SketchUp version: 16.0.19912 64-bit
Skatter version : 0.3.3
OS: Windows 10
Description of the issue: Se bellow
Are you able to reproduce the issue every time? (If yes, how?) : Yes. It seems to be related to the order they are picked

Objects won´t spread as expected using Probability.

Three objects spread in grid pattern on an 10x10m square all on default 100% probability.

1:st pick – Group 1 = 29 items
2:nd pick – Group 2 = 17 items
3:rd pick – Group 3 = 43 items

1:st pick – Group 3 = 29 items
2:nd pick – Group 2 = 17 items
3:rd pick – Group 1 = 43 items

With another set of items (in this case components) on a new square I get another distribution (36-23-31) but still not a third of each as expected.

The problem is the samet for curves and individually picked points.

It’s a probability, it’s not exactly.
If you increase the number of scattered items, these numbers will be closer and closer to a third. (As you would expect with probability in a random system)

If you roll a dice 12 times, you wouldn’t expect to have 2one, 2two, 2*three, etc… You might get 12 times the number five, for instance. The probability is low, but it can happen.

OK! I thought I meant the (exact) distribution among the included objects. For me a function for controlling the mixture is definitely on the wish list. Without that function I can´t use the tool as I had planned n.b. a planting area with a mixture of 5% of species 1, 20% of species 2, 25% of species 3 and 50% of species 4.

Is it possible?

I’ll have to think about how to implement this, but I’ll definitly add it to the todo list :wink:

I would agree with Peter on this. We do need a more “Accurate” distribution control per scattered item. But Peter. let me ask you this… besides the % of the distribution, would you expect the location of same item being scattered to be in the same area? do you need them clustered?

In real life you either have :

  • a "wild" area, where nature decides what goes where and then the current probability solution fits well.
  • The other scenario is a "planned" area, where a human such as yourself decides what goes where and then things tend to be more clustered and organised.

In such a “planned” scenario using a Color Coded texture map could be a better approach to decide what goes where (with various fuzzy controls to get some noise in the borders, some plants mixing an such).

This kind of solution could obviously be coded and Thomas can comment on that better :wink:

Go Skatter!

Clusters have been on the todo list almost since the beginning :wink: it’ll come

Distribution Maps for the different parameters (distribution, scale, rotation, etc.) would be awesome and give a whole new level of control.

Then one could for example create a field with long grass blowing in the wind. A greyscale map would control the z-rotation and you could paint in wind vortexes.

The big question: Will you figure out an ingenious way to paint directly inside SketchUp (like the current paint tool, but with gradients and soft brushes…)? That would be an absolute killer feature (fit for version 2.0)!

The cluster function would be fantastic in many of my other project but in this case (planting and growing a city forest) I am more interested in controlling the exact mixture of planted trees. They are planted as really small plants in a 1x1m grid in a random mix.

Another function on the wish list is a “diforest/thin” out function and this I might want to control in different way. For now I first double the grid size (which I quite likely to be the case for the first thin out), then I use a block containing lines only (dosn´t render) and add that to the mixture (this is where I first noticed that Probability dosn´t work the way I hoped for). If I could control the percentage I would get a fairly good second thin out. But as the trees grow older I need better control of which trees are taken out and which are kept.

You are doing an amazing job with Skatter! Now I am off for Christmas break a day or two…

Marry Christmas to all of you!

Hi Peter, please try the latest version, the probabilities should be much more accurate :

It won’t be 100% because of floating point numbers, but very close. Let me know if you still don’t get the results you expect.