V31 50K Seed Analytics Compiled by BigDunka: https://bit.ly/2sWa4CQ
Corrections compiled by members of the Go Mode Podcast community on the Go Mode Podcast Discord.
Write-up by coolpapabell2282.
Date: Jan. 29, 2020
There seems to be a discrepancy in the data here. You can sort of see this by the fact that, say, Ice Rod cave and Mini-Moldorm cave have the same percentage, even though they’re totally equivalent logically, and Mini-Moldorm Cave has 5 items instead of 1.
So since Dunka shared the spoiler logs from their seeds, I re-did the analysis. I pulled out just the playthrough portion of the log, and got rid of the mentions of collecting crystals, big keys, etc. Then I did a count of how many times each location was visited by the playthrough to collect an item. That data is summarized here:
This data does rely on the playthrough, which definitely has its own bugs and problems, particularly when it comes to swords and bottles, since there are multiples of them in the pool. But to summarize, Skull Woods has fallen way down the list, and generally, locations with lots of items that felt like they should be higher actually are much higher. This makes sense, Skull Woods should drop down – Thieves’ Town has more items and opens up at the same time, it should be more likely to have progression. Espeon and others have noted that it probably should be even a little lower, but the fill algorithm slightly favors dungeons with lots of keys.
The thought is that this is because it makes a list of all locations, and places the dungeon items first, so dungeon locations are under-represented near the front of the list, especially dungeons with a lot of keys to place. Then, the list is reversed before required items are placed, to get away from the overworld-heavy part of the list. Otherwise the filler would heavily prioritize overworld locations (which was the case in some earlier versions of the randomizer). So, relatively speaking, this makes key-heavy dungeons a little over-represented in the part of the list used to actually place items.
How big this effect is is frankly hard to say because the algorithm is generally complex. Which I guess is what makes the game fun. 🙂