How CCTV Games Work

CCTV games are a category of prediction-based betting products built on live surveillance footage and AI-powered object detection. Instead of spinning reels or drawing cards, you predict how many real-world objects - cars, pedestrians, ducks, skiers - will pass through a defined detection zone during a fixed time window. The outcome is determined by what actually happens in the physical world, counted by machine learning models in real time.

This guide covers everything: how rounds are structured, how the AI works, what the bet types mean, and how the camera network operates.

The Format

CCTV games were developed by 155.io, a Swedish gaming studio. The product launched in January 2026 with its first title, Rush Hour, a vehicle-counting game set across major global cities. The concept is straightforward: take real surveillance infrastructure, apply computer vision to count objects with precision, and build a betting market around the count.

The studio's CEO, Sam Jones, described the concept as "Think Big Brother blended with Polymarket" - a phrase that captures both the surveillance aesthetic and the prediction-market structure at the core of each round.

155.io distributes its games through Hub88, an aggregation platform that connects game studios to online casino operators. This means CCTV games are available at Roobet, Stake, Shuffle, and any other Hub88-integrated platform that has elected to carry the content.

Step-by-Step Round Walkthrough

Every round in a CCTV game follows the same three-phase structure: the betting window, the counting phase, and settlement. Understanding each phase is essential before placing a bet.

Phase 1 - The Betting Window

The betting window opens at the start of each round, typically lasting 15 to 30 seconds depending on the title. During this window, you can see the live camera feed and observe current conditions - traffic density, pedestrian volume, weather - before committing to a bet. The active camera location and current round parameters are displayed on screen.

Odds are fixed at round open and do not change during the betting window, regardless of what the camera is showing. This means you can observe conditions without worrying about odds moving against you as you decide.

Bet types and their payouts are available simultaneously: Over, Under, Range, and Exact. You can place multiple bet types in a single round if the platform permits parlays on the same event, though this varies by operator.

Phase 2 - The Counting Phase

Once the betting window closes, the counting phase begins. The camera feed remains live and visible. An AI model - running computer vision on the video stream - draws bounding boxes around objects as they cross the detection zone. Each confirmed crossing increments the round counter, which is displayed in real time.

The counting phase typically lasts 60 seconds. Throughout this window, you can watch the count build in real time. Objects that enter the detection zone but do not fully cross it are not counted. Objects counted include only those that pass through the defined zone in the designated direction (where directional filtering is active).

The visual experience during the counting phase is a key differentiator of the format. Unlike a slot spin that resolves in 2-3 seconds, you are watching a real event unfold over a full minute, with a running count visible the entire time.

Phase 3 - Settlement

At the end of the counting window, the final count is locked. Settlement occurs immediately and automatically. Winning bets are credited to your account balance. The next round's betting window opens within seconds.

Settlement is deterministic: the AI's final count is the outcome. There is no human adjudication. If the AI counted 7 vehicles and your Range bet covered 5-8, you win. If it counted 12, you lose. The count displayed during the round is the count used for settlement - there is no post-round adjustment.

Camera Locations and Rotation System

Rush Hour, the flagship title, operates across a network of city cameras spanning multiple continents and time zones. The rotation system ensures that regardless of when you play, at least some locations are experiencing active traffic conditions.

Current camera cities in the Rush Hour network include:

Tokyo, Japan
New York, United States
London, United Kingdom
Bangkok, Thailand
Paris, France
Sydney, Australia

Additional cities are integrated on a rolling basis as 155.io expands its camera partnerships. Each camera is positioned to capture a high-traffic zone - typically an intersection, a bridge crossing, or a pedestrian thoroughfare - where consistent object volume makes for meaningful rounds.

The rotation system cycles between cameras at defined intervals. Some platforms display the upcoming camera so you can anticipate conditions before the betting window opens. Local time at each camera location is relevant: a Tokyo camera during rush hour will produce different counts than the same camera at 3am local time.

For Duck River, cameras are positioned at river locations where waterfowl movement is predictable but not uniform - creating natural variance around a baseline. For Snow Run, mountain resort infrastructure provides the camera positions.

Detection Zone Mechanics

The detection zone is a defined area within the camera frame. Only objects that fully cross this zone are counted. This is the most important technical concept to understand, because partial crossings, objects that stop before the zone, and objects that cross outside the zone boundary all result in zero contribution to the count.

