GamePress
Hey, Seeker? I have some summoning math questions.
Answers
I'm encouraged to see people take an interest in this.
Let's start with how I calculate the number of orbs expected to find a unit. There are actually ways slightly better than mine, and I think Akariss probably uses them, but my figures are decent.
First we determine the appearance rate of all units. According to information at the bottom of the "details" page of every banner, all units of the same rarity (not colour) share an appearance rate. So for any stone, it's like the game first decides whether it's 3*, 4*, 5* off-focus, or 5* focus, and then from there it picks a random unit in that pool and shows you the colour. Therefore, the appearance rate of any character of a specific rarity is equal to:
(appearance rate of that rarity) / (number of units of that rarity)
Ex: appearance rate of B.Celica = 3% / 4 = 0.75%
Once we have this, we can figure out the appearance rate of stones of any colour by adding together the appearance rates of all characters of all rarities of that colour. So I would have to add the appearance rates of 3* Marth, 4* Marth, repeat for all 3*-4* characters, then 4* Soleil, 5* Soleil, repeat for all 4*-5* characters, 5* Ayra, repeat for all 5* exclusives, and then add any focus units. There is a shortcut, which goes like this: "the 4* pool has has 105 units, and 31 of the are red, so the probability of a 4* unit being red is 31/105." Then I can add the probability of a 3* character being red, a 4* character being red, a 5* off-focus being red, and a 5* focus being red.
I'm not sure if my counts of the units in each pool are dead on because I haven't checked since YT banner, but for CYL2, the percentages should be close to:
R: 30.1%
B: 26.7%
G: 18.2%
C: 24.9%
The probability of finding a specific focus unit on any pull of a given colour is equal to:
(appearance rate of that unit) / (appearance rate of stones of that colour)
Ex: the probability of finding B.Celica on any single red pull is equal to:
0.75% / 30.1% = 2.5%
If this part doesn't make sense, please ask.
As far as I know, every good calculationist (PhoenixMaster, Akariss, etc.) would agree with me up to this point.
From here on I take the "lazy" route because I haven't figured out how to make Excel account for pity rates. What I do is I say "ok, so if the probability is 2.5%, that's equal to 1 in 40. So the math says if you pull 40 reds, you can expect 1 B.Celica." Factually, we know that you can expect it but it won't necessarily happen! So what some people do is they flip it and say "how likely is it that any red stone DOESN'T give you B.Celica?" and there's a not-too-complicated way of figuring out what the probability is that you DON'T get B.Celica in a given number of red pulls. But all the figures I've been giving have used the direct route, just ask the question, "the chance of pulling my desired focus unit is 1 in how many?"
There's one more thing I can do, and the other calculationists will almost certainly agree with me on this part as well. I can predict how often you will be able to pull 1, 2, 3, 4, and 5 stones from a round based on which colours you want. If you want, I can get into the math of it; it involves the probability that any given stone will be a desired colour and the formula for combinations. But what's important is that I can use it to predict how often you're spending 5, 4, or 3 orbs on a pull, which I can then use to calculate an average of how many orbs you're spending per pull. For example, this is oversimplified but just to show how it works, if you want blues, and 4 rounds out of 5 you only get one blue stone and the other round you get two, then you spend 24 orbs per 6 summons for 4 orbs per summon. On the CYL2 banner, the numbers look more like this (again, my counts might be out of date so don't take them for exact figures):
R 4.73 orbs per summon
B 4.78
G 4.88
C 4.80
The things to note are that red is the lowest, green is the highest (most people know from experience that getting more than one green in a round is rare), and the variation across colours is very small. The difference between red and green is the biggest, 0.15, so you spend 1 less orb per 7-ish summons, which means you get an extra pull every 28-35 summons (because a summon usually costs 4-5 orbs). It won't make much difference when the difference in probability of finding the focus unit is something like 1 in 24 vs 1 in 40; you would need to pull 16 more reds than greens so that one extra pull will make only a very small difference in the long run. You'll start seeing numbers like 4.5 orbs per summon only if you pull multiple colours, and of course, if you pull all colours, it does go down to 4.
Once I know the number of orbs per summon, I can multiply that by the number of pulls within which I expect to find the focus unit. For example, I established that B.Celica is around 1 in 40.
40 x 4.73 = 190 orbs. I expect to find B.Celica in 40 summons costing 190 orbs.
