It's not dead; it's resting!
This run has been on hiatus for a little while, but there has been some progress as well. I found a more efficient path for directing the people in Kasandora, but in the process got burnout from doing too many calculations and test runs. Then somehow July flew by too quickly. Now it is time for me to get back on this.
I did all those test runs to see if I can manipulate the construction periods to be faster. So far all of the test runs' constructions are basically the same number of frames not counting random lag. All but 2 of the runs put roads on the same squares, but each have a slightly different path during the "direct" that is only a few frames slower than the best route. Even though there are more routes that I could test, I think I can conclude that they will be the same as the runs that I have done because they would have the same roads build before the construction.
I'm surprised that there were no differences in all those test runs since I encountered variations in my tests for the first construction cycle. In order to get 6 structures built and put the pointer west of the temple where I want it, I had to take a different route that was 4 frames longer. Had I not done that, only 4 structures would have been built. I tested all the patterns getting the 3 squares around the temple during cycle 1, and the patterns that had less than 6 structures built ended the pointer on the temple or the square west of the temple. I think part of the reason is that no field can be built on the square west of the temple, but it doesn't explain why it works in some patterns and not others. This sort of thing also affects the end of the route. If I build over the final lair as soon as possible and leave the pointer there, it reduces the number of structures built during Kasandora's final population growth (which occurs while building Marahna) and the population ends up as 486 instead of 510. By waiting a few construction cycles before building over the lair, the problem is solved.
I think the reason this occurs is that the town is scanned from top to bottom, left to right, and it will not decide to build a house unless there are enough remaining resources. If it is unable to build a field when needed, it won't build any more houses until it can build a field. If there are lots of fields built on the top of the map, then it will take too long to build the necessary field and it won't build as many structures. This is yet another reason to look at the game's code so that this can be optimized better.
While I was away on vacation and unable to TAS, I was playing around with a game where I am trying to get the maximum score (99990) in all the action sequences and the maximum population in the sims without cheat codes. It is going to be difficult to get the maximum score in Aitos Act 1, so I decided to do Northwall before and see how far I could get in the simulation before the sheep's fleece is needed. It turns out that by keeping the population low, the sheep's fleece is not needed! This only saves about 25-30 frames since we have to get the source of magic in Aitos, but it is a timesaver nevertheless.
So I am finally back to work on this and I completed cycle 2 now. Here is a link to the WIP. http://dehacked.2y.net/microstorage.php/info/102606997/ActRaiser_%28U%29_WIP_kas_cycle_2_done.smv