If you’re a casual gamer, then now’s the time for me to tell you this: “Turn back!” This article go into the science of Neko Atsume, and doesn’t stop until it comes out the other side with kittens clinging to us.
“Now I’m all out of Frisky bitz !” - Punctual
The Bear went and posted a link to Data Atsume nominating it for a prestigious award in our small circle of (mostly) friends from college who like fart jokes and the occassional dick pick. He’d talked about cookie clicker. Beknownst to him or not I looked into it a bit. It already had some app that told you everything to buy next to make Grandma’s treats come out faster and faster. I saw the allure, but didn’t get into it.
But this time when we had an award for a site I felt I needed to check out the game in more detail. Data Atsume had a lot of stuff talking about all the super special combos you can do to get the rarest cats, and patterns for optimizing how many unique cats you’ll see.
Maybe I can sit down and take pictures of rare cats all afternoon and it’ll be a days work?
They give me some silver and gold fish and tell me to buy kitten toys. Sure. No problem. Lemme just look at Atsume’s list…
Everything he’s talking about is so expensive. And he’s talking about putting out 2 types of cat food at once? Wait…
Yeah, this has ABSOLUTE ZERO on what to do for your new cat collector, sallying forth into the back yard la première fois. Nothing to tell our newly made photographers on how they might weigh their almighty choice? “WHAT DO I BUY FIRST?!”
I shot a surprised email back to the Bear.
‘They’re telling you what cats want to come, but they’re doing nothing to tell you your roi as a player who hasn’t yet made it rich. Other useful features that are missing include adjusted fish income when talking about “getting the most fish”. This should definitely be an option as the store tells you you can trade 250 silver fish for 10 gold fish.’
I also had done some poking around to see if anyone had a real data driven guide about how to start make some quick silver and gold fish. Data Atsume posts a link to the data set on which jewelloree’s analysis is based. The data set does some exploratory analysis, but I felt it fell flat at giving players what would be useful to the most players.
I decided I needed a guide to help me determine the answer to this question:
I hit on a this Quora guide. Here was someone using the same data. Unfortunately William Chen, although he plays Neko Atsume daily, doesn’t expound on his analysis methods, just throws out some numbers and some recommendations. There’s no way I’m going to trust his data analysis over my own.
It’s about this time, I realize I’m gonna need to run the numbers myself.
I make a local copy of the Neko Atsume Database: “Raw Data” and “Toy Information” and start my journey.
As beginners to Neko Atsume, here’s where we’re starting from, and what we want as an end result.
After all options become available to us, what we do with collecting kittens becomes more complex, and is beyond the scope of this primer. I suggest looking at Data Atsume’s Guide for more ideas. In the meantime, we’re going to work though the many substeps of step 2:
The hardest part about step 2 is determining what toys to buy to make the cats the most thankful. We explore this question in great detail for the rest of the primer.
There was an undertaking track the effects of placing out certain toys. This momentus data collection program was undertaken by several users on reddit. We hope to use their efforts to determine which displayed toys result in the highest return on investment.
With the current data set I don’t think we can accurately predict an hourly fish income rate. Thus, the objective of this primer is to get readers to buy toys that, in so far as we can tell, yield a lot of fish per usage.
“What’s the difference between high fish yield and hourly income?”
We don’t have any data on how long each cat spends so overjoyed with a toy that it can’t pull itself away. Nor does our data set yield any information about how many times per hour a cat actually plays with a toy.
Accordingly, we focus on fish yield.
Here I begin digging around in the actual data. I’m doing everything in R, and using the dplyr package.
require(dplyr)I began by loading the data I’ve downloaded into R. Unfortunately the “Toy Information” spreadsheet tells the price of the item, but uses the color of the cell to denote the type of fish required to purchase the toy. I copied the data to my own spreadsheet and split the one cell into two, indicating both the number of fish and the type of fish required to make a purchase. Below is the result.
items <- read.csv("Copy of Neko Atsume Database - Toy Information.csv") %>% as.tbl
items ## Source: local data frame [126 x 7]
##
## Name Category Size Cat.Capacity Price Type
## (fctr) (fctr) (fctr) (fctr) (int) (fctr)
## 1 Baseball Balls Small 1 90 Silver
## 2 Rubber ball (red) Balls Small 1 60 Silver
## 3 Rubber ball (yellow) Balls Small 1 60 Silver
## 4 Rubber ball (blue) Balls Small 1 60 Silver
## 5 Watermelon ball Balls Small 1 80 Silver
## 6 Ping pong ball Balls Small 1 50 Silver
## 7 Soccer ball Balls Small 1 10 Gold
## 8 Temari ball Balls Small 1 25 Gold
## 9 Stress reliever Balls Small 1 80 Silver
## 10 Ball of yarn Balls Small 1 120 Silver
## .. ... ... ... ... ... ...
## Variables not shown: Description (fctr)Next up was the raw data. We want to get an idea of the data frame, so we run a quick summary on the entire thing. It turns out that we have so many variables we can’t even see everything properly.
