Optimaliserings puzzels

[Bartdelft]

Beste leden,

Ik probeer optimaliserings puzzels op te lossen, maar ik eindig bijna altijd als laatste.
Ik wil verbeteren. Heb je een theorie voor mij?
Wie kan mij helpen?

Bedankt.

[CHalb]

Sorry, mein Niederländisch ist fast nur auf Rätselanleitungen beschränkt und ich kann deine Kernfrage nur raten.
Sorry, my Dutch ist almost reduced to puzzle instructions and I'm not sure about the core of your question.

Can you please ask in English? Oder du fragst auf Deutsch. Then you will propably get answers.
Or maybe some of our Dutch members react   .

[Richard]

Or maybe some of our Dutch members react   .

I will.
Bart originally asked the board of WCPN for tips to become a better solver in Optimizing puzzles.
Since there is very limited knowledge about optimizing puzzles and it is not likely that WCPN will start publishing such puzzles or organise a competition around that theme, we couldn't help him.
Instead, I advised him to ask for help here in the forum, because this place is visited by a large amount of players of all kinds.

So if someone has general tips in getting better in Optimizers: either write it here or send Bart an e-mail via the forum.

[glum_hippo]

It seems like a very broad question.

Die Frage ist nicht sonderlich präzise.

De vraag is moeilijk te beantwoorden omdat ze zo algemeen is. (Thanks Google translate!)

[Semax]

I think, in general you need to be thinking a lot about the puzzle before you even start to make your first solution. First things to do:
- Fully understand all rules of the puzzle. Including initial parameters, like given digits, or given sums, or the shape of the grid, or the properties of the word list, or ...
- Fully understand the measure which is to be optimized. Including smaller measures which are part of the final value, like maximum of sum of digits in a row, or maximum of product of x (x are the sums), when the product of the row sums needs to be optimized.
- For each rule, think about how it is limiting the measure (or parts of the measure) in what way.
- For each pair of rules, think about how they are limiting the measure (or parts of the measure) in what way.
- For each triple of rules, and so on...

Secondly, you could think about optimal solutions for partly rules and partly measures:
- Take a small rule and a small part of the measure, then try to find the optimal solution for them.
- Then take a different rule or a different measure, and do the same.
- Then take combinations of rules and/or bigger parts of the measure.
- Stop, when it takes too much time to do so.
- Find out, which combinations of rules and measures are easy to optimize and which ones are hard to optimize.

Third, start broadly:
- Create a few different solutions, each starting with nothing but the given puzzle. When creating those, make sure that at the beginning you do VERY different things in all of them. For example, start in a different area of the grid, start with different digits or words, start to optimize different part-measures first. For each of the solutions, while creating them, you should already try to optimize them. But try to optimize a different part-measure in each of them.
- When you have those base-solutions, try to optimize each of them a little bit.
- For all optimizations, always keep track of your total optimizing measure. You always want to know what small change gave you how much progress in the measure.
- Try to find out, by which optimizations you can gain the most measure improvements.

Fourth, optimize narrowly:
- Create a new solution by starting from scratch and using any ideas and insights you gained so far.
- Try to optimize it as far as possible.

Fifth, delete + optimize:
- Delete a small random part of your current optimal solution, use the remainders to create a new solution, hopefully improving the optimum.
- Try to optimize it as far as possible.
- Repeat deletion and optimization. Always keep track of the total measure.
- If you feel stuck (can't improve more), go back to any previous solution or create a completely new one from scratch. Then repeat deletion + optimization again.

[Bartdelft]

I think, in general you need to be thinking a lot about the puzzle before you even start to make your first solution. First things to do:
- Fully understand all rules of the puzzle. Including initial parameters, like given digits, or given sums, or the shape of the grid, or the properties of the word list, or ...
- Fully understand the measure which is to be optimized. Including smaller measures which are part of the final value, like maximum of sum of digits in a row, or maximum of product of x (x are the sums), when the product of the row sums needs to be optimized.
- For each rule, think about how it is limiting the measure (or parts of the measure) in what way.
- For each pair of rules, think about how they are limiting the measure (or parts of the measure) in what way.
- For each triple of rules, and so on...

Secondly, you could think about optimal solutions for partly rules and partly measures:
- Take a small rule and a small part of the measure, then try to find the optimal solution for them.
- Then take a different rule or a different measure, and do the same.
- Then take combinations of rules and/or bigger parts of the measure.
- Stop, when it takes too much time to do so.
- Find out, which combinations of rules and measures are easy to optimize and which ones are hard to optimize.

Third, start broadly:
- Create a few different solutions, each starting with nothing but the given puzzle. When creating those, make sure that at the beginning you do VERY different things in all of them. For example, start in a different area of the grid, start with different digits or words, start to optimize different part-measures first. For each of the solutions, while creating them, you should already try to optimize them. But try to optimize a different part-measure in each of them.
- When you have those base-solutions, try to optimize each of them a little bit.
- For all optimizations, always keep track of your total optimizing measure. You always want to know what small change gave you how much progress in the measure.
- Try to find out, by which optimizations you can gain the most measure improvements.

Fourth, optimize narrowly:
- Create a new solution by starting from scratch and using any ideas and insights you gained so far.
- Try to optimize it as far as possible.

Fifth, delete + optimize:
- Delete a small random part of your current optimal solution, use the remainders to create a new solution, hopefully improving the optimum.
- Try to optimize it as far as possible.
- Repeat deletion and optimization. Always keep track of the total measure.
- If you feel stuck (can't improve more), go back to any previous solution or create a completely new one from scratch. Then repeat deletion + optimization again.

Semax,
Vielen Dank fur dein detailliertes Antwort. Ich hatte ferien und war 1,5 woche weg, darum so spat.
Ich habe concludiert, auch fon eine Untersuching aufs intenet, daB man nicht einfach eine Mathematische losing machen kan um dan ein maximum zu errechnen.
DaB ist glucklicherweise nicht der fall, dan bleibt der SpaB auch behalten. Ich habe auch SpaB mit programmieren und habe versucht in GW-BASIC ein losings programm zu schreiben fur SUDOKUS, aber auch daB ist es nicht einfach.
GruBe, sorry und nochmals vielen Dank.  Bart.

[Semax]

Hi Bart,
I know, it's a bit late, but you might want to check out the following optimization contest, because those puzzles are really great
http://anuragthefirst.blogspot.com/2020/...ation.html

Senior already posted a link to it in this forum a week ago
http://forum.logic-masters.de/showthread.php?tid=1880
so I hope you might have seen it. But if not: in that part of the forum, people are posting links to interesting competitions, and from time to time also to optimization contests.