I decided to look at some statistics for my Android game Glow Touch. This is a game that allows users to touch the screen with any number of fingers to see a wonderful light show. You can use dozens of finger combos to control colors and shapes on the screen. There is no objective to the game. It is just a novelty.
The only stats I’m collecting are play time, screen size, and how long a number of fingers are being used. The first thing I looked was an average. Looks like the play time is over 6 mins which is great. There is about a 70% drop in finger count for each extra finger. These numbers also tell us that only about 10% of people ever use 2 hands.
Glow Touch Aug Correlation Stats
Next I checked out the correlation each of these variables had to each other. The stronger the correlation (closer to 100%) implies that a user who does X will also do Y. A zero correlation means there is no meaningful connection between the two variables.
The highest correlation I found was the width and height of a phone. This makes sense. No one has an extremely square or oblong phone. The next variables with a high correlation are finger count transitions. Most people who use three fingers also use four finger. The worst transition is five fingers to six. This makes sense since it takes two hands to have six fingers. Maybe I should suggest to people that they should try it with two hands.
The negative correlation between screen width, height, finger count 1, and 2 tells us that users with smaller phones are more likely to use one and two fingers. One of the most interesting correlations is the play time. It is near zero for all variables. This means there aren’t any large factors that engage users to play longer.
I would like a way of tracking what phase people are using the most. Maybe I can duplicate the best phases and eliminate the least played. I can see that the phase was changed 2k times in 500 sessions. That sounds really low. I need a better way of tracking user behavior for phase flow.
I have a few take always from these numbers. It looks like user engagement is really good. I need to add tips and hints to convince people to use more fingers. I also need to add better phase tracking.
The next interesting graph to look at is the mass of the sun when the app closes. This isn’t a very good way of measuring the mass of the sun since the sun slowly burns away. If you leave the app running before it closes, it will report a much smaller sun. This is why there is a large grouping near 1200, the smallest the sun can get. The most interesting part of this graph is that some players were getting the sun huge! Sometimes so big it would fill the entire screen. The largest sun reported was 480k.
The last thing I’m going to look at is the correlation of all of these factors. How likely is it that if a player has a large sun mass, they will also have a high score. The table on the left shows the correlation of these variables. +1 is a perfect positive correlation and 0 means no correlation at all. Most most interesting thing we see in this table is that there is very little correlation to any of these variables. The biggest correlation is to the total planets on the screen and the the number of scoring planets. This makes sense. Also, the larger the sun’s mass, the higher your score is. That is interesting. This is possibly because these people are playing longer.
You are more likely to accomplish your goals when you write them down and share them. Last year I 
BoulderBuff64's Thoughts 