What Counts

Objects that enter the detection zone boundary and fully exit the other side
Objects detected with confidence above the model's threshold (typically 85-90%)
Objects of the designated type for that game (vehicles in Rush Hour, waterfowl in Duck River, skiers in Snow Run)

What Does Not Count

Objects that enter the zone but stop or reverse before exiting
Objects that pass adjacent to the zone without crossing the boundary
Objects of the wrong category (a cyclist in a vehicle-only zone, for example)
Objects below the confidence threshold - these are flagged as uncertain and excluded
Objects partially occluded to the point where the model cannot confirm category

The detection zone is visible as an overlay on the camera feed during the counting phase. This transparency is intentional - you can see exactly what area the AI is monitoring, and you can observe in real time whether objects are crossing or not.

AI Confidence Scores Explained

The AI model running object detection produces a confidence score for every detection event. A score of 94% means the model assigns 94% probability that the detected object is of the correct category - a vehicle, a duck, a skier - and that it has correctly located its bounding box.

Confidence Thresholds

155.io's models operate with a minimum confidence threshold before a detection is counted. Objects detected below this threshold are discarded. This creates a small category of edge cases: objects that are partially visible, heavily occluded, or ambiguously categorized may be excluded even if a human observer would count them.

In practice, the threshold is calibrated to minimize both false positives (counting something that is not there) and false negatives (missing something that is there). The tradeoff is that rare legitimate crossings are occasionally excluded. This is reflected in the RTP and payout structure - see RTP and house edge explained for detail.

Bounding Boxes

Bounding boxes are the rectangular overlays drawn around each detected object. They are visible on the live feed during the counting phase. A box that appears and then crosses the detection zone line contributes to the count. Watching bounding box behavior is one way to develop intuition for how the model behaves at specific camera locations.

Edge cases with bounding boxes include:

Box merging: two adjacent objects detected as one, resulting in one count instead of two
Box splitting: one object momentarily tracked as two, though duplicate-detection logic prevents double-counting in most cases
Box dropout: the model briefly loses track of an object mid-crossing; depending on implementation, this may or may not complete the count

Bet Types Deep Dive

Every CCTV game offers four core bet types. Understanding the probability and payout of each is essential for informed play.

Bet Type	Payout Multiplier	Approximate Probability	Description
Over	3.60x	~25%	Final count exceeds the set threshold
Under	3.00x	~30%	Final count falls below the set threshold
Range	2.25x	~40%	Final count falls within a specified range
Exact	18.00x	~5%	Final count matches an exact number

The Over Bet

The Over bet pays 3.60x on a win. You are predicting that the object count will exceed a threshold set by the platform for that round. At a roughly 25% win probability, this is a lower-frequency but higher-payout option. Example: the threshold is set at 10 vehicles. You bet Over. The AI counts 14 vehicles. You win 3.60x your stake.

The Over bet performs best when you observe dense, fast-moving conditions during the betting window - heavy traffic, crowded crossings, high activity on the slope.

The Under Bet

The Under bet pays 3.00x and wins at approximately 30% frequency. You are predicting a count below the set threshold. This is the natural counterpart to Over, and its slightly lower payout reflects its marginally higher win probability. Example: threshold is 10 vehicles. You bet Under. The AI counts 7 vehicles. You win 3.00x your stake.

The Range Bet

The Range bet pays 2.25x and wins approximately 40% of the time. You select a range of acceptable counts - for example, 5-9 vehicles - and win if the final count falls within it. This is the most conservative of the four bet types, offering the most frequent wins at the cost of lower payout per win.

Range bets suit conditions where the activity level is readable but not extreme - not clearly very high or very low, but somewhere in the middle band. Example: you observe moderate traffic and predict 5-9 vehicles. The AI counts 7. You win 2.25x your stake.

The Exact Bet

The Exact bet pays 18.00x and wins approximately 5% of the time. You predict an exact count. This is the highest-variance, highest-payout option and carries the highest house edge. Example: you predict exactly 6 vehicles. The AI counts 6. You win 18.00x your stake.

The Exact bet is best understood as a longshot, not a regular strategy. Its 18x payout is attractive, but the 5% win rate and elevated house edge mean expected value is lower than Over or Under. See the RTP guide for the full breakdown.

Count Distribution Patterns

Object counts are not uniformly distributed. They follow patterns shaped by real-world conditions: time of day, day of week, weather, local events, and the specific camera location. Understanding distribution tendencies is one way to bring observation into your betting decisions.

In general, Rush Hour city cameras show:

Higher counts during local peak hours (morning and evening commute windows)
Lower counts in overnight and early morning periods
Predictable low-count periods on weekends at business-district cameras
Higher variance in cameras positioned at arterial routes versus side streets

Counts vary significantly by location. A Tokyo intersection during rush hour will produce a very different distribution than the same hour in Sydney. Counts at any given camera also show round-to-round variance - even during peak periods, individual 60-second windows can produce counts significantly above or below the session average.

