Musings about queue manipulation in run

Public forum for the discussion of strategy.
Post Reply
yojimbos_law
Spy vs. Pie
Posts: 78
Joined: Mon Jun 30, 2014 12:05 am

Musings about queue manipulation in run

Post by yojimbos_law » Tue Jan 05, 2016 6:51 pm

With all of the different and limited banishing sources available, I thought I'd write up my thoughts on strategies for effective queue manipulation, since I like to think I have a pretty good understanding of how this particular thing works and haven't seen much of anything on the interwebs discussing it.

What is this ridiculously voweled, French bullshit?

Quoth the wiki, "The adventure queue is a feature in the game that makes high occurrences of the same adventure in a short period of time less likely. It works by keeping track of the last few adventures, and then sometimes rejecting an adventure if it's rolled again." Effectively, your last five combats and last five noncombats in a zone are stored in an ordered list, which I will refer to as the "combat queue" and "noncombat queue" respectively or simply "queue" when no confusion is likely.

The way the game uses these lists to select your encounter is moderately complicated, and a complete description of the process can be found here, but for this discussion, it suffices to think of the process as follows: if the combat that the game rolls for your next encounter in a zone is in that zone's queue, 75% of the time it will "reject" this choice and repeat that roll (possibly rolling the same encounter again and/or rejecting it again). If a monster is olfacted, in addition to the extra three copies of it added to the zone, the monster will be exempt from being rejected in this manner.

I just want my goddamn blasting soda. How does knowing about this crap matter?

If you have a limitless supply of banishers and olfaction uses, it doesn't matter; olfact any monster you want to see more than once, banish every monster you don't want to see. This was typically the case in SC seal clubber runs given the presence of LTBs (louder than bombs), pantsgiving, batter up!, tennis balls, and a number of older sources. However, in Standard 2016 and a variety of HC paths (especially for you lazy saucerers), banishers are much more limited and it is thus important to evaluate how effective your banishers are in specific situations so that you can make an informed decision about whether you want to banish a particular monster or conserve the banisher for a better use in later turns.

Suppose you're a seal clubber with a few LTBs, batter up, and olfaction available. You're about to adventure in the haunted laundry room in search of your coveted blasting soda. You adventure there, and your first encounter is an undesirable plaid ghost, which you batter up, because the RNG screwed you out of T.U.R.D.S. keys in a-boo peak. Next, you encounter a cabinet which you promptly olfact; sadly, you're not running +1900% item because you're bad, so it doesn't drop its delicious soda. At this point, you have a 4/5 chance to find the cabinet again next turn, but you don't because you're unlucky and suboptimally encounter a laundry press. You take a moment to post a tirade about swingy mechanics in /hardcore and GD, inhale deeply to calm down, allowing the rich aroma of your parents' basement to work its magic on your psyche, and absentmindedly bloody your crotch with your thinking knife as you ponder whether or not to use a LTB.

If you do, you'll certainly never see this turncount ruining asshole again, but you're worried about running into a potentially valuable target for this banisher in later turns, since the level 2-10 and 12 quests are untouched at this point in your run. If you don't, this monster will now be in the laundry room's queue, meaning that you'll have a 1/17 chance to encounter it on any given subsequent turn, so the probability that you never see it again is very high (significantly greater than (16/17)^3 ~= 0.83 if you are at least running +400% item). You expect your LTBs will do a lot of work in the other eleven quests, so you disinfect your thoroughly mutilated genitals, wipe the blood spatter from your hands and forearms with a moist towelette, discard the towelette onto a sizable pile of stained rags in the corner of the dimly lit room, and defeat the laundry press using some overelaborate strategy that minimizes resource expenditure--confident that this will be the last laundry press to mar your speedrun. You drop to casual five turns later after failing to encounter another cabinet, draft a kmail to (#9) requesting the deletion of your account, and let out a sigh of satisfaction knowing that your decision not to banish that laundry press was almost certainly mathematically correct--despite the results.

Changing the scenario slightly, suppose instead that you batter up the first undesirable monster and encounter the second immediately after. Taking the queue into account, you will have a 80% chance to run into the cabinet and olfact it on the following turn if you do not use the LTB. After doing so, you will be in the same situation outlined above, with a 16/17 encounter rate on subsequent cabinets. If you LTB, then, you are effectively saving an average of 1/0.8 - 1 = 1/4 a turn and an olfaction usage in addition to whatever piddly amount would be saved in the earlier scenario. If you are running low on sgeeas or are well-stocked on LTBs, this may be a sufficiently good use for that banisher. At the very least, it's nontrivially more likely to make a difference than in the other example.

Note that in the above discussion, it was important that we needed to encounter the desired monster multiple times. If we had only needed to find it once, then the banishing effect in the second scenario would only be saving 1/4 of a turn, on average, which would not be worth the resource in many runs. It's hard to put a specific value on the turnsavings that one should shoot for with a given banisher, and I don't really have any good rule of thumb to estimate it either. I know that banishes which save more than a half a turn or so are fairly unusual in an ascension context, so if you can get that as a lower bound, it's almost certainly worth the banish. The notion that you should craft a bounding argument to make a simple decision in an ascension is fairly silly, however, so it's typically more productive to look through some simulations to get a sense of an 'average' banish. This is mostly what I've done to improve my familiarity with it, not including an independent study intimately related to it.

Oh, okay. That's neat. Are there any other ways the adventure queue should influence my decisions?

Yes:
  • The queue makes banishing monsters that exist in multiple zones much more effective than it would otherwise be since you're eliminating the relatively high probability of encountering that monster for the first time in the second zone (e.g. at least 1/3 for a given non-tool-holding gremlin).
  • Skipped noncombats are not added to the noncombat queue. This is particularly notable in the castle ground floor and haunted ballroom, where it is correct to not skip any noncombat that you're seeing for the first time if you're looking for the boning knife (also chocolate) or song set. In the ballroom in particular, doing so increases the likelihood that future noncombats are the song set from 50% to 80%, which reduces the tail on this zone's length drastically while also reducing the average number of turns needed to set the song (by about 1, typically).
Edit:
  • Those percents that kolmafia lists as encounter rates use knowledge of your current queue state, which is also tracked by mafia, so using mafia can make implementing a lot of these ideas much more reasonable than it might otherwise be.
TL;DR: this place isn't for your kind. >:[

User avatar
BC_Goldman
My Pie Blown Sky High
Posts: 499
Joined: Tue Jul 22, 2008 9:44 pm
Location: New Jersey

Re: Musings about queue manipulation in run

Post by BC_Goldman » Tue Jan 19, 2016 9:16 pm

You take a moment to post a tirade about swingy mechanics in /hardcore and GD, inhale deeply to calm down, allowing the rich aroma of your parents' basement to work its magic on your psyche, and absentmindedly bloody your crotch with your thinking knife as you ponder whether or not to use a LTB.
Worth reading just for that.

User avatar
Eroquin
Pie of the Sky
Posts: 47
Joined: Sun Dec 20, 2009 12:41 pm

Re: Musings about queue manipulation in run

Post by Eroquin » Tue Feb 02, 2016 9:21 am

After having a very long ballroom this run just due to general bad RNG, it's nice to know I wasn't just making shit up by not skipping the dancers. I mean, I thought that's how it should work, but I also thought Rainbow's Gravitation would be a good short term investment for flipping, so...

Also, I totally didn't know about TURDS key and ghosts. Learn something new, etc.

Post Reply