group at this stage.

My goal at this point is to firmly block off any remaining threats Orange has to increase their score substantially.  Having cut off the bottom with my last two moves, now I close the door along the top as well with Cf11g12-g11, splitting the top group and ensuring it cannot connect to Orange’s central group.

Orange has seen the writing on the wall by this time, but is valiantly fighting to the end.  He extends the southern group with two more pieces, and generates a possible threat of winding around the blockage at j3, though this would require several additional moves to achieve.

I can see that Orange may be hoping to connect around j3, so I close that door as well by placing another bandaged piece at k2.  Now there is no way through on the bottom.

Orange now pursues the last viable option for building a good-sized group, which is to extend their central group as much as possible.  I have almost no presence in the northwestern part of the board, so Orange has some chances to build up a score here potentially.

Finally I respond to the threats in the northwest, and place a bandaged piece at d11, which seems a suitably annoying placement.  The idea is to blockade the 4-group extending from f10 from the Orange pieces in the top-left corner, reducing Orange’s potential for growth.  Orange is still able to grow their central group by connecting two more pieces at b10, but those four pieces around f10 are now isolated.  At best Orange could attach the piece at d10, turning the 4-group into a 5-group, but there’s no way to connect those pieces to the larger group below.

Here I get greedy, and try to connect my central group across the 2nd row along the bottom.  The plan is to construct a pair of pieces that I can then bridge between the bandaged pieces at c2 and f2, but I actually made a mistake.  By choosing the twist at e2f3, I actually ensured that I would be unable to connect my newly-formed pair.

Orange makes precisely the right reply here, placing the Yellow pair where I want it, but unbandaged.  Now I have no chance to connect across the bottom, because in order to secure those pieces I would need to bandage one of them, which means a twist must be made.  But, given that the pair is already in the right position, any twist I make will only move them out of position!

This sort of problem comes up fairly often in Permute, and I was annoyed at myself for not seeing it.  When you have set up your pieces properly, you should have them placed so that they are ready to be twisted into place, but crucially, are not already in place and unbandaged.  That means you can make a move at that location, twist the pieces into place and secure a bandaged piece in the right spot.  If you try a setup for connecting a few pieces but misjudge your plan, as I did, then your pieces may be prematurely placed, and your opponent is then able to prevent your plan at any time by twisting your pieces out of position and locking them down with a bandage.

In truth, this move sequence was overly optimistic anyway; given the two-space gap between c2 and f2 and my lack of well-placed pieces nearby, the only way my attempt to connect would work is if my opponent chose to allow me to make that final twist of the free pair.  I had no way to force that connection in the first place, and Orange had no intention of allowing it, so effectively that was a wasted move.  In a tight endgame, throwing away a move like that could cost me the game!

I cannot secure the connected pair between c2 and f2, so instead I opt to insert another bandage at g2, trapping another Orange piece on the bottom edge at the same time.  Orange then extends that bottom edge group further, but this also connects my pieces to the bottom-right corner group.

I still cannot secure the extension in the south, so instead I opt to harass Orange’s central group some more by blocking two pieces in on the western edge.  This has no effect on the size of my Yellow group, so it is a safe move that still causes some trouble for my opponent.  Orange extends the bottom edge group once more, after which I lock out two more Orange pieces from their central group, and at this point Orange resigns.

If we remove the bandaged pieces we can get a clear picture of the final position:

Permute 12x12 -- move41 -- count-01

There we have it — a final score of Orange 16 to Yellow 43.  Had we played to the bitter end, my final score would have been significantly lower of course, as Orange could disrupt my connecting pair at d2-e2 at any time.  But even then the score would still be convincingly in Yellow’s favour.

Permutation Principles

As you can see, the 12×12 game is intense right from the start, and opening moves are very consequential.  AI testing shows though that between strong players, the outcome is usually decided about 60-70% through the game (during these tests I did not allow the AI to resign).  So, while the opening is very important, most games will be decided during the middlegame.  Our goal in the early stretches of a 12×12 game should be to come out of the opening with a stable central position, with some viable options for future extensions of our largest groups; during the middlegame we want to extend our groups effectively and safely, while denying opportunities for our opponent to attack our groups or build their own.