None of this constitutes a reliable edge. The house margin is built into the odds regardless of observable conditions. But observational awareness of baseline activity levels at a given camera can inform which bet type is appropriate - Range during moderate conditions, Over or Under during clearly extreme conditions.

Multi-Camera Format for Snow Run

Snow Run introduces a multi-angle format not present in Rush Hour or Duck River. Each round uses three simultaneous feeds:

POV camera: mounted at the base of the slope, facing upward toward incoming skiers
Drone camera: aerial overhead view of the full slope segment
Slope-side camera: fixed position at the mid-slope crossing point, the primary detection zone

All three feeds are visible simultaneously, but the count is derived from the slope-side camera's detection zone. The POV and drone feeds provide context - you can see skiers approaching and assess density before they reach the counting zone. This creates a brief informational window during the betting phase that is unique to Snow Run.

The multi-camera approach also enhances the spectator experience, making Snow Run particularly suited to streaming and shared viewing.

History and Distribution

155.io was founded in Sweden and developed CCTV games as a novel format for the online gaming market. The studio's background is in computer vision and real-time data processing - the gaming application emerged from an existing capability in AI-based video analysis.

Rush Hour launched in January 2026 as the first commercial title. Duck River and Snow Run followed as the studio expanded its object categories beyond vehicle counting. The distribution model relies on Hub88, which aggregates game content for online casino operators. Hub88's network includes Roobet, Stake, and Shuffle as anchor platforms, with additional operators integrating the content on an ongoing basis.

Find the full list of platforms carrying CCTV games at where to play.

How CCTV Games Compare to Other Live Game Formats

Feature	CCTV Games	Live Dealer (Blackjack/Roulette)	Virtual Sports	Crash Games
Outcome source	Real-world events	Physical cards/ball	RNG-animated simulation	RNG multiplier
Round length	~75-90 seconds	30-120 seconds	60-180 seconds	10-60 seconds
Observational input	Yes - live conditions visible	Minimal	None	None
Verifiability	AI count visible in real time	Physical process visible	Not verifiable	Provably fair hash
Shared experience	Yes - same feed, same round	Yes - shared table	No	Yes - shared multiplier
Streamability	High	Medium	Low	High

The key distinction of CCTV games versus virtual sports is real-world grounding. Virtual sports use RNG to animate simulated events. CCTV games use actual surveillance footage and actual events - the vehicles on the screen are real vehicles, the count is a real measurement, not a number generated by an algorithm.

Compared to live dealer games, CCTV games offer faster round pacing and a more visually distinct format. The camera network rotation means every round looks different. Live dealer games, by contrast, involve a fixed studio environment with predictable visual elements.

Frequently Asked Questions

Can the AI count be manipulated?

The AI model runs on 155.io's infrastructure and produces counts from live footage. The process is automated with no human intervention. 155.io publishes information about the detection architecture, and the bounding box overlay visible during rounds allows you to observe exactly what is being counted. Platform operators rely on the same count outputs - there is no mechanism for post-hoc adjustment.

Are all players watching the same round?

Yes. All players at any operator carrying the same CCTV game title see the same camera feed and the same count in real time. The round is a shared event - there is no individualized feed. This is one reason CCTV games perform well in streaming contexts.

Is there a best time to play?

Camera activity varies by local time at each location. You can observe current conditions during the betting window. There is no time of day that reliably produces better expected value - the house edge is embedded in the odds and is constant. But if your preferred bet type is Over, playing when cameras are showing high-activity periods increases the likelihood of high counts, which may align with your preference.

Is there a winning strategy?

There is no strategy that overcomes the house edge. Observation of camera conditions can inform bet type selection (Range when conditions are moderate, Over/Under when conditions are clearly extreme), but this is not a guarantee of improved outcomes. Every round carries an independent probability distribution. See RTP and house edge explained for the mathematical framework.

What happens if the camera feed drops mid-round?

Feed interruptions are handled by the platform's terms of service. In most implementations, a round where the feed is interrupted for more than a defined duration is voided and stakes returned. Check the specific rules at your platform of choice.

Are more titles coming?

155.io has indicated an expanding content roadmap beyond the initial three titles. The camera network can support new game types wherever consistent, countable objects are present. Additional urban, natural, and sporting environments are all plausible directions for new titles.

Next Steps

Review the individual title guides to understand game-specific mechanics: Rush Hour, Duck River, Snow Run
Read the RTP and house edge guide to understand expected value before placing bets
Find the platforms currently offering CCTV games at where to play
Start with Range bets to familiarise yourself with round pacing and count behaviour before moving to higher-variance bet types

---

How CCTV Games Work