This third example was quite a bit better. The main difference here is that the map takes up the entire real estate - with the intention that HP bars, buttons etc would all be overlaid.
The other main difference is that this one was generated with individual images and letting the browser do the layering. So the gravel is one layer, the grass is another, the bushes are another, the trees are another... The gates and well are put in independently. The building... well, that's a bit more complicated...
The building was not clickable when I put it below the treeline because there was a transparent image above it consuming the clicks. To get this to work, I created two divs in the same place with different z-indexes. The bottom one had the img. The top one had a href tag with the same size of
transparent image. That allowed it to APPEAR that you were clicking on the building when in actuality you were clicking on nothing. Clicking on the building pops up a fb:dialog.
Also, in the top left there is a link. If you click that link repeatedly, it will toggle the grid on and off. That is done using the facebook clicktotoggle event.
Overall, the biggest issue with this one is that the map wasn't zoomed out as far as in the previous test. I think we need to zoom out a bit in order to appear to have more real estate per screen refresh.