Of course, if we want to reach that point in our games, we need some guiding principles to help us along in our efforts to become stronger players.  Here I will outline some basic principles of Permute play that will help you play more effectively.

Make multipurpose moves

This general principle applies to many games, of course, but in Permute every move is inherently multifaceted and directly affects both players.  Every twist we make moves opposing pieces as well as ours, since faces consisting of only one colour cannot be twisted.  This means that every single move has the potential to grow or shrink our opponent’s groups as well as our own.

When I am choosing a move in Permute, I tend to evaluate it on a few dimensions to see how it measures up against other candidate moves, and I aim for the moves that can accomplish more than one of these general objectives at once:

  • How much does it add to my groups?  All else being equal, a move that grows my groups more is of course preferable to one that grows them less.
  • How much does it disrupt enemy groups?  If I have a choice of two moves, both of which grow my group by 3 pieces, but one shrinks an enemy group by two while the other doesn’t, then the first move is a better choice.
  • Does this move gain space?  Gaining useful space on the board and restricting the opponent’s growth options is really important, especially in the early game.  If a move grows my groups less than another move, but allows me to connect a group to an edge and block enemy expansion for the longer term, that move may well be the more sound strategic choice.
  • Does this move have follow-up options?  A useful beginner’s metric to judge this is whether a move leaves you anchors — single pieces extending from a bandaged piece that can serve as a launchpad for further expansion.  Again this is more of a strategic consideration; sometimes a move that adds fewer pieces can still be stronger, if it produces anchors that allow for a bigger extension later.  Anchors also create threats for your opponent, as they must keep an eye on those anchors and consider which may need to be blocked, which in turn may disrupt their own expansion plans.

Like any complex strategy game, all of these guidelines will have exceptions, and there are many other ways we might judge the effectiveness of a move as we get stronger.  However, I believe that keeping these principles in mind is very valuable as you are learning the game, because each of them focusses on the longer-term consequences of our moves.

In Permute we have to fight against the impulse to play reactively, and start twisting purely in response to local threats from your opponent; each move in Permute can be so destructive that we often will feel the need to react to every attack, leaving ourselves with no chances to attack in return or to build a coherent plan for expanding our groups.  If we instead train ourselves to evaluate each move in terms of its impact on both the groups local to that move and the wider strategic situation, by considering space gains and anchors, then we are more likely to choose moves that not only protect our groups in the short term but also create longer-term challenges for the opponent.

Protect Your Corners!

This is a straightforward tip, but definitely worth emphasising!  In this game we saw the consequences of leaving a vulnerable corner on your largest group — the opponent gets an opportunity to slice the group in two, and if the group is already contained and there is no alternate path to reconstruct the connection, then that one move is enough to ensure your group cannot be saved.  However, we may not want to develop a habit of taking time to protect every corner in every group; instead, we must judge when a group has enough prospects for future expansion that we need to secure its position.

Stronger players may experiment here and there with group sacrifices, where they lure the opponent to spend moves on disrupting a vulnerable corner when their actual growth ambitions lie elsewhere on the board.  However given the high price for getting this kind of sacrifice wrong, this is a tactic that should only be used very carefully!  Generally speaking, if you spot a weak corner in one of your important groups, you should act immediately to protect it.

Bandaging is just as important as twisting

Each move we make in Permute is, to some extent, irreversible.  In Chess, if we move a Knight to the wrong place, we can potentially reverse that move in the future to recover our position, but if we make a bad twist in Permute, the bandaging ensures that at least some of the damage is permanent.  Because of this, we need to be sure that when we bandage a piece as part of an effort to grow a group, we protect as many of our pieces as possible:

Permute 12x12 -- bandaging examples-01

Here we can see some basic examples of this principle.  In the leftmost diagram, Yellow has created a secure L-shaped group — the bandage on d4 protects the corner, and ensures the pieces on d2 and c4 are fully protected as well.  In the middle diagram, Yellow’s misplaced bandage on d5 has protected none of the pieces in the middle of this group!  On the right, this T-junction formation is also safe, giving Yellow three potential avenues of expansion with the bandages on b4, d3 and d5.  Our objective when setting up our groups should be to maximise our reach, while always being careful not to leave weak points that can be twisted out of position.  One-space extensions, T-junctions, and secure corners are important ingredients in any effort to build a large group.

