How to Play HowMany — City Population Game Guide
May 2026
HowMany is Capitalle's daily population estimation game. Each day you are shown a satellite image of a city and asked to guess how many people live there. After each guess you receive a simple higher or lower signal telling you whether the true population is above or below your estimate. The game rewards players who can read urban geography from aerial imagery and who have a strong intuitive feel for global population scales.
The Basic Rules
The game presents a satellite or aerial photograph of a city. You cannot see the city's name or any labels. Your task is to type a population number — how many people you think live in this city — and submit it. The feedback is binary: either the real population is higher than your guess, or it is lower. Based on this feedback, you refine your estimate and guess again. The game ends when you are within a certain percentage range of the correct population, or when you identify the correct population precisely.
Multiple rounds are available each day, and your score is based on how quickly you converge on the correct number. Players who use a systematic binary search strategy — starting with a midpoint estimate and halving or doubling based on feedback — tend to perform better than those who rely purely on intuition, though urban imagery reading skills help narrow your starting estimate significantly.
Reading City Density and Urban Sprawl from Satellite Images
The satellite image is your primary source of information before you make your first guess. Learning to read urban geography from aerial images is a skill that develops quickly with practice. The key visual indicators are building density, street grid pattern, the presence of large infrastructure, and the total area of continuous urban development visible in the frame.
Very dense cities — places like Mumbai, Dhaka, Manila, or Cairo — show almost no empty space at street level. Buildings are packed together, streets are narrow, and there is little or no visible greenery within the urban core. These dense city centers are associated with populations in the millions. By contrast, a city with wide streets, large parking lots, abundant green space, and low-rise buildings is likely an American or Australian city with a much larger footprint per capita — and the population implied by that sprawling footprint might actually be lower than the dense-looking city in Asia.
Grid patterns in street layouts provide geographic clues too. Rigid rectangular grids are characteristic of North American cities planned in the 19th century. Irregular, organic street patterns suggest older European cities or cities that grew without central planning. Radial patterns emanating from a central point are common in French-influenced urban planning. These patterns help you identify the continent and cultural context of the city, which tells you something about both the likely population density and the relationship between the city's visual size and its actual population.
Common Mistakes — City vs Metro Population
The most frequent source of error in HowMany is confusion between a city's proper population and its metropolitan area population. These two numbers can differ enormously. The city of Paris, within its official boundaries (the périphérique ring road), has a population of about 2.1 million. But the Greater Paris metropolitan area has a population of roughly 12 million. Similarly, London proper has about 9 million people, but the wider London commuter belt extends to 13-15 million.
HowMany specifies which definition it uses — typically the city proper or urban agglomeration — and it is important to read that specification carefully. When a satellite image shows a very large urban area with multiple dense clusters, it might be showing a metropolitan area including suburbs and satellite cities, in which case a higher population estimate is appropriate. When it shows a single compact city center, even if very dense, the population figure being asked about may be the city proper figure only.
Asian megacities are particularly prone to this confusion. Shanghai's city proper population is about 25 million, while its metropolitan area is even larger. Tokyo's city proper is about 14 million but the greater metropolitan area is 37-38 million, making it the largest metropolitan area in the world. Knowing these benchmarks helps you calibrate your estimates when you recognise a city from a broad Asian megacity perspective.
Benchmark Cities to Anchor Your Estimates
One of the most useful skills in HowMany is having a mental library of reference populations. Learn the populations of a handful of cities at each scale and use them as anchors when estimating. A city that looks about the same size as Lisbon or Edinburgh is probably in the 500,000 to 800,000 range. A city that looks like it has roughly the same footprint and density as Madrid or Rome is probably in the 3-4 million range for the city proper. A city that has the continuous density and scale of Jakarta or Beijing is in the 10 million-plus territory.
These anchor references work best when you also account for regional density norms. A city in Bangladesh or Egypt of the same satellite-visible footprint as a city in Canada or Australia will have dramatically more people, because South Asian and North African urban density is much higher per square kilometre than North American or Oceanian density. Always adjust your estimate upward for South and Southeast Asia, and downward for North America and Australia, relative to what the satellite image's footprint alone would suggest.
Applying Binary Search to Your Guesses
Once you have made your initial estimate, use a systematic approach to converge on the correct answer. If your guess of 1 million comes back as "higher," your next guess should not be 1.5 million — it should jump to something like 5 million, which splits the remaining possibility space more meaningfully. Think in orders of magnitude: populations range from under 10,000 to over 30 million, and you want each guess to eliminate as much of that range as possible.
This binary search approach — always guessing roughly in the middle of the remaining possibility range — is mathematically optimal for minimizing the number of guesses needed to find any target value. Combined with good initial estimation from the satellite image, it lets you converge on accurate population figures very efficiently.
Combining HowMany with Ranke makes for an excellent daily practice pair. Ranke trains your comparative population knowledge across countries, while HowMany trains your absolute population estimation from visual cues. Together they build a comprehensive understanding of global demographic scale that is useful in trivia contexts far beyond these games.