- Log in to post comments
This is so cool, thank you! Just to start off simple I used your numbers and calculated that B!Hector would cost an average of 172 orbs, so that’s cool. :D
I’d actually love to see the math behind the number of orbs per pull so I could calculate, for example, the rate of pulling a focus hero if I picked only blue and colorless.
- Log in to post comments
Sorry about the delay. I'm glad it's working out for you! You probably have more up-to-date numbers if your counts are fresh.
I had to do some digging because when I came up with the equations originally, I was going mostly by intuition and I actually got it wrong at least once before I found the solution. I'm not sure if I was correct at that time as to why my solution was correct; I knew only that it was correct because the probabilities added up to 100%. After some digging I think I've figured out why it's correct.
I left off at the probabilities of a stone being any given colour. To get the probability of a stone being of a desired colour, just add the probabilities of the colours you want. But to keep it simple, let's just say we want only reds (because 30% is a tidy number to work with).
The probability of getting five non-reds is (0.7)ˆ5. In mathematical probability, that would probably be phrased as "the chance of getting a non-red five times."
Similarly, the probability of getting five reds is (0.3)ˆ5.
Once we start getting into combinations of reds and non-reds, we need to introduce a concept called permutations with repetition. I'll use the probability of only one red stone as an example.
The equation for the probability of getting a red and four non-reds begins with:
0.3 (the probability of getting a red)
multiplied by
0.7ˆ4 (the probability of getting non-reds four times)
but there's one piece missing: that red could be the top stone, the upper left, the upper right, the lower left, or the lower right. To account for all the possibilities of which stone(s) is/are red and which stone(s) are not, we use the formula for permutations with repetition. I'm leaving you to look it up on your own; I'm sure there's an explanation on the web. The two elements that can get repeated are desired colour and undesired colour.
In the case of one red stone and four non-reds, this is equal to:
5! / 4! 1!.
So the probability of entering a summoning circle and finding exactly one red stone is (0.7)ˆ4 times (0.3)ˆ1 times 5! / 4! 1! = 36%.
- Log in to post comments
This is so cool, thank you! I guess the one thing I’m confused by is why the math cares what position the stone is in, but I’m sure I’ll figure it out if I look it up, you’ve already gone above and beyond. The fact that I don’t know all this as well as I wish I did is entirely my own fault for deciding to sleep in rather than take zero period stats back in high school, and then ending up in a non-math field in college...clearly a terrible mistake. ;)
- Log in to post comments
It's a little bit of puzzler to me too, but it's definitely the way I was taught in high school (or close enough anyways hahaha) and the numbers don't add up without that missing piece (and they do with it) so I'm going with it. If you go on stackexchange or something and ask "when rolling four dice, why isn't the probability of rolling two 3's and two 6's simply (1/6)ˆ4?" someone will have the answer and you can help me out too.
What was your major, by the way? Mine also wasn't math haha
- Log in to post comments
I started in comp sci, actually, but math-wise I only got as far as single-variable calculus (which I loved) and linear algebra (which I hated), and I never took stats (which I kind of regret now). In the end my major was music lol. Our program had a heavy emphasis on theory, so I was definitely still exercising that part of my brain, but not in the same way, so it's been a while since I've done real math beyond "calculate whether item x is something broke music major y can afford". Well, that and fire emblem math, I've done a lot of that. :D (When I played Fates I made an excel spreadsheet that calculated everyone's stat caps and growth rates based on the selected class, it was definitely one of my prouder nerdy achievements.)
- Log in to post comments
Oh you're kidding! I started out as a dual program intending on taking science psychology with music; after one term I decided I found music a lot more inspiring and took piano performance and regretted it. Did you have a specialization or write a thesis or anything like that?
- Log in to post comments
Wow, that's a really funny coincidence. :D My specialization was classical voice, although I ended up doing my senior capstone project in a composition class where our teacher told us "you can write a piece for any instrument you can find someone to play", and I said "challenge accepted" and wrote a piece for harpsichord and viola da gamba. Early music is kind of my thing. ;)
- Log in to post comments
Incredible. I had a long...affair, let's call it, with classical voice during my undergrad. The opera program was big at our school, and I attended some summer programs as well, plus my sight-reading was (and maybe still is?) good so I was quite popular among the singers.
