Turning Empty Maps Into Data

Chris Shughrue - April 3, 2020

On March 16, we launched MapJakarta, our first weekly regional contest. Players in the Indonesian capital could now earn extra Bitcoin by mapping places in their city, bringing the global dynamic we’d already created into the backyard of a region. By combining fast-paced competition with tangible rewards, we hoped to incentivize mass participation—filling the blank city map quickly with accurate POI data.

So far, that strategy has paid off. Players built a passionate POI-marking community, despite global events that put a damper on business activity. (We encouraged players to follow local regulations and only play StreetCred if it was safe to take a solo walk). In the first two weeks of competition, StreetCred players in Jakarta created 7,994 places.

Creates overview
Figure 1. Overview of places created by the Jakarta StreetCred community.

City-wide coverage

Within two weeks, we’ve mapped a huge expanse of the city. Players have created places in 340 hexagon game tiles, an area accounting for nearly 40% of the Special Capital Region. The average player has made moves in eight tiles. By mapping across the entire city, these players are helping us get a big-picture view of where POIs are located.

Creates gif
Figure 2. Players have contributed data in 340 tiles throughout the city, with nearly 300 places in the most place-dense tiles.

Bringing tiles into focus

While some players spread out over a large area, others add value by focusing on particular neighborhoods. The average player makes 13 moves per tile. We’ve seen individuals map as many as 110 places in a single hexagon, going over these areas with a fine-toothed comb.

Completeness histogram
Figure 3. Distribution of tile mapping completeness. Tiles at 100% have all places mapped, and lower completeness scores illustrate how many places are left to create.

These highly active tiles allow us to get a better understanding of how many places exist—including places that have yet to be mapped. As players contribute multiple data points in a tile, we are able to model how many places exist in the tile and how completely it has been mapped. We currently have 47 extensively mapped areas in Jakarta. Within these areas, our analysis suggests that 81% of the places that exist have been mapped. As the community continues to play, we’ll be able to estimate these unmapped properties in an increasingly large part of the city.

Completeness gif
Figure 4. Tiles are slowly filled out over the first half of the contest period, with significant additions made in the last days of the most recent contest.

Competition is just getting started

The community is only just getting started. In the first weekly contest, players made an impressive 3,429 moves; by the second week, they were up to 13,892. We don’t necessarily expect 11x increases every week, but the evidence suggests that competition increases substantially with each new awarding of Bitcoin prizes. At this pace, Jakarta players able to rapidly cover the entire map, while also continuing to enrich and update dynamic data in lockstep with the city.

Moves over time
Figure 5. Moves per day versus time.

What can we take away from our Jakarta launch? Solving the toughest problems in mapping is a problem for a crowd. Given the right incentives, individuals can build a thriving mapping community. Crowdsourcing also allows for continuous changes, creating an always-accurate map—no experience in cartography required.