TheSea.ash - Automated Underwater Quests

Theraze

Active member
It's true. :) I run this because it's easier than clicking ~150 times per ascension.

But if I have adventures left, since I use the makeshift scuba rather than effects for my breathing, I know that I'm ready for the boss. If I don't have adventures, that message is lying to me. :)

The eventual fix will be to detect if we're up to the boss before printing that we're ready. But that will take me trying to pull out the right variables... probably just validating that we've finished Ringo, but I might do something fancier. We'll see when my brain organizes a fight between my productive muscle and the laziness muscle. Depending on who wins.
 
I find the script sometimes gets stuck for me on the choice adventures in the Outpost when I'm at the stage that I should be searching for Grandma. This most recent run, it did just that, after I'd found the lock key. So, i manually clicked the right choice in the relay browser and got the stashbox, then opened it and used the map, then asked grandpa about the intense currents.

I then re-ran the script and it found grandma, and went on to the Abyss.

THEN, it came back to the outpost with the statements as follows:

You have completed the Sea Monkee quest.
We need a lockkey before we can open the stashbox.
You need 1 more Mer-kin lockkey to continue.
Conditions list cleared.
Condition added: Mer-kin lockkey

Of course, this will just keep adventuring for ever. I'm guessing it doesn't do any checks to note that i have opened up the intense currents?

Can i change a flag? I'd like to keep this going through to the outfit piece (as i've successfully done in the past) but unsure how to get it back on track...
 

Theraze

Active member
The lockkey message comes up if the Coral Corral isn't unlocked. If for some reason you have a gif besides corrala.gif or corralb.gif for your Corral picture at the seafloor, please let me know what it is so that we can detect that and not break you in the future.

Otherwise, use your trailmap, talk to Grandpa about currents, and re-run the script when the Corral is visible.
 
Thanks Theraze. Looking at it I am wondering if it was a cached image issue in mafia. Running it today after rollover without changing anything it all looks ok and is running as i'd expect it to... Don't have any better answer, but does seem to be cruising now... It's now asking me to train with the lasso, as I'd expect. Cheers.
 

fronobulax

Developer
Probably because you had WHAM off, so it didn't auto-throw the mer scrolls and learn the clues. The script will try to learn the dark skill, will use the knucklebone, and learn the 3 library clues. If you are using WHAM, it will learn the 2 mer-scroll clues if you have them in your inventory to throw.

If you learn the skill-clue manually, that's 7 clues. Or if you eat the worktea, that's 7 clues. Either way.

But if you have Sea questions for the future, probably best to ask them on TheSea's thread rather than on the BCA thread. :)

Feature request - Have TheSea buy a kill and/or heal scroll if it is looking for clues and the corresponding clue has not been found. (I would add, only when WHAM is being used but I don't know how TheSea would detect that). I have seen cases where I never get the clue and I am running WHAM so I presume it is because the appropriate scroll just is not in inventory.

Comment perhaps leading to a FR - How about a scroll guesser capability? If you guess wrong you get 3 turns of the effect for each slot that had the wrong value. The number of turns each guess generated could be used to prune the tree. I suppose I should investigate the logic myself...
 

Crowther

Active member
Comment perhaps leading to a FR - How about a scroll guesser capability? If you guess wrong you get 3 turns of the effect for each slot that had the wrong value. The number of turns each guess generated could be used to prune the tree. I suppose I should investigate the logic myself...
If you look up Mastermind solvers, you'll get some ideas how to write a solver. I wrote one in C. I'm sure others have written them. Basically, you want to make a guess that partitions the number of possible solutions as well as possible (well meaning minimizing the size of the partition with the largest number of solutions).
 

Winterbay

New member
That doesn't help. The problem is that WHAM (and SS specifically) will only use those items after the dreadscroll has been acquired and TheSea seems to abort after I have received the dreadscroll so the items never get used...
Code:
[1345] Mer-kin Library
Encounter: Hook, Line and Sinker
You acquire      an item: Mer-kin dreadscroll

Conditions satisfied after 6      adventures.
[COLOR=black]You have 1 Mer-kin knucklebone, no      adventuring necessary.[/COLOR]
Using 1 Mer-kin knucklebone...
Mer-kin      knucklebone clue: south
Finished using 1 Mer-kin knucklebone.
Time      to figure out the dreadscroll!
We don't know a definite solution for 1
We      don't know a definite solution for 2
We don't know a definite solution      for 3
We don't know a definite solution for 5
We don't know a      definite solution for 6
We don't know a definite solution for 7
We      don't know a definite solution for 8
We are aborting with 7 unknown      scroll segments.
No luck with figuring out the deepscroll. You'll      probably need to figure out a few more steps yourself.
You need 50 sand      dollars to complete this quest.
Buying the boot and collecting our      well-earned reward.
Verifying ingredients for damp old boot (1)...
Purchasing      damp old boot (1 @ 50 sand dollars)...
Visiting the Big Brother...
You      acquire an item: damp old boot
Big Brother successfully looted!
Successfully      created damp old boot (1)
You acquire an item: damp old wallet
Using      1 damp old wallet...
You acquire sand dollar (66)
Finished using 1      damp old wallet.

After this the script stopped.
 

lostcalpolydude

Developer
That doesn't help. The problem is that WHAM (and SS specifically) will only use those items after the dreadscroll has been acquired

Sounds like that should be changed in WHAM/SS then, to check for vocabulary training or whatever instead (a good indication that that path is being taken, and a requirement for heal/killscrolls being useful), regardless of what TheSea does.
 

Theraze

Active member
For your case, Winterbay, if you're using BBB, the next step is obtaining 3 choiceadvs at the library. I don't force-run that, because if people aren't using BBB or something similar, then they need to manually pick the right choices to learn the clues. Unless your luck is stupendous and you get 3 choiceadvs in a row, you should run into at least one combat which should let you throw the mer-scrolls.

