The original version of Trellis was among the connection games described by R. Wayne Schmittberger in his article, "Making Connections" (June 2000 Games). It took place on a 17x17 checkered grid (counting the points rather than the squares), which was divided into adjoined 9x9 quadrants. Five games took place at the same time --- one in each quadrant and one on the full board. The object of each game was to connect the top to the bottom (for Black) and the left to the right (for White). The player who won at least three of these five games was the winner. You played just one stone on your turn.
Shortly after the article was published, I tried applying the quadrant idea to the game Hex (i.e., Black tries to connect the upper right to the lower left for each quadrant as well as for the full board, and White tries to connect the upper left to the lower right for each quadrant as well as for the full board --- three or more out of five wins). Because of the tension between the two "acute" quadrants and the two "obtuse" quadrants, the strategy of this game was more interesting than that of Trellis, I thought. I also began experimenting with ways to spice up the tactics of Trellis' checkered grid. I came upon the idea of playing two stones at a prescribed distance apart. A three-point separation turned out to be the most appropriate.
So the original game of Trellis split into two games: square Trellis (15x15 checkered grid, no quadrants, two stones per turn) and quadrant Trellis (14x14 rhombus board divided into 7x7 quadrants, just one stone per turn). Both were described in my article, "New Connection Games --- Part One" (October 2001 Games). Since then I've chosen square Trellis to be the standard version and changed the name of quadrant Trellis to quadrant Hex, since it is really a Hex variant rather than its own game.
Important addendum: I've discovered that the essential connective structure of Trellis is that of the game Bridg-It. In both games directly adjacent same-color placements connect, and indirectly adjacent same-color placements connect in color-specific alternation. There are two differences between the games: 1)There's a forty-five degree difference in board orientation 2)In Trellis a move consists of two placements, exactly three units apart. Thus if Trellis is played with only one placement per turn, it is equivalent to "Diagonal Bridg-It." (Consider that stones and bridges are only visual symbols --- what is really placed on a board is a position. So if you rotate a Trellis board forty-five degrees and square it off, you have a Bridg-It board. And if you rotate a Bridg-It board forty-five degrees and square it off, you have a Trellis board.) Bridg-It is a solved game --- a directly-adjacent-point pairing strategy exists for the first player. This explains why upon continued play of Trellis' original version, I began to find its tactics predictable. It also explains why the "exactly-three-points-apart" rule works so well: Because three is an odd number, the directly-adjacent-point pairing strategy is very quickly foiled. (I had also experimented with two-point and four-point separations, which I now see did not work for the reason that even numbers do nothing to prevent directly-adjacent-point pairing.) It was never intended as such, but I suppose Trellis can be viewed as "redeemed Bridg-It."