Conversely, if we are attacking the opponent, we should place our bandages so as to create maximum disruption.  Let’s go back to that fateful 12th move in the game, where I was able to split Orange’s group via the vulnerable corner, and zoom in a bit:

Permute 12x12 -- attacking examples-01

On the left is the move I made.  When twisting at c2d3, I placed my bandage at d2, a diagonal step away from the diagonal corner I had just disrupted.  This ensures that not only is the vulnerable corner at c3 permanently cut, but also the Orange piece at d3 is blocked in, preventing Orange from twisting around the obstruction to continue building the group.  The Orange pair at c1-c2 is still twistable, but cannot get around the bandage at d2.

In the middle diagram, we see the position if I had bandaged at c3 instead — note that Orange now has two pairs at c1-c2 and d3-e3 that can potentially be brought to bear, expanding the group around my obstructing piece.  On the right we see a potential result of this mistake: Orange twists Ad2e3-d2, completely isolating my piece at c3 and building the group southwards with a secure connection.

The lesson here is to always observe the local patterns of bandaged pieces when attacking an enemy group, and think not just about the immediate damage you can do, but also whether you can potentially block off any more pieces by setting up your bandages differently.  Make sure your attack has as much lasting impact as possible, and then you can make advances elsewhere while your opponent scrambles to repair what damage they can!

Split the board when ahead

At several points in the game above, you saw me aim to split the board into sections by connecting my main group across the whole board.  When you have a decent-sized lead and the opportunity, connecting securely can be very beneficial; by cutting across the board, you prevent the opponent making board-spanning connections of their own, and restrict their options for catching up to your lead.

Let us go back to the game for a moment and look more closely at my middlegame connection to the eastern edge of the board:

On the left is the move I played in the game, which connected my largest group to the eastern edge.  Orange is now firmly cut off from connecting their upper group to the bottom half of the board, and their attempt to move south has been contained.

On the right, we see what could have happened if I had not attempted to reach the edge, but instead had played elsewhere (in this case, bolstering my group in the bottom-right with Ch2i3-h3).  Orange is then able to generate threats by connecting significantly more pieces down that right side.

In this particular position Orange cannot make massive gains this way; on the right we see that after a couple more moves, Yellow is able to contain the advance once again.  But Orange did manage to gain two more pieces even in this constrained position, and in a close game two pieces may make all the difference.

So, when the opportunity to cut off your opponent presents itself, taking advantage of that opportunity is often a good way to consolidate your advantage and deny your opponent chances to create more threats.

Twists are local, scores are global

Finally, an important reminder — Permute is a game full of complex tactics, but the ultimate goal of the game is inherently global in nature.  When fighting for your groups to stay live, never forget that all moves should be in service of the larger goal: building the biggest group of pieces on the board.  If you get deep into a tactical battle and realise you will not be able to make any more profitable exchanges, do not be afraid to abandon your group and seek a bigger move elsewhere!

Similarly, we can easily get tied down defending only our largest group, and forget that the group scores in Permute are cascading — this means that if the two players tie for the largest group, then we compare the second-largest, then the third-largest if those are tied, and so on.  So if a game is close, remember to nurture a secondary group in addition to your largest one.  Tight games between equally-skilled opponents will most certainly have games go to the second-largest group, and occasionally the third-largest (though this is rare).

Remember too that playing strategically and planning your group-building effectively will also help you tactically; the more secure your groups are, the better-placed you are to defend them, and the more you constrict your opponent’s growth options with well-planned extensions, the more vulnerable they are to attack.

Next moves

At this point, hopefully you have a clearer picture of how a game of Permute flows, and you have picked up some useful tools in your toolbox for your next twisty battle.  Permute feels quite odd at first if you are used to something like Go or Chess, given the fact that the board starts full and every move directly affects both players’ pieces.  But once we spend some time getting acquainted with the properties of the game, these unusual aspects will start to feel natural, and you can focus on increasing your playing strength.

