Topic: Frisian Balancing

Nordfriese wrote:

Oh. I just saw that the growth program has of course its own built-in remove=(chance) statement, so the additional remove=# statements in the bush programs are superfluous. Just remove them and the growth chances should go up drastically…

I have done this and it should be sufficient. In my first attempt I got the code in immovables.cc wrong. The immovables with low probability are not removed by the probability check but replaced with a new instance of themselves (i.e. they do not grow) so they need an extra cycle each time this happens.
with the deleted remove statements in their grow programs I believe we have sufficient chance to have a berry farm support one collector even on blackland. However I'd really love to have a special blackland berry bush that fits into the scenery.

Top

Quote

Nordfriese wrote:

However I'd really love to have a special blackland berry bush that fits into the scenery

Just add one with suitable values and placeholder graphics in the balancing branch in review, and I´ll add the graphics in the branch then

oh that is very kind of you. I will do so then.

Top

Quote

Pushed a set of graphics. How do you like them?

---

Attachment: image/png

Top

Quote

hessenfarmer wrote:

WorldSavior wrote:

Additional idea: What about letting the berry farmer plant always the best possible berry, at least for the place which he chooses?

According to the code that should be the case. However there is a crux in the code. for each node the 6 best bushes are identified and then based on a weighted randomness one is chosen.

So it's not always the best bush, but it can also be the 6th best

so with one bush at 70% other around 1% we have a 25% chance of planting the wrong bush.

Why?

Do they work different to trees? Somebody said that the best 6 trees are chosen and then weighted again by their growth possibility, the better, the more probable. In this example the chance to plant the wrong bush would not be 50%, but 0.01/0.71.

Nordfriese wrote:

Pushed a set of graphics. How do you like them?

I like them, they fit well to blackland. They have some candle-look so to speak, and this fits well to the blackland trees which have some lamp-look sometimes.

Wanted to save the world, then I got widetracked

Top

Quote

WorldSavior wrote:

hessenfarmer wrote:

WorldSavior wrote:

Additional idea: What about letting the berry farmer plant always the best possible berry, at least for the place which he chooses?

According to the code that should be the case. However there is a crux in the code. for each node the 6 best bushes are identified and then based on a weighted randomness one is chosen.

So it's not always the best bush, but it can also be the 6th best

indeed. Maybe this Code could use some optimization.

so with one bush at 70% other around 1% we have a 25% chance of planting the wrong bush.

Why?

Ok here is how it works:
1. first we check all immovables which have the attribute we want (e.g. tree or bush) for their growprobability and have the 6 best suited with their related probability in a sorted table of the length 6 or lower if less immovables exist with this attribute.
2. afterwards we sum up all probabilities of these candidates. (e.g. as in my example above 70+1+1+1+1+1=75)
3. we create a random number in range 0 to sum of probabilites.
4. we start with the first (highest probable) entry and subtract the probability from our random number. if the result is below zero we plant this immovable. Else we take the result as new value to subtract from for the next candidate.

this rresults in a chance of 5 / 75 in our example which is 6%. so it is not as high as I first I expected (got something wrong from the code) but still high. if you calculate this with 70+10+10+10+10+10 you'll end up with a 50/120 = 41 % probability that a 10% immovable is planted instead of a 70% immovable.

Do they work different to trees? Somebody said that the best 6 trees are chosen and then weighted again by their growth possibility, the better, the more probable. In this example the chance to plant the wrong bush would not be 50%, but 0.01/0.71.

it works the same for all immovables, so yes the better growthprobability the more often the immovable is chosen. but the algorithm evaluates basically the probability of 1 immovable against the sum of 5 other immovables which is not perfect I believe.
maybe we should square the probability of each immovable to increase chance of immovables with higher probability.

Top

Quote