Copyright (c) 2000 Joćo Neto

Gonnect is a two player strategy game played on a 13x13 Go board.

  • RULES - The rules of GO apply, except:
    • Players may not pass;
    • The PIE rule: White may choose as his 1st move to exchange places with Black.
  • GOAL - A player wins if he can create a chain of connected stones between any two opposite board sides (left-right or bottom-up) or if his adversary has no valid move.

Historical note: At July 12, 2000, Gonnect was proposed with an extra rule stating "Players cannot create uncapturable stone structures (i.e., all pieces can eventually be captured), unless it is a winning move", but was remarked by Eric Osman, Matti Siivola and Richard Rognlie that both rules forbidding suicide and passing would make this complex and inelegant rule, unnecessary.

  • Some initial notes:
    • Black cannot make a good 1st move, because White would use the PIE rule to exchange positions.
    • Perhaps, the best 1st move for White is to play at the board center.
    • Notice that the "No pass" rule, together with the "No suicide" rule, implies that there is no safe uncapturable two hole structures, since sooner or later a player needs to fill them in order to keep moving.
    • Too small boards (9x9 or less) would give too much advantage to Black. I would suggest a 13x13 board for some nice and quick games, or even a 19x19 standard Go board for deeper battles.
    • A suicidal move cannot be made even if it would result on a connecting winning line!
    • There is already a Gonnect engine in Richard's PbM game server! Try it!
    • An article about Gonnect appeared on Abstract Games Magazine 6.

Blocking the Opponent

Given an unresolved local battle between players A and B:- if the unresolved area is not central to A's connection, then A is safe to abandon pieces in that area and play around the periphery of the battle, often ending up in a superior position despite losing many pieces. - if A must connect through the unresolved area, then any captures by B in that area are bad (much worse than captures in Go) as not only is a potential connection for A removed, but B consolidates their blocking connection. Sounds obvious, but these qualities distinguish Gonnect substantially from Go.

Each player will have at least two structure types: connecting chains (to win) and blocking chains (to prevent the opponent's win). Also, there will be stones that can be used to block the opponent's connecting stones. A player may insert a stone into the opponent's structure in order to delay a connection. let's call it a defensive blocking move.


From left to right (and up to down) we have:

  • The triplet
  • The ladder (or strong Keima)
  • The open triangle
  • The arrow
  • The extended ladder

Loose diagonal connections, like Keima (Knight's extension) introduce delays for final orthogonal connections. Orthogonal connections might be even more valuable than at Go.

Connection Bestiary

From left ro right (and up to down) we have:

  • Strong and very slow
  • Equally strong and twice as faster
  • Almost as strong as the first and faster
  • Fast, but not that strong
  • Slow (but faster than the first) and strong
  • Strong and fast

These connections can mix with each other (eg, the fifth is a mixture of the first two). It depends on the adjacent piece context. It the connection is threaten, a strong connection may be needed. If the danger is not very close, a faster connection can be more useful.

Placing Stones

It seems that a good rule of thumb in Gonnect is not to play isolated pieces adjacent to enemy pieces. Playing an isolated piece adjacent to an enemy piece removes one of its freedoms and gives the opponent a head star in surrounding that piece.

Even for offensive attacks, White should play in E4 to capture the Black stone [this is a basic Go tactic]


Basic Patterns

[The X] From our game it looks like the following pattern is strong for either player. If White has connected above and below, he still needs to surround the entire formation and removes it to join the two. This situation suggests the opposite of Go philosophy: - in Go it is more important to build a strong periphery around a region (outwards in), - whereas here a player builds a strong center that reaches outwards.

[The ladder] This is a basic Go tactic. If Black plays e6, he threatens to capture d6, and White responds d5, and then: d4 e5 f5 e4 g4... and the White chain is dead when reaching and edge. If White as some piece near the diagonal line e5-j1, the ladder cannot work, and Black would get himself into a dangerous position.


[The triplet Jump]

Moves: e7 d7 e6 d6 e5 d5 e3. White tries to escape Black's opposition, making a two point jump. Point e4 is relatively safe. If Black e4 then f4 d4 d2 After that, if Black move f5, then: g5 f6 f7 g6 h6 (a Go ladder).

Breaking Connection

Black is heading to the bottom. How White should start blocking it?

