Theater Betting Episode 5: Mathemagician

Day 4 Postmortem

The results look a bit more normal now, but A still won.

Lynch Ceia

please and thank you



A triple crown, just like I predicted.

God, I knew this would happen too. 

I said C neutral, A/B winning is a coin flip, then I guessed wrong. 

I'm sticking to B, because it's about time for C to win. 

By the way, EN didn't manage to advance a single node yesterday. We're still on Hard 6: 

If anyone's wondering how to beat the Doppelsoldners with an RFHG team, check out kfourit's video below!

K4 uploads Theater runs every day, so please consider subscribing to see more in the future!

For the Doppelsoldners, you don't necessarily need the Dorothy or Carcano M91/38 that K4 brought - any standard RFHG will do fine in the same fight. 

The M4 cannon team also works, if you know what you're doing. 

I uploaded my run too. 

B lost


This is the worst timeline, Kira won and Ceia brought us all down.

I'm going back to sleep like a sad man.

We made the news apparently. 

I'm gonna lose all credibility after Theater 1. 

smh ceia

Never believe the lies of an Anmiface.

Can't trust them, with how innocent they look, but inside is STONE COLD HEART OF ICE.

On the bright side, this graph is finally looking closer to typical older server charts. 

Day 5 Bet Brainstorming

Alright kwes, get in here. 

You got someone else for this?

Since we clearly have no idea what we're doing, I decided to add in somebody who does. 

Hi, I'm Kwespell, and I spent 11k parts for 7 CE.

So far, I'm at like 3 wins and 1 draw. Hope that keeps up. 

What did you guess?

I went B-B-A-A. 

B got a neutral score on Day 2, so I still made a profit. 

How about today? 

It's gonna be a coinflip between B/C. 

No shit Sherlock. 

It can't be A 4 days in a row right???

Kwespell is a computer scientist, and he does a lot of machine learning work. 

With access to all the historical Theater data from KR, TW, and CN, it should theoretically be possible to train something that has better odds than what we've been able to do so far. 

Last I heard, he wanted to go with a random decision forest implementation. 

Oh I forgot about that entire thing.

I haven't even set it up. 

Wait, where's your guess coming from then? 

Just eyeballing the Theater plots from older servers. 

My thing will just say "ignore my vote and vote for not Ceia".

I'm streaking B and I ain't gonna stop now.

Rush B

It made sense until i thought a bit and it didn't, and after thinking a bit it did make sense.

I need some water :ribbleed:

God, I seriously need to win today




Kwes please, you’re doing way better than everyone else here.


I'm unsubscribing.

Alright let's try this again. C'mon in Blount.

Hey, I'm DahBlount (DahBlount#0387).

You will find me wandering the GFC Discord, quite often consuming potentially illicit substances in the name of math and theorycrafting.

I do occasionally pop into GFEN as well.

I spend much of my time yelling at people who complain about bad RNG and can sometimes be seen ranting about how "you too can improve your crafting statistics".

DahBlount is a statistician and very well-versed in analytical mathematics. 

I'm an Electrical Engineer and Computer Scientist, not a statistician.

Too late, you're a statistician now. 

For theater, we're looking for an analytical solution that gives us the probability of a given area (A, B, or C) being the least "popular" the next day. 

How did you tackle it? 

Well, C is definitely not the best choice, doesn't take math to see that. 

It looks like zone A was, on average, the least popular for the first Theater. 

I focused on data from the first Theater in foreign servers since this will be our first Theater in EN and the results between each server for a given Theater run look pretty similar. 

This also cuts down on the data set size, which saves me some trouble orienting it for SciPy usage. 

After a lot of grueling manual work, I've managed to do some regression analysis using Decision Trees. 

Would you mind elaborating a bit more on that?

It's machine learning basically, but nothing like a neural network.

Essentially, it uses statistics to make decisions about a regression or classification (in this case the win/lose of each sector). 

I'm also testing a log-likelihood, recursive spatial model, that one basically plots the points for all datasets as distinct directional graphs then use log-likelihood to recursively refine a solution for the next day given the previous days results.

It's something I've seen used for gambling before with pretty good results, although that was hands for Blackjack :P

With all of that set up, I could, in theory, use the outputs of them both combined with what I know about EN's population to come up with a "best guess" answer. 

So what does all of that say about EN's day 5 results? 

Oh, I gave up on my statistical models. EN thus far actually defy all logic and the actual results don't match any of what my algorithms gave back. 

Not again!

I honestly wasn't too sure as to how bad EN actually was until now. On top of the results from CT, it's safe to say I've lost all faith in EN. 

If you want to hear my wild-ass guess, after the 3 As in a row and how close together yesterdays results actually were...

Here's the idea: all the folks already in B and C are probably going to toss things up a bit, so I'm almost certain A is going to lose today. 

I think more people from B than C are going to jump ship since I believe that C mostly consists of folks who want to "lose less" so to speak, so I'm going with B for today. 

If we can't rely on maths and statistics, we'll just have to leave it up to the OJ sims!

Just saying, but yesterday's OJ sims were 100% accurate. 

Oh, I'll run a game of OJ right now!

Day 5 Bets

B Again?!

Looks like C will win at this rate. 

Guess I'm going for that negative score. 

>Ceia voted for B


I already mentioned many times I'm taking B because I got a B on my exam


Can't blame me this time.

Sure we can.


I for one, blame Ceia.

In all seriousness, the latest scouting results are actually kinda normal. 

I just wanted top 10%...

Fury, if you just do Theater daily and can clear all the content, you stay top 10% even if you don't minmax anything. 

But I got knocked back to 21%.

You're fine, you haven't spent your materials, so your actual score is much higher than it might appear. 

Being top 10% not gonna be related to betting.

Core 8 clears are like 30k points each, scouting/investigation changes your score by much less than that, and materials are earned as a fixed % of your score anyway.

Early rankings won't really show you anything.

When we start doing core dailies and everyone dumps their materials on Core facilities, then it starts mattering. 

I see why every guide said to invest in HOC buildings now. 

5/5 HOC charge improved my quality of life so much. 

Dalaos will throw everything at HOC building. 

We don't want echelon slots, we need more HOC charges. 

If they still haven't seen my guide video, then I'm gonna lose hope. 

I bet EN will bloat echelon slots again and make Judge harder for no reason. 


The OJ sims are in guys, B > C > A. 

Same here. 

This panel was sponsored by Fruitbat Factory and 100% Orange Juice, now available on Steam

I wish, lol. 

We've got a 7/12 B vote for some reason. Guess no one here learned their lesson.  

Remember to send all 54 drones available today to your zone of choice (which may or may not be the same as what we picked)! If you guess right, you'll fully cancel out any wrong guesses from days 2 and 3!

For endgame players that just want to maximize their points, each day after the ceasefire, you can always find the TL;DR on what zones are optimal to do for that day at the start of the Theater Guide!

Oh, and don't forget to submit the CBT form every day when you investigate in-game! Here's the histogram of scores after four days, with only 13 users maintaining a perfect pick record: 


Enjoyed the article?
Consider supporting GamePress and the author of this article by joining GamePress Boost!

About the Author(s)

(AKA Nonotan, Discord Cleista#7481)

Guide writer and editor for Girls' Frontline, and one of the few people who like its gameplay more than the gacha. Got MICA to pay me 8000 gems once. 

Send guide suggestions and ideas via DM on Discord, or drop some in the #gf_site_suggestions channel on the Community Discord