Decide on good starting points

[MichaelClerx]

• If the sea is sloped, we could use random points
• We could use some randomised distance from e.g. notts maths
• We could use a list of starting points, e.g. http://www.math.uwaterloo.ca/tsp/pubs/data.html

[mirams]

I think (for a competition) I would just let people pick those and just give them the domain (a 1x1 square).

[MichaelClerx]

You mean a square that encompasses the full island?

[mirams]

Yeah, which you can rotate for different competition entrants perhaps

[MichaelClerx]

One procedure would be:

1. transform the four corner points of the grid
2. take min and max x and y coords of resulting points

This would give different search space to land mass ratios depending on the rotation, so maybe better to

1. rotate around center instead of (0, 0)
2. use fixed box around result, i.e. with sides that are 2*unrotated-height long
3. scale and translate island + box
4. return box bounds

[mirams]

I think I would just draw a big box, then translate/rotate UK so it stays within it but don't scale it (so that gradients remain actual gradients of hills).

[MichaelClerx]

Hmmmm... I was thinking to scale x,y,z with the same factor

[MichaelClerx]

I suppose it's hard enough without that

[mirams]

OK, I guess that is fairly arbitrary, in terms of whether you choose box bounds of 1 or 2 or what