  • [White plays at e5] Too soon. Black may play f5, then: b4 (if f4 then e3) f4 e3 (if f3 then e3) f3 and White cannot stop it.
  • [f5] After Black's e5, then e4 f4. Both e4 and f5 are weak pieces.
  • [e4] If Black responds f5, then f4, and the blocking is strong. If Black responds f4, then f3. This option seems sound.
  • [f4] If e4 then e3 f5 g4, and White creates a ladder structure. If d4 then d3 c3 and Black seems to escape. However, this move may be good near the edge.
  • [g4] This move makes sense if the chain cannot turn into an opposite direction. After e4 (if e3 then e4) White moves e3 f4 f3.
  • [e3] Extends option [e4]. If f4 then f3 g3 g2 h3...
  • [f3] Similar to [g4], but if White has some influence in the right, it's stronger because it gives more protection to the weaker bottom edge.
Connection Races

White must avoid connection races (each player forgets the other and tries to finish his own chain), at least until he gets the advantage. If a player knows that his connection time is smaller than the opponent's, he must insert delays by placing defensive blocking stones. [check blocking the opponent]

E.g., If Black moves B4, even if White plays at B3, White suffers a 2 move delay, since he must play A4 and then B4. This move is a defensive blocking move, a player defends himself by increasing the opponent's connection time (i.e.. the number of stones needed to win)


Consider the following situation where White has achieved a deadlock at A11-B10.

White is blocked from connecting vertically or horizontally. To win Black just has to wait until the board fills up and White is forced to play B13 or D13, reducing their top left formation to one eye and making it vulnerable. How Black can maximize their chances of forcing White to do this? Black must ensure they have more eyes than White when the eye-filling stage is reached, regardless of whether the eyes belong to formations that offer any connective benefit. 


What about the next pattern?

This appears to be the best way to pack as many eyes into as small a space as possible. Any game between two expert players will eventually be deadlocked, reach the eye-filling stage, then be won by the player with the most eyes. Of course, while you were doing this, your opponent would be playing on awkward places to interfere. And which would be more important - making 1-point eyes in your own territory or messing up his? We just don't know as yet!

[Bill Taylor] Whoever has to crack first loses; which means, roughly, whoever has the less territory as in Go. But not exactly the same - there is an art to filling in one's own territory, and wasting moves in the opponent territory, that regular Go does not encompass. It has been studied for endgame-like positions, and leads to some fascinating Conway-style results there. But no-one has yet studied things for filling in very large territories. There are many general proverbs waiting to be discovered! No doubt if Gonnect (or no-pass Go) ever catches on they will become more well-known.

Gonnect Stages

(1) Board-filling stage: first half of the game where players run for connection until the board is deadlocked 

(2) Eye-filling stage: second half of the game where each player is forced to fill in their eyes. The first player forced to reduce one of their group's eyes to a single point then loses that group, and the game returns briefly to the board-filling stage. This is the reason for the "no passing" rule.

We've found the following recipe successful: play for a deadlock, anticipating the eye-filling stage, while creating as many eyes as possible. The opponent usually makes a mistake (not defensive enough) allowing a premature connection to win the game.

Gonnect turns out not to be only a territory-grabbing game, with connection theme being only secondary in importance. The connection aspect is uppermost, with the territory grabbing component more a contingency plan for later. In fact, these complementary strategies combine beautifully to make Gonnect as  interesting as both Go and Hex!

Some opinions

About the depth of Gonnect compared to Go, I start to understand. There is the new dimension found in the richer endgame where we might need to waste time in filling territories, as you said. So there is also something of Tanbo in Gonnect, not only Go and Hex. And there is the constant threat that a connection may be achieved quite early. This is something decreasing the depth. We seem to see where we want to go and who is ahead before it would be possible in a Go game. But this may well be a beginner's illusion. The game is anyway more exciting than Go, where I'm lost with its early lack of tension. Claude Chaunier

A fascinating game - part race, part blockage, part territory! I think this may be one of your prize inventions. But as you say, it's amazing no-one ever thought of it before.  Bill Taylor