I also studied continuo harpsichord as a course before attending Tafelmusik's summer institute during my master's. Early music is great. We played, perhaps in not-so-historically-exact fashion, Mondonville's Dominus Regnavit with an orchestra of at least 100 members, just the whole cast. Boatloads of fun.
Have you made a career out of it?
- Log in to post comments
Oh man, a good accompanist is worth their weight in gold, I’ll bet you were popular!
WOW, that sounds like one hell of a great experience. I finally got to be in my school’s early music ensemble during my last quarter (it had gone on hiatus before that) and it was wonderful, but our program was definitely a lot smaller and a lot less prestigious. ;)
Sadly I have not made a career out of it, though it would have been fun if I could have. At the moment I am an unemployed person with a semi-useless bachelor’s degree (my last job was in a more IT-ish direction at a community college that ran out of money to pay me - thanks, budget cuts), although I do sing in a local community choir and I’m on the board as well, so I definitely try to keep music as part of my life.
- Log in to post comments
What happens when you pull all colours is that the calculation suddenly becomes really simple. The answer to "what is the probability that a 4* unit is the colour I want" is simply 58%, the appearance rate for any 4* unit. At this point, we're treating all colours as equal, so the probability of any specific pull revealing a focus unit is 3%, which means we expect a focus unit 1 in 33 pulls. (33.3 repeating if you want to be precise, but in practice there's no such thing as 1/3 of a pull.) By comparison, the numbers I currently have for CYL2 banner are: 1 in 40 if you pull only reds, 1 in 36 if you pull only blues, 1 in 24 if you pull only greens, 1 in 33 if you pull only colourless. So yes, if you pull all four colours you do see more focus units than if you pull only reds.
The reason why I still don't go for the CYL2 banner is because I can probably get the heroes cheaper if I go for them 3-focus unit banners with 3%/3%. Assuming no colour sharing (as is the case with CYL2, i.e. all other things are indeed equal), I'm looking at 1 in 30 for B.Celica, 1 in 27 for B.Hector, 1 in 18 for B.Ephraim, and 1 in 25 for B.Veronica. (You can compare those to the CYL2 banner figures I gave; they assume the same 3*, 4*, and off-focus 5* pools.)
There's no getting around the fact that greens are generally less expensive to pull; and at the same time, even for greens, we should look for the best opportunities to pull. For example, M.Grima on the fighter skills banner was around 1 in 18. It's just that for reds, you will tend to pay a lot more severely if you don't get picky with the orb prices you're getting. For example, compare the 1 in 40 on CYL2 with the coming legendary banner. If you like 2/3 reds on the legendary banner, you'll get one around 1 pull in 23 (but keep in mind that if you only want one Swordhart, for example, once you get him your rate drops down to 1 desired 5* in 45 pulls because you don't want him anymore).
You are correct in that Celica has the same chance of appearing as any other focus unit. That would be true whether or not you pulled all colours; it's just that if you don't pull reds, you'll never know if she appeared or not. The fact has not changed that if you isolate the red stones, you will still have more non-focus reds than if you isolate any other colour.
- Log in to post comments
This makes sense, thank you! Out of curiosity (no worries if you don’t know), have you or anyone managed to make the calculations account for what happens when you can’t pull your focus color? It happens most frequently with green, of course, and I’d wonder how much the fact that sometimes you have to waste 5 orbs pulling a color you don’t want impacts the idea that greens are the least expensive to pull. Are they still the least expensive when I have to pull from the wrong color what feels like every other banner?
- Log in to post comments
This is a very good point. My original calculations did not account for this, and the orb estimates I've given are, as you seem to understand, for the number of orbs spent specifically on that colour and that colour only, so anytime you enter a round and there's no desired orb, my estimate loses value. I'm glad you pointed this out; I'm refining the formula to account for the "no desired stone" possibility.
I did think of a way rather quickly after you asked the question, but when I started working on it, I ran into some unexpected results which made me realize there's an entirely other way of calculating the expected number of orbs spent to find a character (which, admittedly, would probably lead us to the same numbers, which helps my uncertainty about my intuition in that part about calculating the probabilities of finding 0, 1, 2, 3, 4, or 5 orbs of a desired colour.