raw <- read.csv("Neko Atsume Database - Raw Data.csv") %>% as.tbl
raw## Source: local data frame [12,528 x 10]
##
## Name Toy Fish Type
## (fctr) (fctr) (int) (fctr)
## 1 Snowball / Shironeko-san Ball of yarn 3 Silver
## 2 Smokey / Kuroneko-san Scratching board 5 Silver
## 3 Smokey / Kuroneko-san Rubber ball (red) 1 Silver
## 4 Shadow / Haiiro-san Rubber ball (red) 7 Silver
## 5 Sunny / Tobimike-san Mister mouse 2 Silver
## 6 Pumpkin / Shirochatora-san Cat metropolis 2 Gold
## 7 Bandit / Sabigara-san Cat metropolis 3 Gold
## 8 Pumpkin / Shirochatora-san Baseball 11 Silver
## 9 Socks / Kutsushita-san Cat metropolis 3 Gold
## 10 Callie / Mike-san Ball of yarn 1 Silver
## .. ... ... ... ...
## Variables not shown: Silver.Equivalent (int), Gold.Equivalent (dbl),
## Food.condition (fctr), Date (fctr), Theme (fctr), Toy.Size (fctr)Next we look at the outputs
raw %>% colnames()## [1] "Name" "Toy" "Fish"
## [4] "Type" "Silver.Equivalent" "Gold.Equivalent"
## [7] "Food.condition" "Date" "Theme"
## [10] "Toy.Size"The initial data frame has way more variables than we’re interested in handling in this primer, so we’ll pare things down considerably. We’re most interested in the following columns:
data <- select(raw, c(Toy:Silver.Equivalent,Food.condition,Toy.Size)) %>% as.tbl
data## Source: local data frame [12,528 x 6]
##
## Toy Fish Type Silver.Equivalent Food.condition
## (fctr) (int) (fctr) (int) (fctr)
## 1 Ball of yarn 3 Silver 3 Frisky bitz
## 2 Scratching board 5 Silver 5 Frisky bitz
## 3 Rubber ball (red) 1 Silver 1 Frisky bitz
## 4 Rubber ball (red) 7 Silver 7 Frisky bitz
## 5 Mister mouse 2 Silver 2 Frisky bitz
## 6 Cat metropolis 2 Gold 100 Frisky bitz
## 7 Cat metropolis 3 Gold 150 Frisky bitz
## 8 Baseball 11 Silver 11 Frisky bitz
## 9 Cat metropolis 3 Gold 150 Frisky bitz
## 10 Ball of yarn 1 Silver 1 Frisky bitz
## .. ... ... ... ... ...
## Variables not shown: Toy.Size (fctr)Aside about Cat Food
You’ll notice above that the food above is referred to as Food.condition. This is because many cats will only make appearances with specific foods out. From the outset, setting out food is already going to influence our yield.
The first thing that we’ll do here is only select the food Frisky bitz. It is by far the most affordable of the cat foods you have to pay for, and independent sources all agree that it brings a high average fish yield. An exploratory analysis has lead me to agree for now.
Fdata <- filter(data, Food.condition == "Frisky bitz")Along with this, we’ll make sure we remove all foods from the list of toys, since we have set these up as two distinct realms. With this complete, we no longer care about the category of toy or the toy’s description. So we ditch those two variables.
items <- filter(items, Category != "Food") %>%
select(-c(Category, Description))A further note on cat food. Not all foods last the same amount of time (if uneaten). All of the ‘premium’ foods (ones that cost gold fish) seemingly last three hours, while Frisky bitz lasts 6 hours, and Thrifty bitz lasts 12 hours.
Aside about Toys
Toys come in two sizes: Small and Large. Small toys can be placed anywhere and hold a maximum of 1 cat. Large toys have to be placed in a central position, and take up the same amount of space as two small toys. Many times a large toy can hold more than 1 cat.
With the current data set it is hard to know whether having space for more cats on a toy makes it more likely that high numbers of cats will visit said toy. For the present we will assume that each space on a large toy has the same chance of being visited by any 1 small toy.