Although Go and Gonnect have even more in common than I at first though--in the end, it is the player who constructs the strongest groups who wins, regardless of how the win condition is defined. It is obvious from looking at our board that you and I are playing Gonnect and not Go, but I think the positions could be analyzed without even thinking about what game we are playing. That is, I am about 90% certain you are going to win, and if we suddenly switched to playing by the rules of Go, I would still be about 90% sure you are going to win. There is some elbow room here, but I bet this would apply to most Go or Gonnect matches. And it is easy to see the basis of my prediction by looking at the board: you have two strong groups, groups which I should probably consider nearly impossible to capture. I have none, and it is obvious that my pieces were only played in response to your groups. I tell my friend with whom I play Gonnect, but is much less interested in games than me, "Don't look at the board and see pieces; look at the board and see groups." What makes Gonnect great--which I'm sure you realize--is the way it forces the players to constantly care about the move immediately at hand. Go is too spread out; a strong player can take it easy for the opening moves, like the hare racing the tortoise, knowing that they will ultimately win in the end game. All Gonnect does--the simple twist of rules--is to raise the possibility that there may very well not be an end game. Carpe diem. Consider we poor human beings. For the most part, we feel little direction in our lives, often pulled in multiple directions at once. We envy the characters we see in drama--their (fictional) lives are directed, they have a goal, and the plot is the essence of fulfilling that goal, there is never a wasted scene, or else we call it bad drama. We live our lives as though they were Go games, but we secretly (sometimes not secretly) wish they were Gonnect games. I suggested a business to my friend: clients come to me, name a date in the future, and pay me a few thousand dollars. On that date, I murder them. From now until that date, they live with the absolute knowledge that they will die in X years (or less, by other causes); anything they want to do, they do, because the possibility of the "end game" has been removed; for them, there is only the present. That is what you've done to Go in my opinion, you've taken an enjoyable game and made it as enjoyable as it could be. - Chris (San Diego - CA)

Go and Gonnect

Gonnect is not just a variant of Go.  What happens is that Go and Gonnect both use a common set of rules, namely, those for capturing with no liberties, Ko, ..., but each game uses them in a different role, Go for territory grabbing, and Gonnect for connection patterns. We don't know which is the best game, but Gonnect is not a subset of Go, they intersect in some places, but differ in others. What Go players are using when they play Gonnect is the knowledge they have in those common rules, but there are other parts in Gonnect that experience Gonnect players will get and play consistly better.

The same can be said about FIDE Chess and most Chess Variants. Why good chess players behave well in CVs? It's not because CVs are a subset of Chess (which is not the case for certain) but use the concepts of space, tactics, opening experience, and so on... to play those new games that share many features with FIDE Chess.

Another difference between Go and Gonnect is that Go is rather vague during the opening stages. This is because players can safely space pieces about the board to claim territory without too much concern for exact placement. In contrast the placement of pieces in Gonnect is critical from the very first move (as in Hex!) and a weak move can be attacked directly from the second move.

A sample game

c3 f6, f3 g3, g4 h3, h4 j3, j4 l3, f4 l5, h6 j9, k7 l7, k8 m8, l9 m10, m9 n9, l10 m11, l11 m12, l12 m13, l13 l8, k9 j6, j7 

White resigns 1 - 0

A sample game

c2 d4, e2 g3, g3 h3, h2 j2, f3 g5, f5 g6, f6 j3, g8 g7, f7 h8, h12 g9, f12 f8, e8 e9, d12 l11, d9 d8, e7 c8, d10 l9, k10 l10, j8 k9, j9 l7, k4 j7

Black resigns 0 - 1

A sample game

b12 g7, g11 k10, k11 l11, j9 k8, k9 l9, j8 j6, k7 l8, l6 j7, k6 k4, l4 l3, k5 j5, k3 j4, m3 j11, h10 k12, l2 e5, e8 d6, c7 c6, c3 b7, b5 c8, c9 d8, d9 b9, b10 d3, d2 f3, f2 e2, e1 g2, c2 f1, h2 j2, j3 h3, e7 d7, f9 g9, g10 c10, b11 e11, f10 h9, j10 d10, e9 g12, f12 f13, h12 e12, g13 a10, a11 a9, f6 f11, g12 c12, g8 b13, d4 e4, d5 e6, f7 c11, e3 f4, f2 g1

Black resigns  0 -1 


Should black play at c6? (No!) why not?

Check the answer here!

Another challenge: What would be really nice for a Gonnect puzzle, would be a winning move that had two good-looking replies - but one which lost to a connection tesuji, and the other lost to a Go tesuji that prevented connections for either player. - Bill Taylor


We want to thank all those friends that are even now, playing and discussing with us about Gonnect. Without them, this page would never be done. Neto wants to send a special thanks to Claude Chaunier and Bill Taylor from all the profound and fruitful discussions that we had and continue to have in all these years.