From here, I recommend having a bunch of games on the 12×12 board, and focussing on developing plans for each of your groups as you play.  Always keep an eye on the score, and check that each move is advancing your own groups while hindering your opponent’s.  Once you become confident navigating the sharp openings of the 12×12 board, then you can try moving up to 16×16, where you have an even more dizzying array of opening options, and the game opens out to become even more strategically interesting.

I have a large backlog of posts to work on at the moment, but at some point down the line I will come back with some more Permute tips, shown off through a fully-analysed 16×16 game.  If any of you out there are interested in a game on the larger board, let me know in the comments; I am always happy to find new opponents!

Some of you may have noticed too that Ai Ai lets you play Permute with four colours, which makes the game feel even more like a twisty puzzle.  I am still in the process of getting acquainted with the nuances of the four-colour game — which works very well with the 1[1]2[2]* move protocol — but I will cover this at some point as well.

In the meantime, I hope you will give Permute a try — Permute is my game so I naturally will have some bias towards it, but at least I can say that after many, many games of Permute so far, I have yet to get bored with it.  I hope some of you out there will have as much fun with it as I have!

Until next time, here’s a little preview of what’s to come — a tense game of 16×16 Permute that came down to a single point:

permute16-5s-tight1

Tagged , ,

A Beginner’s Guide to Hex

Some time ago I talked a bit about Hex on this blog, discussing its history and how it kickstarted the entire connection game genre. Since then, a few readers have asked for a bit more detail on how to actually play Hex. So, for this post I have teamed up with Matthew Seymour, author of the brilliant Hex: A Strategy Guide, and we have put together this beginner’s guide to Hex strategy and tactics.

Below we will introduce you to basic tactics, templates, openings, and strategic considerations. Each section is very brief but will give you enough to get you started on improving your Hex play; after you tackle each section, please continue your studies with Matthew’s guide, which has much more detail and numerous examples of each critical concept.

Basic Tactics

The edges

Edge play in Hex is obviously critical, since in order to win we must connect our two edges across the entire board. Initially, connecting a chain of stones successfully to the edge seems like a baffling enterprise — an adept opponent can bend and twist your attempted connections away from the goal, and it can be difficult to understand how to set yourself up for a strong connection.

Edge templates are extremely useful tools for understanding edge play. In our previous discussion of Hex, we met the bridge, a configuration of two stones that proves to be virtually connected even though the stones aren’t adjacent.  Similarly, edge templates show us configurations of stones that, with correct play, are guaranteed to result in connect to the edge. Templates indicate patterns where, even if your opponent has the first move, you will always be able to connect successfully. Your opponent may intrude on your template, but so long as you defend your template at every step, you will succeed.

This means however that templates are always just a move away from breaking. If you have two overlapping templates, and your opponent plays a move that intrudes on both, you can only potentially save one of them.

Here are some examples of basic edge templates:

Our first edge template is the very basic A-2 template, consisting of a stone on the second row. If White tries to block at A, Black can respond at B (second diagram), and vice versa. Either way, White has no means of stopping black. This template is very similar to the bridge template.

One of the most useful templates is template A-3, consisting of a single stone on the third row. This template comes up very frequently. We’ll analyse the situation by considering Black’s possible threats. On one hand, he could play at A and connect to the edge with template A-2. Or, he could play at B, and connect by a chain of a bridge and template A-2. The important part here is that these threats don’t overlap. If White plays at A, or one of the two hexes below it, Black simply plays at B and connects by that route. And if White plays in any of the 5 hexes on the right, Black plays at A and connects by that route.

We can succinctly convey this information in what’s called a pivot diagram. The two small dots indicate Black’s alternative moves. We can see that the rightmost dot connects back to the top stone with a bridge. Both dots are connected to the edge by one of two possible moves (template A-2).

Lastly, let’s look at Template A-4, on the fourth row:

diagram5-01

Similar to A-3, Black has two threats: to play adjacent at A, connecting to the edge with A-3 (left), or to bridge to the side at B, also connecting with A-3 (right).

These alternatives can get around most possible blocks by White. But there’s one issue, the point C. Both the threats above overlap at this point. So Black needs a response to a block at this point. The solution is shown below (note that if White played (3) on the left side, Black could bridge to the right instead).

diagram8-01

There are a large number of edge templates. You can find an excellent catalogue here or here.

