Just going to start by saying there has been a lot of discussion about the structure and rewards and communication surrounding the banquet. This is not going to touch any of those subjects. There are plenty of threads doing that. This is strictly going to focus on the numbers, and what the numbers say, period. If that's something you're interested, and math doesn't put you to sleep, grab a frosty beverage and pull up your chair. If you're looking for a dissection of what went right and wrong, this is not the thread for you.
I wanted to put this out earlier, but it took some time to gather up all the numbers and perform the analysis, so this is a bit later than I expected. Also, I wanted to take some time to really think about what the numbers were saying, and not just dump a bunch of calculations. So here we go.
First: how many players actually participated in the banquet event? The only requirements were be conqueror or above, and open at least one banquet crystal. It isn't easy to figure out how many players this was, because this would require someone to open the bare minimum (one GBC) and then report their rank, which no one did (nor did I on any alt). And even so, if a lot of people actually did that they would all be tied and have different ranks. However, there's another way to narrow this number down: we can look at reported ranks. Every time someone accurately reported their rank and bracket, they created a data point that served to narrow down all possible values for total number of players. If someone says their rank is 30000 and their rank was 4-5%, then that means their highest possible rank was exactly 4%, and their lowest possible rank was 6% (I'm assuming the 4%-5% bracket means 4% up to the next bracket which starts at 6%). In this case, that would mean the lowest number of participating players is 30000/.06 and the highest number is 30000/0.04. Which narrows the total number of participating players to a range of between 500k and 750k. If I do that for all reported data points I could find (from Line chats, from reddit and forums, and special shout to
@RichTheMan who collected data for his banquet video which I
stole incorporated into my analysis) and overlap the ranges to find the narrowest possible range, I end up with a final range of between 533k and 602k players. Because of the way the data fell, I am inclined to believe the number is closer to 553k than 602k (the lower number's associated data points fell much closer to a rank bracket cutoff) I'm going to estimate the total number of participating players was 550k players.
That number is interesting, because during Crystal Cleanse I determined there were at least (and probably just about) 785k active players that were Proven and above (the minimum required to participate). The banquet required Conqueror and above. My guess is the vast majority of active players conqueror and above opened at least one crystal, since they were free, so the total number of participants in the banquet is about the number of active players conqueror and above. So between crystal cleanse and banquet, we now have a vague idea of the active player population distribution:
~ One million active players.
~ 215k below Proven
~ 235k Proven
~ 550k Conqueror and above
This strongly suggests the median MCOC player is Uncollected. Of course, these numbers include all active players, including players that might have just downloaded the game, will play for a few days, then quit. But still, I think it is safe to assume that the median active players
among those likely to stick around for a while is probably at or near Cavalier.
Now, the big banana. How much revenue did the banquet generate? Well this one is very tricky and requires a significant amount of judgment being applied to the relatively little hard data I have. I'm not going to flood the post with numbers and calculations (its a lot), but I will summarize. First, I needed to find some estimated point distribution across the participants. I did that by taking all data for which I have accurate scoring data and precise rank data (which was not a lot) and created a kind of histographic pigeonhole lower bound estimate for the total amount of points scored by different people. To explain: suppose I have these three data points:
Someone scored 91,160 points and ranked 2066. Someone else scored 90,000 points and ranked 2134. What does this tell me about ranks 2067 through 2133? Well, all of them must have scored between 90,000 and 91,160. I don't know how much, but what I know absolutely for certain is that all of them scored 90,000 points or higher. 90,000 is the lower bound on the total number of points those 68 people scored. Which means they cumulatively scored at least 68 x 90,000 = 6,120,000. The same logic says the 9,530 players between ranks 2134 and 11,664 scored at least 41040 x 9,530 = 391,111,200. Which you can see reflected in that table excerpt. Doing that for all the data I have from rank 1 down to rank 260663 I get a total cumulative point total of 3,971,880,070. That's the minimum amount of points that group of players scored, the real number is somewhat higher, but not *too* much higher. I can say that for a couple reasons. First, I have enough data points that the estimate should be close. And second the distribution follows a roughly exponential decay curve:
I mean, that's actually a log scale, and the curve *still* looks logarithmic. Its a doubly exponential drop off. So scores will tend to be weighted towards the bottom of the bins not the top. So let's go with 3.97 billion for now.
