I would like to work toward a truly random hex-generation system that produces good results, but maybe that is too much, so lets start with extrapolation. Extrapolation can work almost as well as complete random generation. One can start with a rough map at a global scale using the Traveller style icosahedral globe with hexes. There is a nice random generator of those with customizable scales from like 7 hex per triangle up to 35 hex per. Anyhow, a random/automated extrapolation routine could basically turn a global map into a local map with only a few passes, and charting the entire globe could be done on a progressive and as needed basis from a very coarse high level map.
Before I get into my other thoughts, I want to address that I always use horizontal (latitudinally aligned) hex rows. E-W travel is always true to a hex row in this case, but N-S travel, especially when extrapolated to an icosahdral-projection globe, is only true to a longitudinal hex row when at or near the center of a given triangle of the of the icosahedron. Otherwise, you are actually tracking a path that curves off of true north or south by following a hex row. Horizontal rows won't fix that, but ceases to present a false north heading to confuse. Furthermore, the hex rows at the edges of the triangle, do still maintain a true north heading. Predominate travel patterns in open seas tend to be E-W also. Some early wargamers used the E-W convention that was picked up by Traveller, but it was discarded along the way because on skirmish maps, troop advancement (on the vertical) was easier to track, as well as frontal engagement... I find battle mat logic a poor justification for making a mess of global mapping... but enough of that digression.
To start, each global or macro hex has a terrain assigned. At least a majority if not more than half the sub-hexes should match the macro-hex they are being extrapolated from. Also, each macro-hex has six neighboring macro hexes with terrain assigned. Those neighbors correspond to a pie wedge of the given macro-hex, we should assume that some percentage of those pie-wedges will have terrain matching the neighboring macro-hex.
This gives us a couple of baselines to work on... the hex is probably best processed in triangles (pie-wedges), and a series of tables can be produced to represent those correspondences. At this point, I should probably specify the terrain types, but I am winging it, so lets assume around 6-10 major terrain types and hope everything falls into place later. These tables should be weighted in a way, either by dice mechanic or raw sampling, to statistically deliver the terrain indicated by the macro-hex at least 50% of the time. The remaining spread of results would be a litany of other terrain types sorted by expected prevalence.
SO: The intersections of Mountains and Ocean would probably contain a wide variety of terrains including the principles of course, along with just about everything in between. I imagine a terrain table between Mountains and Hills on the other hand would be predominately hills and mountains with some thin chances of a couple of other terrain types.Terrain variations between Ocean and Ocean on the other hand, would probably offer almost nothing but ocean, with a few land types scattered at the most extreme distribution of the table.
I am also considering additional weighting of the rolls/results based on already filled in adjacent sub-hexes. I can image a number of ways to make this process work in a virtual system, but in a manual tabular system where simplicity is king, this may be best represented as a 'wild-card' result on the table. More filled in adjacent sub-hexes should have a greater impact on the result. Perhaps, in a system using 3d6 (3-18) to produce results, each adjacent, filled-in hex adds 1 to the roll, and results of 18+ are a wild-card result based on those adjacencies...
SO: a sub-hex with no adjacencies would roll as normal, but for every adjacent hex filled in, the next die-toss is influenced slightly in that one's favor.
Obviously, Genesor* Fiat will win out if some die-results fail to satisfy the desire or expectations of the world creator. This is a process to aid and expedite the creative, not stifle it, after all.
Find the URL for the world generator, and knock one of those out for experimental purposes.Link included.
- Make (and print) a grid template with the wedge/triangles marked.
- Nail down the major terrain types to include (will probably need to have more granular terrain options for finer map scales).
- Decide on a dice mechanic and table format that can produce sensible results based on the adjacent sub-hex fill.
*I am coining and claiming the word Genesor as "one who creates (worlds)"