Interior templates: connecting chains

We’ve already encountered one interior template — a template for connections in the centre of the board, away from the edges: the bridge. As you might expect, other templates exist for more complex configurations of interior stones. Knowing these templates is very helpful indeed, as once you achieve such a configuration on the board and recognise that template, you can play elsewhere, knowing that you’ll be able to successfully defend that template against later intrusion.

Some well-known interior templates are the Wheel, the Crescent, the Span, and the Trapezoid.

Defending at a distance

We’ve talked a lot so far about connecting — positive, attacking play. But what to do if our opponent has us on the back foot, and a deadly connection is looming? How can we stop them?

First, we should remember that as satisfying as that positive, attacking play may be, Hex is what is called in combinatorial game theory a hot game. This means that it is always beneficial to make a move in Hex, and no stone of ours on the board is ever a detriment to us. On top of that, one player must always win in Hex, so if we successfully block the opponent from any possible connection between their edges, we’ve in fact won the game. So don’t neglect defensive play — in Hex, it’s precisely as powerful as attacking play, and will win you the game just as effectively! Defensive moves in Hex are also offensive moves.

Having said that, defence in Hex can be a delicate affair. As we’ve seen in numerous examples, stones can be connected even if they aren’t adjacent, and skilled players can move across the board at high speed, staying connected the whole way. Therefore, if we attempt to block simply with adjacent blocks — playing right next to opposing stones — a skilled opponent can easily bend around us. Likewise, if we block at a distance but misjudge the situation, the opponent may still get round us by using bridges to cover enough ground to do an end-run around our defences.

In practice we may need to combine adjacent blocks with more distant blocks in many cases. The adjacent blocks restrict the opponent’s options for bridging forward, while the distant blocks contain those advances:

The classic block short-circuits the opponent from afar, allowing the defender to respond whether the opponent advances directly forward or takes a more oblique approach:

In either case, in order to defend we need to evaluate the opponent’s options for onward connection, and place our stones in anticipation of those options. If we defend reactively, and follow the opponent around right at their heels, then we’ve no hope of survival. If we instead constrict their choices and contain their subsequent advances, then we may just keep them at bay.

Ladders

Often, when approaching an edge of the board, you’ll end up in a situation like this:

ladder1-alt-01

White wants to connect and Black can’t allow it, so Black blocks at (3). White can’t connect right away but can force Black to carry on blocking all along the edge. This series of back-and-forth adjacent plays is called a ladder. White is completely in control here; Black has to respond to every ladder stone White plays, otherwise White’s connection is assured.

Go players will be familiar with sente — the concept of maintaining the initiative, by making moves that force an immediate response from the opponent. When you have sente you are in control of the game; you are making profitable moves, and all your opponent can do is match you, stone for stone, unable to direct play to their advantage. Sente is just as important in Hex as it is in Go, and ladders are one common manifestation of it.

However, you have probably noticed that if White continues playing the ladder here, it’s Black that ends up connecting across the board:

ladder1-01

To make the ladder profitable for them, White needs to incorporate some additional tactical plays. If White had an additional stone in place to form a ladder escape, then when the ladder reaches that stone, they can connect to the edge with ease.

ladder1-escape-01

Of course, when a ladder is already forming, taking a turn to place a ladder escape stone simply dooms the ladder. So players will often place ladder escape stones during the opening phase of the game, to allow for profitable ladder play later.

Another option is to place a stone that is both a ladder escape and a threat to connect by another route. Your opponent will be forced to block either the threat or the ladder, and then you can connect by the other means. In the diagram below, Black plays a stone at 1. This can escape the ladder, but it also threatens to connect via A (with two bridges). White can’t block both approaches, so Black will connect. This is called a ladder escape fork.

diagram14-01

If you have no ladder escapes or forks available, you’ll have no choice but to “break the ladder”, as Black does with move 7 in the diagram below (note that 7 is connected with Template A-3). In the acute corner, this tends to reverse the roles: notice how now it’s White who’s the attacking player with a ladder.

diagram13-01

Strategic Considerations

Openings

The opening in Hex is an interesting moment, as the first player needs to consider not just what is the best move to play, but what is the best move to play that won’t get swapped. In Go or Chess, you can play your opening move without fear of that move suddenly becoming your opponent’s opening, but not so with Hex!

