A tool to analyze application statistics!?

Question

Game development

We are looking for a tool for detailed analysis of statistics and evaluation of player behavior!
Those. in addition to online statistics (it can also be collected from nginx \ apache logs), I would like to investigate the dependence of events in the application, evaluate the actions of players in certain situations.
Those. example - how many players gain more than 3 levels during the first hour, how many of them are chosen
At the moment - due to lack of experience - there is no complete information about which statistics are important for research, so I want to collect everything in a row, and then analyze it.
At the moment - in the process of being implemented in the google analytics application, unfortunately, it is not clear yet - what specific analysis opportunities it provides.
And it’s not clear what he has with the limits - we have about 3,500 players per day * 100 events * 30 days = 10,500,000 requests per month, Google says about support.google.com/analytics/bin/answer.py?hl= ru&a... 10,000,000 requests per month.
The game is a firmum on flash for social networks.
Question to the owners of such games - how do you research the statistics in your games - what tools do you use!
Links to the mat.chast on the issue are also welcome!
The possibility of inventing your own bicycle for collecting and analyzing statistics is not ruled out!

Answer 1

You can make your bike in R, a programming language for statistical data processing. For him, there are a lot of packages with already implemented statistical algorithms for all occasions. There were introductory articles on Habré: here , here and here .

Answer 2

You've already been advised R, but I wouldn't recommend it. The fact is that, as a programming language, it is simply terrible (it is already used on chewed, ready-made data, it is better to do all the logic for processing logs on python / java / etc), so it’s better to do this:
0) For complex analytics of GA capabilities, it may not enough, in this case you can implement the analytics system yourself, but keep in mind that this is quite laborious. There are no ready-made tools as such (if you need more features than GA), so you have to implement everything yourself.
1) Develop a clear plan for exactly what you want to measure. I understand that it is difficult to do this without experience, but believe me, it is better to do it now than in a month, when you have already killed a lot of time writing server logic for processing statistics. If GA is enough for your tasks, use it and don't worry. If not, see paragraph 2.
2) Discard everything unnecessary and of little significance from paragraph 1. Leave only the most important thing that you are guaranteed to need.
3) Implement it in the simplest possible way (but so that during the implementation you would create an infrastructure that allows you to write additional metrics as quickly as possible).
4) Get the first data and try to solve the problem with the existing metrics. At this point, it will become clear to you what metrics are missing. Return to point 1.