There is an auto-solver for a single unknown entry, frono. If you look at the solveHatred function, you can remove the "unknowncount > 1" abort field if you really want it to try to guess with multiple segments, though as the code currently runs, you'll need to run the solveHatred function as many times as specific unknown entries you'd like it to learn. And it expects full success, so you'd probably want to mess with the deep-tainted mind check as well... unknowncount-1 *3, something like that. Or even just if the rounds are less than unknowncount*3 or whatever the specific is... I always skip the food since that's the one I don't know if it will happen on this day or not. *shrugs*
But I don't think I want to officially add support for solving 8 unknown questions. :) I could see someone running it and cursing as it burns through 300 rounds slowly.
 

Winterbay

New member
That would be slightly more annoying than me setting my Seaky Pete CCS to "attack" and then losing 100+ fights against a protector spirit before I noticed :)

Also, it appears that much of my problem was that I had forbidden BBB to change choiceAdventures, probably as a way to combat interaction with BCA so the library ones never got set.
 

Theraze

Active member
Funny that you have your AoSP CCS set to attack and mine is set to WHAM. :) Which makes it abort winless after the first spirit without elemental damage, generally the hospital, but at least it's only a few wasted turns instead of 100+. :)
 

heeheehee

Developer
If you look up Mastermind solvers, you'll get some ideas how to write a solver. I wrote one in C. I'm sure others have written them. Basically, you want to make a guess that partitions the number of possible solutions as well as possible (well meaning minimizing the size of the partition with the largest number of solutions).

The main problem is that you only get feedback regarding the correct number, as opposed to how many were in the incorrect position but correct value as well as the correct number. Your best bet is probably to just tweak one parameter at a time, since if you vary two things, it's harder to determine exactly what information you got.
 
Last edited:

Winterbay

New member
Funny that you have your AoSP CCS set to attack and mine is set to WHAM. :) Which makes it abort winless after the first spirit without elemental damage, generally the hospital, but at least it's only a few wasted turns instead of 100+. :)

Well, I was under way of creating a more detailed one and then got interrupted and then forgot about it and then ran BCA while playing GuildWars2 so didn't notice for like an hour :)
 

Crowther

Active member
The main problem is that you only get feedback regarding the correct number, as opposed to how many were in the incorrect position but correct value as well as the correct number. Your best bet is probably to just tweak one parameter at a time, since if you vary two things, it's harder to determine exactly what information you got.
Harder for you and me, but not harder for an algorithm. The algorithm, which I believe is optimal, normally only varies one parameter at a time. However, once it guesses wrong twice on a slot, it will tweak that slot and the next one. While it's not exactly the same as Mastermind, the idea of partitioning the results should work optimally for this problem as well. Sometimes tweaking two spots gives more information than tweaking just one.
 

heeheehee

Developer
Okay. Let me work this out. If you simply try a third time on that same slot (after two failures), you resolve uncertainty in one coin flip. If you try both of them, you have a 25% chance of knowing both for sure (+2 or -1 score), 37.5% chance of knowing that one of those can be deduced (+1 score) but not which one, and 37.5% chance of no score change (so both incorrect->incorrect, or one incorrect->correct, one correct->incorrect).

In principle, the second slot really only has ~1.5 bits of entropy, since we can resolve it in one guess 50% of the time and two guesses 50% of the time. (It's a little less than this because you have info about the sum of the scores of that and all later slots)

I'll be approximating left and right in this part. Simple approach reduces entropy by 1. Case 1 resolves 2.5 bits. Case 2 resolves 0.67 bits (2/3 chance of resolving 1 bit, 1/3 chance of resolving 1.5 bits, but it's uncertain which is which; we can resolve this in a single coin flip. I'm not sure how your algorithm deals with this case). Case 3 resolves the same amount as case 2. This adds up to, uh, 1.125 bits of information. I guess the algorithm slightly better. It'll save maybe most of 1 adv, so... :|
 

Crowther

Active member
Agreed, slightly better for a lot more work. That's just the one (only?) deviation I noticed. I wrote the code, because I had multiple slots unknown. Too little fish or something. I wanted to guess and burn the turns finishing other quests outside the sea and guess again, so I could ascend quickly. Partly I wanted to see what optimal looked like.
 

digitrev

New member
I've written a (fairly naive) piece of code to do this in python. The worst-case scenarios for guessing n unknown slots are as follows. I can't get the 7 or 8 unknown case because my terrible code scales horribly.
  • 1 unknown - 4 guesses
  • 2 unknowns - 5 guesses
  • 3 unknowns - 6 guesses
  • 4 unknowns - 10 guesses
  • 5 unknowns - 16 guesses
  • 6 unknowns - 16 guesses

Repeating this, but for how many adventures of Deep-Tainted Mind you'll get in the worst-case
  • 1 unknown - 9 adv
  • 2 unknowns - 24 adv
  • 3 unknowns - 45 adv
  • 4 unknowns - 96 adv
  • 5 unknowns - 192 adv
  • 6 unknowns - 201 adv

So that provides you with an upper limit of how many adventures you should spend to get the next clue, assuming you can handle the adventures of Deep Tainted Mind.
 
Last edited:

Theraze

Active member
Yeah, I'm not doing anything that will potentially burn off 200 turns without warning to guess your 4 unknown slots. Sorry! :)
 

digitrev

New member
Yeah, I'm not doing anything that will potentially burn off 200 turns without warning to guess your 4 unknown slots. Sorry! :)

Oh, I certainly wasn't recommending that you do that. I'm just doing the math for fun. It's a weird habit I picked up at some point. Stupid complicated undergrad.
 
Top