We create a new variable called efficiency factor that takes into account the number of locations used and the number of cats the structure is able to hold. We then generate the efficiency factor for each toy.
effCalc <- function(Size,Capacity) {
Capacity <- as.integer(Capacity)
if(Size=="Small") return(Capacity)
else if(Size == "Large") return(Capacity/2)
}
Eff.Factor <- apply(select(items, c(Size,Cat.Capacity)), 1, function(x) effCalc(x[1],x[2]))
items<- cbind(items,Eff.Factor) Now we can dispense with the variables Cat.Capacity, since it is a part of efficiency.
items <- items %>% select(-Cat.Capacity) %>% as.tbl
items## Source: local data frame [120 x 5]
##
## Name Size Price Type Eff.Factor
## (fctr) (fctr) (int) (fctr) (dbl)
## 1 Baseball Small 90 Silver 1
## 2 Rubber ball (red) Small 60 Silver 1
## 3 Rubber ball (yellow) Small 60 Silver 1
## 4 Rubber ball (blue) Small 60 Silver 1
## 5 Watermelon ball Small 80 Silver 1
## 6 Ping pong ball Small 50 Silver 1
## 7 Soccer ball Small 10 Gold 1
## 8 Temari ball Small 25 Gold 1
## 9 Stress reliever Small 80 Silver 1
## 10 Ball of yarn Small 120 Silver 1
## .. ... ... ... ... ...Next we’ll summarize toy yields in terms of silver fish equivalents. To do this we first comb through the data looking for all toys of the same name. We take these groups of toys and calculate the mean, standard deviation, and number of visits for each.
sfeSummary <- Fdata %>% # group the data into which type of toy it comes from
group_by(Toy) %>% # make calculations on the sfe and find the number of visits
summarize(meanYield = mean(Silver.Equivalent), sdYield = sd(Silver.Equivalent), Visits = n())
sfeSummary %>% arrange(desc(meanYield)) %>% as.tbl # print sorted table## Source: local data frame [121 x 4]
##
## Toy meanYield sdYield Visits
## (fctr) (dbl) (dbl) (int)
## 1 Warm sock 153.50000 174.02203 4
## 2 Rubber ball (blue) 102.28571 130.27754 7
## 3 Frisky bitz 66.67089 73.23701 79
## 4 Cushion (pink) 58.75000 111.94731 16
## 5 Beach umbrella 53.10714 105.66100 28
## 6 Twisty rail 48.40000 112.95688 15
## 7 Scratching log 45.73077 106.42915 26
## 8 Snow sled 42.91176 99.32358 34
## 9 Burger cushion 42.76923 100.64058 13
## 10 Cat pancake 42.05882 106.93893 17
## .. ... ... ... ...A glance here shows that the two highest yields have only a few visits, and that the standard deviation is insanely large. This is a strong sign that there is not enough data to be confident that these visits to be representative of the true mean.
In the next part of our analysis, we merge the sfeSummary and items data sets.
sfe <- merge(sfeSummary,items, by.x = "Toy", by.y = "Name") We create a new variable that will replace mean yield and Eff.Factor, called adjusted yield.
sfe <- sfe %>% mutate(adjYield = meanYield*Eff.Factor) # make new variable
sfe <- sfe %>% select(-c(meanYield,Eff.Factor)) %>% select(1,7,2:6) # delete old variables and rearrange
sfe %>% arrange(desc(adjYield)) %>% head(20) # print sorted table## Toy adjYield sdYield Visits Size Price Type
## 1 Warm sock 153.50000 174.02203 4 Small 11 Gold
## 2 Rubber ball (blue) 102.28571 130.27754 7 Small 60 Silver
## 3 Beach umbrella 79.66071 105.66100 28 Large 340 Silver
## 4 Cushion (pink) 58.75000 111.94731 16 Small 100 Silver
## 5 Paper umbrella 56.07692 79.34738 26 Large 50 Gold
## 6 Cat metropolis 52.91429 58.24166 315 Large 50 Gold
## 7 Tunnel (3D piece) 49.80392 68.58137 255 Large 20 Gold
## 8 Scratching log 45.73077 106.42915 26 Small 30 Gold
## 9 Giant cushion 45.32143 68.08211 56 Small 210 Silver
## 10 Snow sled 42.91176 99.32358 34 Large 140 Silver
## 11 Burger cushion 42.76923 100.64058 13 Small 15 Gold
## 12 Cat pancake 42.05882 106.93893 17 Small 24 Gold
## 13 Cushion (brown) 41.00000 92.22256 7 Small 100 Silver
## 14 Basket case 40.92857 90.76891 14 Small 320 Silver
## 15 Fruit basket 39.27586 84.89680 58 Small 80 Silver
## 16 Bucket (red) 39.12500 105.41949 8 Small 60 Silver
## 17 Black head space 37.76000 89.11943 25 Small 380 Silver
## 18 Butterfly swarm 37.14286 84.86900 14 Small 12 Gold
## 19 Bureau with pot 35.50000 53.52107 100 Large 950 Silver
## 20 Ball of yarn 35.05357 83.44414 56 Small 120 SilverWith this, I’m done for now. Good luck photographing cats and collecting fish!