Tero dataset:

The data is json-formatted in utf8 encoding.

---

Important note:

Until further notice, due to privacy concerns, the data format has been modified as follows:

1. No user_id is mentioned.
2. All latency datapoint are rounded to the closest hour.

Files:

• locations.json : single json file containing all user locations.
• raw/latency : contains the latency digits as extracted from gaming thumbnails after comparing the values from 3 OCR engines, separated by game and month.
• raw/alternative_values : contains all values extracted from gaming thumbnails if there are at least two different values; i.e. not all points from latency have an entry in alternative values.
• processed/latency : contains latency values after cleaning, removing glitches and replacing suspect values using alternative values if possible.
• processed/glitches : contains latency values marked as glitches (latency decreases due to image processing errors).
• processed/spikes : contains latency values marked as spikes (latency increases).
• processed/clusters : contains json files with the latency clusters per country and game.

locations.json:

• The json file contains:
  • data : list of user locations, formatted as follows: {user_id, location: {country, alpha-2 country_code, (region), (city), latitude, longitude}, (temporal index)}. Parameters in brackets are optional.
  • last updated : unix timestamp indicating the last time the data was modified.
• Region refers to the largest administrative subdivision of the country; i.e. state in the US, province in Canada, and region in France.
• Locations include the coordinates of their geographical center: these coordinates are not the user coordinates, but a generic point common to all users in the same location.
• Each user id can be associated with one or more locations. Users with more than one location appear several times on the list, each time with a different location. In these cases, the parameter temporal index is used to give partial order to the different locations.
• All location data except the country code is in local language.

raw/latency:

• Each zip file contains a list of json files, one per day of the month (if data is available).
• Each daily json file contains:
  • data : list of latency entries per user, formatted as follows: {user_id, game_id, values: [{date as unix timestamp, latency}]}. The user_id corresponds to locations as listed in the locations.json file.
  • last updated : unix timestamp indicating the last time the data was modified.

raw/alternative_values:

• Each zip file contains a list of json files, one per day of the month (if data is available).
• Each daily json file contains:
  • data : list of alternative latency values per user, formatted as follows: {user_id, game_id, values: [{date as unix timestamp, list of alternatives}]}. The user_id corresponds to locations as listed in the locations.json file.
  • last updated : unix timestamp indicating the last time the data was modified.

processed/latency:

• Each zip file contains a list of json files, one per day of the month (if data is available).
• Each daily json file contains:
  • data : list of latency values per user after running a custom cleaning procedure that removes glitches (latency decreases due to image processing errors). The data field is formatted as follows: {user_id, game_id, values: [{date as unix timestamp, latency}]}. The user_id correspond to locations as listed in the locations.json file.
  • last updated : unix timestamp indicating the last time the data was modified.

processed/glitches:

• Each zip file contains a list of json files, one per day of the month (if data is available).
• Each daily json file contains:
  • data : list of latency values discarded from the pool as possible glitches, divided per user, formatted as follows: {user_id, game_id, values: [{date as unix timestamp, latency}]}.
  • last updated : unix timestamp indicating the last time the data was modified.

processed/spikes:

• Each zip file contains a list of json files, one per day of the month (if data is available).
• Each daily json file contains:
  • data : list of latency values marked as spikes, divided per user, formatted as follows: {user_id, game_id, values: [{date as unix timestamp, latency}]}.
  • last updated : unix timestamp indicating the last time the data was modified.

processed/clusters:

• Each json contains the clusters for a given game in a country.
• Each daily json file contains:
  • game_id : code of the game from which latency was extracted.
  • country : alpha-2 country code.
  • data : list of clusters per geographical region (administrative subdivision), formatted as follows: {region, min: minimum latency value of the cluster, max: maximum latency value of the cluster, coverage: percentage of users in the region contained in the cluster, n_users: number of users in the cluster}.
  • last updated : unix timestamp indicating the last time the data was modified.