The logistics behind Niantic’s inability to deal with cheaters and how bots can fix it

Pokémon GO Datamine

If you are like us, you’ve probably often wondered GPS spoofing and location mocking still exists in Pokémon GO. It’s 2018 after all — Niantic has surely had ample time develop a solution, test it and to ban everyone caught cheating?

In this article, we want to share a different perspective on the issue of cheaters in Pokémon GO: logistics of a widespread ban wave. As you will soon see, it’s not all sunshines and rainbows when you take this into account.

Hooman is… outdated?

Let’s pretend for a moment that Niantic has access to the following data, tools and technologies:

  • A complete audit log of every player’s movement since the date they started
  • A set of algorithms that can detect cheating with 100% certainty using the above mentioned data set
  • A data center that can detect cheaters instantly, on a press of a button

In this imaginary scenario, computational power and detection accuracy are factors that can be ignored, allowing us to focus on the aftermath of a “widespread ban wave”.

What is a widespread ban wave, anyways? Let’s say (conservatively) that around 5% of the Pokémon GO player base cheats by using GPS mocking tools.

Pokémon GO has around 65 million monthly active players (April 2017 data, source John Hanke), which leads us to this:

  • Estimated player base size: 65M
  • Estimated percentage of cheaters: 5%
  • Estimated number of cheaters: 3.25 million players

Let’s say that 10% of banned players decide to complain about the ban and request evidence for the reasoning behind their account being terminated. After all, many of these players are paying customers and, in their mind, they will not give up their accounts easily.

  • 10% out of 3.25M is 325 000 complaints
  • On average, let’s say that a support agent can successfully resolve a complaints every 15 minutes during a typical workday. Let’s also say that Niantic has 100 support agents on their disposal, resulting in 3200 complaints solved per day.

This leads us to a simple formula for time required to clear through the complaint backlog:

325 000 complaints / 3200 solved complaints per day
101.5 days worth of work

In other words, for the next three months Niantic’s complete support team would be completely prevented from offering real support to the rest of the player base (61.75M).

Actual face of a Niantic Customer Service agent after dealing with 32 cheating complaints.

Remember, these numbers are calculated by using an imaginary setup where Niantic has infinite computational power, a 100% accurate detection algorithm and a bullet proof log of every action you ever took.

The reality is much more grim, which leaves only one viable solution: support and ban bots. Surely these machine-learning powered tools will be able to deal with a potential complaint outbreak when push comes to show?

OK Computer

Yes, the answer to the question posed above is yes. Bots, machine learning and a strict category system for complaint management can and will be able to deal with the aftermath of a widespread ban wave.

Unfortunately, these systems take a really long time to develop, train and they generally get better with time if the initial setup was good (see training function MLand supervised learning for more info).

With the help of bots, Niantic can set up a system where it’s very difficult to reach an actual human with your complaint. For example, here’s a possible bot-and-human-powered complaint scenario:

  • A player posts a complaint by selecting categories and subcategories: Complaint > Ban Complaint > Banned for GPS Spoofing > Enter player name
  • A support bot responds with a greeting and proceeds to fetch the decision data set from the previously mentioned “all knowing database”
  • The bot processes the data and presents the data to the player grouped by incidents and ToS violations
  • The player can either agree with a decision or escalate the complaint to human powered customer support. Escalation can only occur if the player is able to provide evidence (pictures, receipts, etc) that overrule bot’s decision
  • In case of escalation, human customer support deals with the issue, further improving training level of the bot system for future cases.
We, for one, welcome our robotic overlords


League of Legends has the Tribunal. Ingress has Operation Portal Recon. These are tools created for dealing with massive backlogs and helping actual flesh-and-blood humans with their work. If we are ever to get rid of bots, Pokémon GO also needs that tool.

P.S. Niantic, if you’re reading this, we have a name suggestion for the: Banizard.

Antonio started the Hub in July 2016 and hasn't had much sleep since. Software developer. Discord username: Zeroghan. 28 years old.