I’m glad you liked it!
(. . . could you spoiler your strategy and win chance? I know this challenge is three years old, and what you mention here isn’t game-breaking info, but I want to keep it possible for people looking/playing through the archives to seek clarifications in the comments without unwittingly/unwillingly learning anything else about the scenario.)
That makes sense, ty.
The leaderboard will track how well you’ve done relative to random/best play at the # of soldiers you chose to bring.
Could you elaborate on this? I think I’d do better relative to best play with
high numbers of soldiers,
and do better relative to random play with
low numbers of soldiers,
so it’s not clear which way I should lean; also, I don’t know how you plan to quantify “relative to”.
What we’re facing:
A horrifying number of Tyrants,
A large quantity of Scarabs and Abominations, and
A below-par-given-they-showed-up-at-all-but-still-significantly-above-zero count of Crawlers and Venompedes.
Relevant Weapons:
Artillery is the optimal counter for Tyrants.
Miniguns are very good at handling Scarabs (to the point that bringing more than one would likely be overkill), and pretty useless at most handling most other xenos (to the point that bringing more than one would likely harm our chances).
Lances are good counters for anything which isn’t a Tyrant or a Scarab. (And also not-terrible vs Tyrants)
Torpedos are slightly better than Lances when facing Abominations, and only slightly worse than Artillery when facing Tyrants.
(As far as I can tell, the other four weapons aren’t worth considering.)
Current strategies per number of soldiers:
8 Soldiers: 3 Artillery, 2 Lances, 1 Minigun, 2 Torpedos.
(My model says this gives me >99% chance of survival, but also says that about just bringing one of every weapon. We can be more daring!)
7 Soldiers: 3 Artillery, 2 Lances, 1 Minigun, 1 Torpedo.
(My model says this gives me ~95% chance of survival.)
6 Soldiers: 2 Artillery, 2 Lances, 1 Minigun, 1 Torpedo.
(My model says this gives me about a 2⁄3 chance of waking up the next morning.)
5 Soldiers: 2 Artillery, 1 Lance, 1 Minigun, 1 Torpedo.
(My model says this has slightly worse odds than a game of Russian Roulette with five bullets loaded.)
4 Soldiers: 1 Artillery, 1 Lance, 1 Minigun, 1 Torpedo.
(My model says this almost gives me an entire 1% survival chance.)
If I have to pick one strategy:
7 Soldiers: 3 Artillery, 2 Lances, 1 Minigun, 1 Torpedo.
Description of an investigative cul-de-sac:
I notice that
Duels between a Tyrant and an Artilleryman always end well.
Duels between a Tyrant and a Minigunner, Phaser or Flamethrower always end badly.
Tyrant vs Artilleryman 2v2s . . . don’t happen, ever. (Turns out the quartermasters do display some nonrandom behaviors, and one of these is a bias towards weapon variety.)
2v2s involving two Tyrants, an Artilleryman, and someone who’d lose a 1v1 against a Tyrant . . . end well pretty much exactly half the time, regardless of which [MPF] is used.
I reason that
This is what we’d see in a turn-based fight where humans aggressively heroically always take the first move, and the xenos move randomly. The Artilleryman caps a Tyrant every time; the remaining Tyrant then picks a random human to squish; they pick the dud half the time; we get the coinflip we see.
But then
I find out that there are 2v1 fights between two Tyrants and a lone Artilleryman, and these have the exact same 50% win chance; the dud isn’t even useful as a decoy; my hypothesis is falsified.
From all this I conclude
Absolutely nothing.
Like, conceptually it’s absolutely unpredictable
That’s exactly what I was going for; I wanted phenomena which couldn’t have been predicted without using the dataset.
Misc. prelim notes:
There’s a random element. (Existence proof: 16079 and 17759 were the same fight but we only lost 17759.)
There’s an implicit chrono effect: It looks like this war has been developing not necessarily to our advantage. (Luckily it seems like this is probably ‘just’ enemies outnumbering our troops more frequently in later rows, and not anyone actually getting better/worse at their job.)
