Suggestion for improved algorithm

[Sparlock]

I have a working C# implementation of an algorithm which would significantly decrease your number of comparisons and increase your accuracy to 100% for the entirety of the top 10 list. I think that it might be adapted to fit this project.

Basically, you increase the number of Pokémon in the comparison to 4 and implent a modified merge sort which keeps track of the results of all the comparisons simultaneously.

[fio4ri]

Sorry, I had gotten pretty busy and ended up forgetting to write a response. If you are still around that is something I would be interested in hearing more about.

[Sparlock]

Hey, no need to be sorry. It's my birthday AND it's less than a week before Sun and Moon come out. Perfect timing to perfect your program. Haha.

Basically, you wrap each Pokemon in an object containing an empty linked list of Pokemon. After each comparison, you add all of the "losers" to the linked list of the winner. When you get to the final Pokemon, you take it's linked list of losers, and you use that list to start the next round of comparisons. Rinse and repeat until you have a list of 10 Pokemon.

As far as speed is concerned, you will get your number 1 Pokemon in a little less than 300 comparisons, and the first level linked list (because, keep in mind, the losers will ALSO have a linked list of all the Pokemon THEY beat, so that data structure looks a bit like a "binary" tree) will only have around 15 Pokemon you need to worry about drawing from for your next set of comparisons. In my testing with the XY Pokedex, it would run a little over 400 comparisons before getting the complete list.

I've attached the relevant code. I actually ended up naming the algorithm a "Cluster Sort" because, although I started working with a merge sort, the final result didn't really resemble it. lol

HumanSort.zip

[fio4ri]

Great thank you very much! Also happy belated birthday! I am a little bit busy right now, but I will take a look at it soon (December or thereabout).