My attempt to solve Lower Prison (updated: 200 runs now)

After getting DESTROYED by seemingly horrible layouts in today's Brutus race, I decided to do a hundred (and one!) runs of Lower Prison to see if I could make better sense of it [EDIT: I've added another hundred runs for 200 total!]. Below are the results.



Each data point tells you: was the correct direction from the waypoint left/down ("west") or right/up ("east"); and did the layout generally move in an upwards ("up"), downwards ("down"), or straight ("straight") direction towards the exit (I'll explain DOGLEG later).

In the raw data (below) I also recorded whether a master was encountered somewhere in the layout. I did not think to note when i saw a shrine until it was too late. Note that I may have just missed some masters that existed but I'm sure that I stumbled in to the majority of them. Masters and Shrines really messed with my intuition sometimes as I tried to learn the layouts and guess the correct path, but I'm not even 100% sure if they actually screw up "standard" layouts (if there is such a thing).

[EDIT: raw data now 200 runs total. The second half has shrines noted as well as masters]


As for DOGLEG: for a while this was the only 100% foolproof pattern I found. I noticed that when the west direction IMMEDIATELY deadended (so there's just 1 tiny empty room in that direction), the correct path always seemed to start east, but then immediately wrap back around to the west (passing underneath the waypoint again), in an L-shape basically, and the exit would be to the west. This worked the first 9 times that I got this pattern!... but I decided to keep going until I got the western deadend ONE LAST TIME to "confirm", and sure enough it wasn't a dogleg that time. It just went east. Fuck. (This happens to be the 101st data point, btw) [EDIT: after encountering more doglegs in the second hundred runs, I feel that this pattern is VERY reliable after all].


Dogleg Example


Anyway, the data generally says that it's way better to go East, and down/straight instead of up. There are some more specific room patterns that you need to recognize to maximize your success rate though. For example, it generally seems better to travel through as many doors as possible (marked in yellow on the map).

After the first 50 or so runs, the general heuristic I settled on for running this zone was:

- Go west, if it looks like a deadend (and there are a few pretty obvious cases, which is why I suggest this even though east seems statistically better), chances are it is a deadend, so go east.
- If you see Chatters in the first or second room, GO THE OTHER WAY (unless you just came from a direction that was clearly a deadend!). If you see chatters much later on a particular side (east or west), you are probably on the correct side.
- If you're going west, generally you should heard straight/down; if east, go straight/up (unless its a dogleg)

To confirm all of this, I'm going to do another 100 runs and see what my success rate is, and to collect more data! (Yes I am crazy, but this was actually kinda fun and good practice). Then I guess I'll try Upper? (ugh...) [EDIT: I've done this now - see reply post below.]

GGG THIS IS THE IMPORTANT PART: pathing RNG is stupid from a racing perspective. Fixed seeds without the map revealed (like the boss kill races) don't really help with this, since you can still guess wrong while someone else guesses right. If there really ARE patterns that you can learn for every zone (which my limited data suggests, but certainly does not confirm), then great, but I still think it would be better if there were "breadcrumbs" in each zone (e.g. The torches in Prison could lead to the exit; add some skeleton/corpse breadcrumbs to Submerged Passage; follow the red carpet in Solaris; *SOME "CORRUPTION" EFFECT DOWN PATHS THAT LEAD TO A CORRUPTED ZONE*; etc.). Then your layout generator can generate as many deadends as it wants, and noone that is paying attention will complain.

russia confirmed the next world super power

"

Fightgarr wrote:

Then I guess I'll try Upper? (ugh...)

rip

You are doing Sanic's work, my friend.

e: funnier name

Mind doing something similar for The Vaal Ruins too? :^)

"

passofexhale wrote:

Mind doing something similar for The Vaal Ruins too? :^)

I honestly don't find vaal ruins very confusing. Its long and horrible sure, but you don't actually get LOST very badly. It's always a long wavy zig-zag hallway, and once you know the orientation, its really not that confusing.

In other news, I've completed a second batch of 100 runs of lower, AND I ran upper this time as well (not collecting any data, just trying to learn... anything). I had an amazing 82% success rate on lower prison layouts this time, so studying paid off! Upper however, is a hopeless mess. Sometimes (like, less than half the time!) it spirals inwards and this helps you find the exit door once you are going in the correct general direction (you usually have to turn upwards to hit the door... most experienced racers will know this already), but good luck getting to that point. The deadends are huge and unrecognizable. So, fuck Upper. GGG breadcrumbs PLZ

fightgarr if i do this type of research for every zone in the game do you think i can finally get gooder?

Yeah, seems impossible to know where to go. Intuition is used, but it is often wrong on Prison layouts.

And that bothers me how Gucci knew perfectly where to go on his fastest Brutus kill.

To me, as an intermediate racer, to me it looks like he cheated. Say he did not, but it feels like he did, from my racing view.

He perfetly did Mud flats, even if that was unusual layout.
He perfectly traveled Submergedd pasage. (my brothers wasted at least 1 minute more on that particular layut that particular race, it was nasty layout!)
He perfectly traveled Lower prison.
He did 95% well on Uppoer prison.

No matter how good i am and even when i guess right, i end up spending at least 10-30 seconds more, and he just straight from point to point runned exacrtly where needed. Especially on Lower prison, and Mud flats.

"

Lordemore wrote:

And that bothers me how Gucci knew perfectly where to go on his fastest Brutus kill.

The more excellent players in a Brutus kill race, the more likely that there will be a lucky one that makes all the right coin flip decisions. Comparing yourself to that, even as a great racer, let alone a mediocre one, isn't really worthwhile IMO. It's just variance. He won all the coinflips this time, next time it will be someone else, and if you're good enough, there's a chance it will be you. Play enough races, and if you're good enough, eventually it will be you.

Yeah... still learning. :)

Will practice more in next race season.

I just hope one day that they make something like alternate art Lighting Coil, or any verified strong item that is popular. (not just alt art belts, rings, ammys)