As the second player, you need to do the opposite calculation: has my opponent played a move that, with perfect play, would give them a winning advantage? If so, I should swap; if not, I can safely play on.

On smaller boards, Hex has been solved, meaning that we know the precise outcome of any given opening move. That leads to diagrams like these:

Above are three diagrams showing the ultimate winner, with perfect play, of opening moves played at every cell on the board. There’s an important trend to notice here — the winning openings for any given board size are not straightforwardly extendable to larger boards! While we can see a general theme that opening moves in the centre are stronger than those on the edges, the specific outcomes of those edge cells change as we change the board. That means that on the boards we humans play Hex on — from 11×11 upward — not only do we not have these convenient maps of what moves win or lose, but we cannot use the opening maps from smaller boards as a definitive indication of the outcome of any opening on the bigger ones.

Matthew’s guide focusses on the 13×13 game, and for openings on that board, he’s produced a swap map that can help guide you in the opening. The cells with black dots are Matthew recommendations for good opening moves for Black. When you are the second player, if your opponent opens anywhere in the shaded area, you should swap — those moves have the potential to give a winning advantage, so you’re better off taking that stone for yourself. If your opponent plays outside that shaded zone, let them carry on — you can possibly do better by playing your own opening.

swap-map-13x13-01

For larger board sizes, like 15×15 and 19×19, we don’t yet have enough games played at a high level to put together a reasonable swap map. However, we can make some reasonable inferences about good opening moves; in particular, opening in the obtuse corners seems a good way to go on all board sizes.

These are very simple principles, but should be enough to get you started. One thing to bear in mind is that we humans are far from perfect play, even on 11×11, so both sides are likely to make mistakes, not just in the opening but throughout the game. So our goal at this stage should be simply to ensure that our opening doesn’t obviously disadvantage us; we don’t need to fret too much about whether a particular move is 100% winning or losing.

Playing in the corners

The Hex board has two types of corners — acute and obtuse — that have different properties. Corners are the only parts of the board where your stones can both strengthen your own position and weaken your opponent’s, and for that reason, players tend to play stones in the corners early in the game. Typically you will want to play stones in at least one corner on each of your edges during the opening.

The corners being so important often leads to pitched battles to establish control over them, and so strong players may study corner patterns (think joseki in Go) to navigate these tactical scuffles. If you don’t have a presence in a corner and your opponent does, invading is useful in order to reduce their influence there, and corner patterns will help you to reduce that influence. Conversely, if your opponent invades your corner, you can use these patterns to settle the fight and maintain as much of your initial influence as possible. The challenge in these situations is judging when there is no more profit to be gained, and thus when it’s time to move on from the corner battle and establish yourself elsewhere.

There is a lot to discover in these corner patterns, but don’t worry too much about these early in your Hex journey; as you start to face stronger opposition and find your corner play is letting you down, refer to Matthew’s guide for detailed examples of how to fight for the corners.

Influence

We’ve alluded to this concept in the previous section, so now let’s expand on what influence means in Hex strategy. Stones in Hex are not just localised points — they have impact on the board around them and on other nearby stones. Every stone has the potential to connect to something or to block something else, and when placing our stones we need to consider the influence of the stones around our planned placement.

In the early stages of a Hex game, gaining influence is important. We would do well to place our stones around the board, to spread them out; this maximises the potential influence of each stone. Conversely, if we don’t spread our stones out, we may have a strong influence in a particular area but will be weak elsewhere. If we are struggling to find an effective place to play, we can look at our relative influence on different areas of the board; if we find some areas where we have low influence, those might be good places to play our next moves.

As you might expect, stones in the corners have a high degree of influence — the proximity to two edges means those stones are better able to restrict your opponent’s activity in that area and force them to work around you in a limited space. Placements in the corners also are tougher to block, and provide you with ladder escape stones for later in the game.

The edges are somewhat less intuitive. We might feel secure playing near the centre of our own edges, as this seems a useful way to block the opponent, but in practice these kinds of placements do not provide strong influence. Instead, we should play near our opponent’s edges — this forces them to work around you and makes it harder for them to connect.

