Havannah strategy has evolved over the years. As I write this, winter solstice 2016, my rating at Little Golem is 1929 whereas the highest ranked player rates 2377. That should tell you enough about the game and my ability, or lack thereof, to comment on strategy. This is merely about basics.

A very basic property of Havannah is this: of all games that can end in a draw, it must have one of the smallest margins.
It's easy to construct a drawn position, but in thirty years of play, there has been only one recorded base-8 draw.
There are two immediate consequences:

  • The first¬†player to complete a ring, bridge or fork, is the winner, so white has a first move advantage, and it is not diminished by a drawing margin.
  • Since a draw is no option, a sound defensive strategy will eventually turn into an attack all by itself!

To balance the first move advantage Havannah uses the swap rule.
The second property was not immediately obvious in the first year of the game's existence, when it was extensively played at the University of Twente and its games club 'Fanatic'.
It took Roelof Moll, a local Chess player who had played only for a couple of months, to point it out. He started winning consistently by following his Chess instinct and taking the center. He didn't care for speed, he cared for safety.
His reasoning was that it doesn't matter how 'fast' a group threatens to connect, if it's dead or forced to go roundabouts.
Cutting the opponent's groups from above (that is: from the center) limiting their options to at most two sides and one corner, he proved that all our previous strategies were in dire need of reconsideration. From his contribution came the concepts of snake strategy, with the emphasis on speed, and spider strategy, with the emphasis on safety.
It gave rise to the Safety Speed Dilemma, illustrated here in a nutshell, but actually pervading Havannah in all strategical and tactical aspects.

Havannah board White 1 brings the intended connection one step closer, but isn't save: black can cut with 2.

White 3 is safe: barring a simultaneous ringthreat, black cannot cut. But white has not gained any tempo: he still needs two moves to connect.

From this example an inductive understanding may already emerge that safety and speed tend to follow different routes! This principle pervades Havannah from the very basic tactics shown, to the intuitive realms of opening strategy.

A frame
The main strategic goal in Havannah is is the establishment of a frame, a connection aiming at a ring, bridge or fork, that, though still incomplete, cannot be broken by the opponent. With safety taken care of, it gives rise to two simple strategic truths:

  • Attacking a frame pushes it right into victory! The only defense is: having a faster frame, or at least threatening a faster connection in the process of making one. This may seem quite obvious, but even experienced players do not always recognize a frame until they discover that their attempts to cut or block were in fact counterproductive and would better have been left undone. In short: only defend if it can be defended and only cut when it can be cut.
  • Balanced games, that is: games that are not decided by tactical oversight, will eventually take the character of a race. Assuming that both players eventually frame, both will engage in the difficult process of counting. The faster player will make it a race without bothering about the opponent other than to answer local tactical threats.

Sorry, you need a Java enabled browser to view the Havannah Player!
The minimum number of stones to make a ring is six, but how about a frame? Surprisingly it takes more stones to frame one! The three attempts shown here illustrate why. In the top right example, the order of moves matters.
This "easy to threaten, hard to frame" property of the ring, makes it more of a tactical than a strategical weapon. Ring threats are usually used with other objectives in mind than actually completing one, such as connecting in sente, cutting, or stopping a low running opponent's chain.

Special attention should always be given to the possibility of a double ring threat such as black would call off over himself by playing 10 and 11 at j14 and j17 in the above example. Between experienced player's, double threats may be cleverly implemented, but seldom in the expectation that the opponent will actually fail to notice.

Unlike sides, corners are mutually exclusive, so the options for making a bridge are limited to begin with. A frame requires that the two corners involved are occupied before the frame is established, lest the opponent might refute the plan by snatching away one corner. This means investing in moves whose efficiency is granted only in a narrow context, inviting the opponent cut and render them useless. Because of these properties, the bridge, like the ring, is mainly a tactical weapon. Cornerpoints, apart from their inherent danger as part of a bridge, play an important role in joseki (corner disputes) because they form a connection between two sides. Three stones can add up to two sides and a corner, and if a chain containing such a triplet comes out in the open, it may connect up successfully almost anywhere.

There's a major difference between the fork and the other two winning structures:

  • Although a fork requires at least twelve stones, twice the number for a ring, a frame¬†takes less than half this number.

Sorry, you need a Java enabled browser to view the Havannah Player!
The example shows the same 5-stone frame twice. In the basic tactics section is illustrated why it is a frame: black can cut on f4 but after e2 further cutting attempts will push white straight into victory.
Black therefore should leave white alone! Let's assume he has a frame elsewhere on the board, of approximately equal speed, so that the game turns into a race.

The question of course is: Just exactly how fast is white? The answer is: the number of moves needed to complete the fork, minus the number of black stones white can draw into the position with moves that
  1. add to the completion of the fork, while at the same time
  2. threatening a ring or bridge in less moves than the forkframe itself.

The sequence at the top shows one of the many scenarios to do it in 10 moves. Black sente moves are indicated by the move counter, but do not appear on the board: they're supposed to take place elsewhere without influencing the situation at hand.
White makes 16 moves, 6 of which require a local answer because they constitute ring threats. Of course this is not the final answer to the origial question: make white 7 a one point jump to k18, and white can run the 18-line in sente for quite a while.
This low route has a resource to do it in 9 moves, as shown at the bottom. White makes 19 moves, 10 of which require a local answer because they constitute a ring- or bridge threat.
This is what the race stage often about: the one extra tempo you can squeeze out of a position using tactical threats. Of course the bottom sequence may not be the final answer to the origial question either.