The challenge
Supporting multiple regulated products while keeping up with a changing fraud landscape
Underdog runs fantasy sports contests and prediction markets in the same user interface, and the two don’t share the same map. A state may allow one and not the other, and those rules change as regulations evolve. Every entry, across every product and state, depends on confirming where the player actually is at that moment.
The volume behind those checks can be significant. During the Super Bowl, Underdog peaked at roughly 400,000 active users in a single hour.
As its footprint grew, Underdog needed a geo-compliance approach that could adapt quickly across products and jurisdictions. Location verification needed to accurately determine a player’s location without creating unnecessary friction for legitimate users. That was especially important near state borders and in dense cities, where small inaccuracies can cause a player to fail a check even when they’re physically located in an approved jurisdiction.
New-customer promotions introduced another challenge. Fraudsters can create large numbers of accounts across different identities, devices, and IP addresses, making each account appear legitimate when viewed on its own.
Traditional identity checks can help determine whether an individual account looks legitimate, but they aren’t designed to show when hundreds of seemingly unrelated accounts appear within minutes of each other from the same small area. Underdog needed a way to identify those relationships earlier and act before promotions were claimed and cashed out.
“Fantasy sports and prediction markets don’t share the same regulatory map, but we run both in the same app. As we expanded into prediction markets and into more states, we needed geo-compliance that could flex product by product and stay accurate near state borders, where small inaccuracies cost you legitimate players.”
—Casi Bogacz, Fraud Operations Manager
The solution
Modern, flexible geo-compliance built for multiple products and jurisdictions
Radar runs location verification across Underdog’s fantasy and prediction markets footprint on iOS, Android, and web. Each Underdog product brings its own unique jurisdictional considerations. Radar’s technological approach allows the same integration to return a different answer for prediction markets than it does for fantasy without requiring a separate build, enabling Underdog to bring new states online within days.
The teams also worked together to improve accuracy near state borders. After Underdog found that some players near state lines were failing checks they should have passed, Radar changed how buffer zones are evaluated when a player’s location overlaps two permitted states.
“The same integration gives us a different answer for prediction markets than it does for fantasy, without a separate build. That’s what lets us bring new states online in days instead of weeks. It made a real difference near state borders too, where we used to lose legitimate players to inaccuracies that had nothing to do with where they actually were.”
—Casi Bogacz, Fraud Operations Manager
Radar’s location and device signals also gave Underdog a new way to investigate coordinated fraud.
Underdog’s team began seeing short bursts of high-density location checks concentrated in small areas, well beyond what ordinary traffic would produce. Many different users were reporting identical latitude and longitude coordinates, sometimes pointing to open fields, intersections, or the geographic center of a city rather than a plausible device location.
Looking across other risk signals revealed more connections. A high proportion of the traffic came through residential proxies. Some devices reported time zones that didn’t match their locations. Browser inconsistencies and device integrity signals pointed to emulators rather than real handsets.
Any one of those signals could have an innocent explanation. Together, across a dense cluster of accounts, they gave Underdog a much stronger indication that the activity was coordinated.
"Individually, a lot of what we were seeing could have had an innocent explanation. What changed is that we could look across a cluster of accounts at once and see the pattern, instead of one account at a time. That’s the difference between catching a ring early and reviewing hundreds of accounts after the fact."
—Casi Bogacz, Fraud Operations Manager
Working with Radar’s fraud operations team, Underdog turned those patterns into rules that combine signals like high velocity, never-before-seen accounts, implausible coordinates, and flagged residential proxies. The rules run at the first location check, before identity verification, so suspicious activity can be stopped before promotions are claimed and cashed out.
Reports monitor for the same patterns over time, and identified hotspots can be geofenced directly. This enables the Underdog team to investigate and act on coordinated groups instead of reviewing each account in isolation.
The investigation helped identify a synthetic identity ring farming new-customer promotions using device emulators, spoofed locations, and residential proxies to imitate legitimate account creation at scale.
Radar now sits alongside Underdog’s existing identity and monitoring vendors as the location and device layer, feeding into the decisions the team already makes rather than adding another system to manage.
The results
Fewer legitimate players blocked and coordinated fraud caught earlier
With Radar, Underdog has a more flexible geo-compliance foundation that supports multiple products, adapts quickly as new states come online, and provides location and device signals the team can use to uncover coordinated fraud.
Pass rates near state borders have improved by 34.3%, helping more legitimate players access products they’re eligible to use. Radar also helps Underdog block an average of 2.1K risky devices each month.
The results:
- 34.3% improvement in pass rates near state borders
- 2.1K risky devices blocked per month on average
Next, Underdog is feeding confirmed fraud cases back into Radar so risk scoring can learn from patterns its team has already verified. The team is also moving to act on fraud rules through webhooks, allowing flagged accounts to be suspended automatically instead of waiting in a review queue.
“We operate two different regulatory models and three different licenses in the predictions space, and we rely on Radar to stay compliant across all of them.”
—Sam Baker, SVP Product
