I admire the simplicity of an elegant design idea, hopefully most games designers and developers do. The concept of a deck builder is extremely elegant. The act of building a deck is satisfying and includes action, engine and scoring in a single simple act. So it goes on the quite short list of ideas I wish I had been able to come up with.
However, there are a few gaps in this very elegant design. One of the biggest is that just as players get their deck built the game tends to end, which can be quite unsatisfying. This is usually compounded by the scoring of deck builders which is based on counting through the built up deck in slightly fiddly fashion. Sadly most games which seek to remedy these problems simply tack on a "something to do with the deck" second act which undermines the very elegance which makes the deck builder so perfect as a game design idea.
So my big idea was a deck un-builder where the game ends when a player tries to draw but cannot, even with a reshuffle. Once the game ends players score only the cards in their hand. However, for a long time the concept of handing players a deck pre-built and asking them to un-build it sat uncomfortably. Still, the idea of un-building a deck was quite clear and the different deck's themes wrote themselves. One deck would extend a player's hand, one would build a tableaux, one would return trashed cards, one would hand their cards to an opponent. This also slotted quite neatly into a vague fantasy setting; scholar, builder, grave digger and thief.
Still, at this point Moonflight was sitting on the shelf because its special idea was mechanically killing itself. Whilst working on a few other games the solution occurred, Moonflight has to be a deck builder and un-builder. Quite simply each card would play two ways, one way in the first half in game, then another way once the market place empties. In the second half of the game cards trash the carefully built deck down to the game end trigger and scoring.
This version of game play means play is all one thing, the same engine powering the building of the deck powers its destruction, allowing the elegance of the deck builder to be preserved. Since the decks are not pre-built the game has a much greater level of re-play value and freedom, also the inequality caused by deck learning is lessened.
The difficult part remained, the theory was that when the game flicked from state to state a deck's engine would persist, recognisable but altered. So all I had to do was write and balance a couple of hundred interrelated cards all at once. Easy.