Serefina's Summoner War Research: De-Mystified

Alright, so a lot of you who know me in game and out tend to ask me, "what do you research anyway?". So to explain that, I've put together a quick little presentation that explains how Summoner War summoning works on a very basic level, and what my research aims to do.

I originally wanted to update this deck (as it was made prior to the 2015 full year analysis), but my PC is still out of commission and I'm stuck with what data exists in my old laptop. That said, keep in mind the graphs and charts aren't fully reflective of current summon trends, though much of the conceptual implications remain true. I'll be updating these after I fix my PC once and for all.

How Probabilities Work and What That Means for Random-Chance Based Games

That means that the odds of getting a specific monster or star grade in Summoner's War should not be impacted by what you summoned before, the number of scrolls you used, or what monsters you have. There should be fixed odds that distribute themselves in a random fashion that follows a normally probabilistic distribution over time (Guassian curve anyone?).

Of course that doesn't mean outliners, or unusual occurrences such as high levels of duplication can't occur; it just means it's really unlikely. Kind of like winning the lottery---or getting pooped on by pigeons three times in a year.

WOW! 0.000095%? So why the hell did I just get five water harpies in a row when I summoned? Shouldn't that rate be stupid low? That's bonkers!

Well, yes. That's assuming the RNG is truly random, rates aren't weighed, and there are no controlling factors programmed in, but do you really think a game company isn't going to add some controls for challenging game play if not profit?

The RNG isn't Truly Random

 Now cosmic noise methods and those like them are really, really expensive. So you can bet Com2us isn't shelling out dollars for it. However, not being truly random might not be a bad thing. Adding weights or controls can keep you form getting extreme situations such as everyone getting Camilla or no one getting Camilla ever. Extreme situations are bad for profit. They're also demoralizing; but with real life and things like clicking banner ads or dying by falling air conditioner (or pianos), these sorts of situations can and do happen.

What Summoner Rates Would Look Like with No Weight

Well gee, doesn't that look strange? Shouldn't I, you know, have like, 20 5* by now because I've used like, 1000 scrolls if not more? Yes. Yes you should if the rates weren't weighed. But rates are in fact weighed. That means, Com2us is essentially saying hey "people are going to get a 4* ten times more often than a 5* and a 3 * 20 times more often".

To go back to the marbles example; it's like being handed a bag of marbles that can contain red, green, or blue marbles, but Com2us is filling 10 bags with evenly distributed marbles, 20 bags with mostly green marbles, 30 bags with mostly blue marbles, and 40 bags with mostly red marbles, and you get handed a bag at random. The odds of getting a bag of mostly red marbles is much higher, and within that bag you have a 50% change of getting a red marble.

So...if its based off an algorithm with weights, does that mean we can figure out the algorithm?

While predicting the algorithm would be nice; it would require way too much data and some serious hacking skills to do it. Much better to just make broad estimates, play for enjoyment, and most importantly, make the best of what you got. Summoner's War does a pretty good job of providing "free" monsters that can do almost everything needed when combined correctly. It just takes a lot of skill and strategy.

Besides, getting everything you ever wanted would get boring anyway.

What Serefina's Analysis Does

What Summoner's War Summon Rates Really Look Like

When I say lightning, I mean getting those sparks from an uncommon or mystical scroll. Using a combination of survey data (thanks everyone who fills out my questionnaires), Youtube Data (it's a pain in the neck to log all those 100 scroll summons but it helps a lot!), and uncommon scroll tests (which is why I never have any); there have been distinct patterns for when you get lightning.

It's not so much a, lets use 100 scrolls thing, more of a "after this content update, I'm getting  lightning within my first 20 scrolls". Which means spamming really doesn't help after you pass this threshold. However what that frequency will be seems to change every time they add more monsters as you'll see later on.

Elements Like to Reoccur on Certain Days

Though lightning frequencies might be hard to log unless you're as crazy as I am; predicting day of week skews takes little more than observation once you know to look for it. After a major update, you'll notice that certain monsters (like wind hell lady's on Monday's) occur a lot more often on Monday than Sunday. The same seems to hold true for Ifrit, suggesting there might be some sort of implication day of week has. Or it might be something to do with how date affects CPU Clock.

This is actually only from 12/1/2014 through 6/10/2015, so it's a bit outdated, but I'll get you guys similar charts when I finish my analysis. You can also copy and paste the database data into excel and create this chart yourself if you are data-savvy.

Time Affects Summoning

Now this is where things get dicey. When looking at individual days and individual hours, we see a lot of skews; especially weekend skews. However as with the days when LD and Mystic scrolls are given as rewards, or the first three days when ToA resets, there can be a large influx of 5* due to everyone and their mother summoning at the same time.

That suggest that there are distinct player behavior factors that influence dates in the sample, but it doesn't completely discount the hypothesis that date or day or week may also be controlled in the game's coding.

As with the previous observation; we can see that content addition seems to spike 5 Star Summon volumes likely due to player behavior. While the code may or may not be "jacked up" to increase odds, players are more likely to summon when new content is added as it provides immediate incentive, which may explain the trends you see in the abridged graphs.

However, on a week over week basis we see a lot of these trends continue, suggesting that there is some sort of basic population control based off total volume of summons or number of summoners that keeps everyone feeling like they have a chance to get their very own Dark Oracle.

That means most weeks will see roughly the same volume of 5 Star summoned, and that weeks that start off high will taper, while weeks that start off low will see more towards the end of the week.

Of course, behavioral factors cannot be discounted, and can be argued to be the true underlying cause.

Summary of Findings