Hello,

I am no expert at this game, or speedrunning in general, but I did a little digging into the assets of the game and made a few interesting discoveries, and thought I'd share them in the hopes that it could help somebody else.

First of all, here's a spreadsheet I extracted from the game code which contains all the dialogue transcripts: UPDATE: I have added a column to the end which includes all the search terms that will pull up the word. Please see the note in a post below on why words with apostrophes (can't, haven't) don't appear. https://drive.google.com/file/d/1hTXypS_gDJKskTqzQRvBxNoA613-HQEc/view?usp=sharing

But the most interesting thing I found was the code that checks the victory requirements. (For anyone interested I found this code in Assembly-UnityScript.dll, NewUIDatabase)

So basically the victory requirements are:

1. Must have seen 109 or more clips
2. Must have a plot score of 5

The plot score is a hidden counter that ensures that the player has seen clips that contain 5 pieces of key information:

1. Florence
2. Sister/Twin
3. Attic
4. Baby
5. Murder

If you look in the spreadsheet, it has columns for each of these plot points. So for example, if a row contains TRUE under 'Attic', that clip will give you the plot point for 'Attic'.

In my run, I use the following words at the beginning to knock out these plot points:

stories - (tells about florence) clothes - (tells about baby and murder) school - (tells about sister/twins and attic)

Then after that, I just use the shortest words possible to get to 109.

Just to make this post simpler I'm inventing the term 'Collision' to mean when one clip is being pulled up twice in the same run by two different search terms. This wastes time because the second search term only gets us 4 clips closer to 109 instead of 5 clips.

There is a delay of exactly 1 second every time you run a search. With 109 clips required to be seen, we could complete the game in exactly 22 search terms if we have 1 or no collisions. However 2 collisions would require 1 additional search term, 7 collisions would require 2 extra search terms and so on. As I said, each search takes 1 second for the computer to process, but of course there's also the human aspect of typing the word, clicking the button, etc...

So, the most optimal word list would cover the following:

1. Hit all 5 plot points
2. Have 22 search terms
3. All search terms have 5 words and no collisions (compromise: there is wiggle room for 1 search term with 4 words OR 1 collision)
4. Search terms be as short as possible (may save 1 or 2 seconds max due to typing, also longer words are easier to mess up)

One last minor thing to consider... there's a few search terms that can be typed exclusively with the left hand (sex, cat, bar, ever, dead). In general these feel faster, but that could just be me.

So, of course I have to mention, I purposely didn't look at anyone else's word lists before doing my own investigation, but after I chose my route and made my first run, I took a look at the competition and saw that, particularly FreezardB's word list, is already super optimal, way more so than mine! :)

Anyway, this game isn't quite to the level of optimization where any of this is even necessary, but I'm the kind of guy that enjoys stuff like this haha

So I'll leave you with a spreadsheet of the current wordlists:

https://docs.google.com/spreadsheets/d/1jrMUhalGDQMMQGlPbe3huOeVS0iNW5zQxEK1AqKR-xw/edit?usp=sharing

Also, can someone change the colors so it's not blue on blue in the forums please? :D

UPDATE: I was able to deconstruct the way the search engine works through the code, and I have added a column to the first spreadsheet link that contains a list of the search terms that bring up each clip.

You may notice that words with apostrophes do not appear (can't, haven't etc...). This is why:

EVERY character that's not a letter or a number is converted into a space by the search engine. So if you search for "can't" it converts the search into "can t". Something like "P${i!ck'$le}" would turn into a search for "P i ck le".

So basically, to find transcripts in the spreadsheet that will come up for "can't", you can filter for lines that have the search term "can" and "t".

I wrote a script to analyze the spreadsheet and come up with what is a VERY slightly more optimal word list.

It also contains no collisions, but the main difference is just shorter word length, since Freezard's list was pretty much already fully optimal.

cloth (satisfies 'baby' and 'murder' plot) call (satisfies 'florence' plot) mum (satisfies 'sister/twin' and 'attic' plot) wig no do don (you can just add an "n" to "do" and save a keystroke) bar ok get say had sex put her door each hear made very dead ever

Anyway, I'll make a guide soon to sum everything up. This has been fun :D

Interesting finds, if you could document these in a guide that will make running this game a lot easier. I have also responded to your request of changing the text color for easier reading, your feedback is appreciated and always welcome. Edit: I also apologize I haven't seen this earlier..