These points encompass more than half of all the estimated participating players (550k). The bottom half all scored 5655 points or less. How much I can't say precisely, because I have no data below that score. But I can still make some interesting statements about those players. If we use our participation estimates and divide, we get an average point score of 4284 points for the bottom half of the players. In other words, the scoring doesn't keep dropping like a stone down there. And that makes intuitive sense: participation is hard at the top, because you have to spend money and units. But scoring at the bottom is easy because most of the crystals are free. So we wouldn't expect scoring to drop to zero, we would expect a scoring floor below which most players would not descend below. 4.3k points is just about double the maximum points we'd expect a player to be able to score just from free crystals from ticket exchanges and milestone rewards (solo, banquet, and accolade). 2k more points is like five crystals.
Now, most of those crystals were likely bought with units earned in-game, and few of them were purchased with cash, so none of that scoring is likely to have generated actual revenue. In fact, we can go further. Almost *none* of these points generated revenue. Why? Because most players don't spend on mobile games. The industry average is between 3% and 5%, and if MCOC was converting several times more players, that's something they would probably be bragging about, as it would be a huge industry outlier. So actually, only about 5% of all the participants are likely generating revenue for MCOC, because only that many ever spend.
Now, there's some complexities to that industry average. It encompasses all players who play all mobile games, even for a single day. It counts very ephemeral players. And it averages together games of a wide range of longevity. Presumably, the longer a game has been around, the more veteran players it has, the more time it has had to convert them into spenders. MCOC's spender ratio
for its most active players might be significantly higher than that vanilla 5%, and remember the participants to banquet must be conqueror or higher, which means they are among the top 75% of active players.
Let's say that maybe among all active players, including players that have only been here for a day, MCOC is typical for mobile games and only 5% or so spend, but among all conquerors and higher, that number is higher. Say 10%. Well, then that means only about 100,000 players spend in a game with one million total players. We can use that figure as a guestimate for how many points
the spenders scored in the banquet. We will assume they are the top 100k of all scores. This isn't strictly true, but it is likely to be a very close estimate of the overall situation. When I calculate an estimate for total points only counting the top 100k players or so, I get about 2.9 billion points (specifically, this is the histographic estimate for all scoring down to rank 118413 which is a bin-boundary). The average number of points scored by this group of players is about 24.5k.
So now we have an average score for the most casual free players of about 4.3k, and the average score from spenders of about 24.5k. We can assume the spenders are doing the same things as the free players, and spending on top. To account for that point differential requires opening approximately an additional 48 SBCs.
What do 48 SBCs cost? Well, it depends. There was a way to buy ten for ~$50 USD, which means those cost $5 USD a piece. But those opportunities were limited. You could buy them with units at an exchange rate of about $10 USD per crystal at normal unit bundle prices, but there were ways to do that more efficiently buying unit bundles with more units. Those other opportunities priced crystals anywhere from $6 to $8 USD. This means the amount of revenue those crystals could have generated was anywhere between $28.2m USD and $56.3m USD.
Given the uncertainties involved, I can't say much more than that. The average figure is $47m USD and my guess is that's not too far from the truth. I suspect the majority of spending is still via in-app purchase, which means after the app store tax from Apple and Google Kabam ne Netmarble makes about $33m USD from that. Call it $35m USD with some spending coming from the webstore (which has a much lower margin). That would represent a bit more than half of all the revenue reported by Netmarble as coming from MCOC during the fourth quarter of last year, which seems to land in the general ballpark of what might be reasonable as the banquet cannot represent all of the Q4 spending (most Cyber Weekend spending would not directly go towards banquet points alone, for example).
What do all these numbers tell us about overall participation in the banquet event? Well, it tells me a few things. First of all, it tells me that there was a lot of casual activity, and very little evidence of players completely checking out of the event. A lot of free to play players got in, got their free crystals, spent a small amount of their unit stash, and were fine with that. Those players probably got a very good return on their investment, and I can't imagine most of them being terribly unhappy with their results, both because they didn't spend any cash and did not deplete their units much, and were probably mostly of lower progression and thus the relative value of the rewards were much higher for them on average.
There was also quite a bit of activity among the likely spenders. The average number of crystals spenders likely bought across all spenders was on the order of 50 SBCs. That's not bad, although I have no idea where that number lands in terms of good or bad participation. It seems like a lot to me, but I only have one banquet event to compare to (in terms of solid data).
$35m USD is a lot of money to get out of a spending population of between 50k and 100k players. A lot of games would kill for that, although that number isn't necessarily better than previous years. MCOC spending has declined over the years as the game has gotten older, and whether the 2024 banquet will reverse the trend or not will have to wait for February's Netmarble reporting to determine.
And that's about all I can squeeze out of the banquet data. See you all when the next Realm event comes around.