The logic of it goes like this: suppose we're on a 3%/3% banner with three focus units. 1 in 100 stones hides each of the focus units, and 100 stones divided by 5 stones per round = 20 rounds. So we can expect to find the focus unit once in 20 rounds, and meanwhile, we need to reveal (pull) all stones of the correct colour in those 20 rounds. The number of orbs we therefore prepare to spend equals the number of orbs we spend revealing all those stones plus 5 orbs per round in which no stone of desired colour appears. The second part is pretty straightforward: it equals the probability of entering a summoning circle to find 0 stones of desired colour x 20. The first part is similar: it equals 5x the probability of entering a summoning circle to find 1 stone of the desired colour plus 9x the probability of entering a summoning circle to find 2 stones of the desired colour plus 13x for three stones plus 17x for four stones plus 20x for five stones.
If we want focus units from more than one colour, then we adjust accordingly the frequency of entering a summoning round to find 0, 1, 2, 3, 4, or 5 desired stones, and if we are on a banner with different focus appearance rates (ex: four focus units instead of three), then we adjust the number of stones and the number of rounds needed to see that many stones (ex: for four focus units on 3%/3%, 133 stones to see each focus unit once, so 27 rounds).
I henceforth present the new estimates for the CYL2 banner, compared to the previous estimates (in parentheses), which didn't account for rounds when you don't see an orb you want:
R 207 (old estimate: 190)
B 194 (old estimate: 170)
G 165 (old estimate: 119)
C 187 (old estimate: 159)
Again, my counts on the pools may be a bit off because I haven't updated since before YT banner, but these numbers should be pretty close. The differences between the old and new estimates suggest 4-5 rounds out of 27 in which you don't see a stone you want (only 3-ish for red), which seems to line up with my experience and probably that of most players.
One lovely advantage of this new method is that it's now much easier to have a number for multiple colours. However, like my old method, it still doesn't account for pity rates.
My hearty congratulations to you, SourPeridot, and my thanks. You've actually spurred me to improve my methodology, plus, in the process I've actually come up with some ideas about a way to calculate the most efficient number of orbs for a buffer designed to help moderate the swings of RNG. The idea of the buffer is to set aside a certain number of orbs that we only use when variance runs against us and to replenish that padding whenever variance runs in our favour. The threshold can be set at various levels of defensiveness (ex: very defensive would enable us to continue pulling until we reach 90% chance of obtaining the desired focus unit; less defensive might be something like 70%), but the number of orbs required for the buffer increases exponentially the more defensively we try to be (ex: the improvement in likelihood of finding the focus unit in 90 orbs vs 60 is much greater than between 270 orbs and 300) so the idea is to find the goldilocks figure where favourable variance equals unfavourable variance. It's not a foolproof strategy and sometimes variance really will run so bad that the buffer won't enable us to exploit a pity rate, but the idea is to reduce this occurrence. So thanks again for helping me progress.
- Log in to post comments
Wow, thanks, I’m honored to have contributed! The updated calculations are really good to have and the buffer thing sounds really useful, I hope you’ll share more when you figure it out!
There is one thing here I’m confused by, and maybe it would be more obvious if I weren’t still working on wrapping my head around the math, but: once you accounted for the possibility of 0 desired stones, why did the number of orbs for red go up by more than for blue? You’d think red would go up least since the odds of getting no red stones are smaller, but the difference isn’t that big, so maybe it’s just an artifact of the change in the way you did the math?
Once again, I want to say thanks so much for taking the time to explain this to me, this has been really useful and I will definitely be returning to this post for reference. :D
- Log in to post comments
I've been working on this buffer idea. The prospects look worse than I anticipated right now, but this is really pushing the limits of what I know how to do so I'm not even entirely sure what's going on right now haha
I think you made a mistake in your arithmetic. 207 - 190 = 17, but 194 - 170 = 24. Reds did go up less than for blues. Your intuition was correct.
I'm thankful for this chance to spell it out. If anyone ever asks I'll be able to reference this post.
- Log in to post comments
Update: figured out the premise of the solution, but it requires some calculus which I might or might not know. I'm going to see if I can somehow make Excel perform some magic for me haha
Update 2: decided to just brute force it with Excel for an approximation, and assuming a 3%/3% banner with three focus units and pulling blues, I got a figure of 86%ile for an 180-orb buffer. 86% seems suspiciously high. I have no idea right now whether I did it right.
- Log in to post comments

Fate Grand Order
Dragonball Legends
IV Calc
Heroes