Beginning Go players often play in a style referred to as Puppy Go, where they continually play very close to every one of their opponent’s moves, following them around the board like an excitable puppy. We can easily be tempted to play Puppy Hex in a very similar way. Unfortunately this is an adorable, but poor strategy; in a Puppy Hex scenario your opponent is dictating play completely, and since you are always one stone behind they will have free choice of where to establish influence and you will always be playing catch-up. Always keep an eye on the broader board situation, and try to take the initiative when the situation allows it — don’t let your opponent drag you around by the nose!

As you become more comfortable playing in an influence-oriented style, you can start to focus on making moves that serve multiple purposes. Gaining influence is good, but gaining influence and blocking the opponent is even better! This is a challenging step, requiring you to have both tactical and strategic vision, but as you gain more experience and become able to recognise common tactical motifs, you’ll be better able to keep these in mind as you seek to expand your presence across the board as well.

As a final note, we should remember that Hex is fundamentally a scalable game — we can play Hex on any size board we like without changing the rules, but the feel of play will change. Hex on larger boards is a challenging and rewarding affair, but specific tips on those epic battles is beyond the scope of this article. However, we encourage you to try larger boards, as they by necessity will make you play in an influence-oriented style. With so much additional empty space on the board, you’ll need to learn to anticipate where battles for influence and territory will rage, long before they actually happen. That experience can help you on the smaller boards too, training you to think globally more consistently.

Territory

Territory is a critical concept to understand in Hex strategy. Think of territory as the potential your stones create for future connection; the more territory you control, the more tactical options you have for later attempts to form connections between your stones.

As a starting point, we might say that each stone creates territory in the area immediately around itself; in other words, the empty hexes immediately adjacent to it. However, as we see below, this definition falls apart fairly quickly:

useless-stones-01

These intrusions by Black gain no useful territory. In both cases, White simply blocks any onward connections, so the ‘territory’ gained (the shaded cells) offers nothing that Black didn’t already have!

If we believe that stones create territory regardless of their disposition, then we will run into situations like the above, where our stone is effectively a wasted move, as it will never actually be able to connect to anything. Instead we should restrict the definition a bit more: the territory around our stones consists of the adjacent hexes that could in theory participate in a connection. If we want to invade somewhere and gain influence from that play, we need to be certain that the placement provides useful territory; if the stone does not gain territory, then we have simply placed a stone for no real purpose. Without territory we cannot claim influence, as the enemy can simply work around us at no real cost.

Taking the initiative

As in many other abstract games, in Hex gaining the initiative is of huge importance. Recall the Puppy Hex discussion earlier — imagine if we could force the opponent to play Puppy Hex. If we can place stones with aplomb while our opponent can do nothing but respond, we can dictate the pace of play and dominate the board at our leisure.

Here we will go in-depth into some Go terms we mentioned earlier: sente and gote. In Go, when we play a stone that forces the opponent to respond — because a group is threatened with capture, for example — we say that is sente, meaning we are gaining the initiative. Our next move after the sente move is essentially free; the opponent’s response is mandatory, so our next placement can be anywhere we like, and we can use that to gain influence or territory. Conversely, the forced response the sente move creates is gote — we are forced to be the puppy for that move and play where the opponent demands.

In Hex we also have sente and gote moves. For example, we may recognise that our opponent has an edge template in play, so we may choose to intrude on that template and gain some influence. That move is sente because it demands a response; the opponent must play to save the template, otherwise that connection is lost. At that moment our opponent’s move is gote, lending us the initiative.

As we gain more experience of Hex strategy, we will be better able to identify opportunities to gain sente. At the same time, we must be mindful of our opponent’s threats, and remember that playing gote moves to save a critical connection is vital too! We should try to avoid being the Hex puppy whenever possible, but sometimes there’s no escaping it.

Tenuki

Let’s look at another situation:

tenuki1-01

Here Black is threatening to cut the stone A off from the top-right edge, and the straightforward response would be for us to save the connection and take gote, such as by responding at L3. After all, by not playing there we lose the connection.