The number of troops sent scales with the size of the enemy forces, making inference trickier; however, I haven’t seen anything contradicting the hypothesis that loadouts are decided by throwing darts at a board.
Specific weapons counter specific enemies: in particular, the Minigun is usually pretty lousy, but drops Scarabs like flies.
I expected to find synergies between weapons, and didn’t. I did, however, find some antisynergies: Miniguns and Flamethrowers are hella redundant (presumably because they’re both anti-Scarab bugspray), and the [MPR] set all clash with each other (“Why do you need Gun? You already have Gun!”)
Guaranteed victories seem possible. (A single soldier with a minigun can perfectly-reliably survive 5 Scarabs, but not 6.)
Thanks for running this when my one was going to be late, and thanks for checking with me beforehand.
(Also, thanks for the scenario, like, in general: it looks like a fun one!)
I (to my own surprise) got an “above average” score when I took this test a few years back, which I attribute mostly to the lack of emotional and circumstantial ‘noise’ in the images. I don’t think being able to tell what is being emoted by a professional actor told to display exactly one (1) emotion, with no mediating factors, has much connection with being able to read actual people.
(. . . though a level-2 version with tags like “excited but hesitant” or “proud and angry” or “cheerful; unrelatedly, lowkey seasick” could actually be extremely useful, now I think on it.)
Typos:
“Al gore”->”Al Gore”
“newpaper”->”newspaper”
“south park”->”South Park”
“scott alexander”->”Scott Alexander”
“a littler deeper”->”a little deeper”
“Ai”->”AI”
(. . . I’m now really curious as to why you keep decapitalizing names and proper nouns.)
Regarding the actual content of the post: appreciated, approved, and strong-upvoted. Thank you.
This, linked at “Never.” in the OP.
an alliance socialist nations
an alliance of socialist nations
I didn’t like this post, but I did very much like the “insight porn” post it linked to. (Unfortunately LW doesn’t let you simultaneously downvote and strong-upvote a post, so consider my weak-upvote as a sum-of-vibes.)
If someone says ‘What’s for supper?’ a beginner will desperately try to think up something original. He will carefully evaluate dozens of options in his mind.
“Is this funny?” “Will this not reveal something weird about myself?”
It will take him ages to come up with something and eventually he will say something “fried mermaid”.
An improv pro would simply respond “fish”.
Taken—almost verbatim, without attribution—from Impro, by Keith Johnstone. (I don’t know whether LW would consider this plagiarism, or consider that to be bad.)
taking it out early and letting it sit
What I actually usually do is move it from the freezer to the refrigerator like 15min before I eat it, so the change in temperature is more predictable and evenly distributed (instead of some parts being melted while others stay too cold).
Is the point that it’s initially too hard to scoop?
That and it being too cold to properly enjoy the taste.
(The votes on my original comment make me think most people are less concerned about their dessert-that’s-supposed-to-be-cold being too cold. Typical-mind strikes again, I guess.)
I enjoyed the exercise, thanks!
You’re welcome, and thank you for playing.
(I wrote a custom loss function for the NN)
I’m curious how you defined that. (i.e. was it “gradient = x for rows where predicted>actual, gradient = −8x for rows where actual>predicted”, or something finickier?)
I think this is because
LightGBM and its kin are tools for creating decision forests, not decision trees. If you use standard hyperparameters while creating a single-tree model then they will under-train, resulting in the “predict in a way that’s correlated with reality but ridiculously conservative in its deviations from the average” behavior you see here. Setting num_boost_round (or whatever parameter decides the number of trees) to 200 or so should go some way to fixing that problem (while giving you the new problem of having produced an incomprehensible-to-humans black-box model which can only be evaluated by its output).
(I would have said this sooner but helping a player while the challenge was still running seemed like a bad look.)
If I’m following your notation right, it looks like you mixed up Flamethrowers and Miniguns.