However, in this situation we can see that White has an opportunity to make an intrusion of their own, on the other end of the board. Black’s threat depends on using the stone B to connect to the bottom-right edge of the board. White’s board situation will allow them to make other connections, even if they sacrifice the connection under attack by Black, but Black’s situation is just as fragile. In cases like this we may elect to tenuki — to play away from the threat and allow our opponent to break the connection. Instead of defending against the threat we attack elsewhere, and now they must make a choice: either save their own template, or finish ours off. If they finish ours off, they must make a second move, giving us influence elsewhere; if they take gote to save their own connection, then we have regained the initiative.

In this game, White elected to attack Black’s B stone with move (2), rather than save the connection of A to the edge. Black elected to save the connection, playing out a standard joseki sequence, leaving White with the initiative.

tenuki2-01

Tenuki is an advanced concept, and often difficult to judge. In general, you will have more opportunities for playing away from threats in the early- and middlegame, when the board is less full and there will be opportunities for other connections. In the late game, typically both players will have committed many stones to particular connections, and there is inherently less flexibility; if we ignore a threat, we are more likely to hand the win to the opponent.

The Joy of Hex

We’ve covered a lot of ground in this post — over the course of these few sections we’ve gone from the basics of the board geometry through to advanced strategic play. Yet for all that, we’ve barely scratched the surface. From here, you can move on to Matthew’s detailed guide to Hex, and dig deeper into all of these concepts. While you’re there, be sure to try out his fantastic collection of 500 Hex puzzles (also available in PDF, in Hex style and Go style) to sharpen your tactical vision.  If you need help with openings, he also used over 6,000 online games on 13×13 to generate a very useful opening database.

Having said that, resist the temptation to power through all this material. Take some time with these concepts, apply them to your games, and move on only when you feel comfortable and confident. Remember too that Hex is perhaps the most famous modern abstract strategy game, but it is still very new in the grand scheme of things. Traditional games like Go, Chess and Shogi have had centuries for strategies to be developed, whereas in Hex we are all still beginners in some sense! So there is always more to discover and more to learn.

If nothing else, we hope this brief introduction will give you an appreciation for Hex’s incredible depth and nuance. Hex is a disarmingly simple game, so much so that a brand-new player may be tempted to ask ‘…that’s it?!’ when told the rules for the first time. But within that sparse framework lies a world of intricate tactical and strategic variety. This simplicity means Hex also has amazing flexibility — we can play lightning-fast blitz games on 11×11 boards, strategic masterclasses on 19×19, or mind-bending, baffling escapades as long as a game of Go on 26×26. Each one of these configurations is rich with possibility. Learning Hex also benefits you in other connection games — the tactics you learn here can transfer to other games, like the Game of Y (more on that in our next post).

Above all, we hope you have fun with the game! Go spend some time testing out your strategies online, entering tournaments, analysing games and writing about them. But alongside that, teach your friends and family (when Covid restrictions allow!), help them learn some basic tactical and strategic concepts, and show them why you love it. Every new player we bring to the game makes Hex’s future ever brighter, so the more we help others to see what the game can offer, the more enjoyment we’ll all have in the years to come.

Next moves

In the second and final part of our Hex mini-series, we will analyse a complete game of Hex in detail, and show how the concepts we’ve introduced here play out in a game between strong players.

Then we will analyse a brief game of Y, as well, to demonstrate how Hex concepts transfer to other, related games — and we’ll point out how some concepts change when we move to a different game.

Extra nerd stuff

Check out these papers if you’d like to know more about how the small-board swap maps above were generated:

SOLVING 7×7 HEX: VIRTUAL CONNECTIONS. AND GAME-STATE REDUCTION. R. Hayward, Y. Bjomsson, M. Johanson, M. Kan, N. Po, J. van Rijswijck. Department of Computing Science, University of Alberta, Edmonton, Alberta, Canada.

Solving 8×8 Hex.  Henderson, Arneson and Hayward, IJCAI 2009.

9×9 Hex: Scalable Parallel Depth-First Proof Number Search.  Paulewicz and Hayward, Proc. Computers and Games CG2013, Springer LNCS 8427 (2014) 138-150.

Tagged , , , ,