The numbers behind Squid Game

Episode Summary

Executive Summary: This episode uses Squid Game to explore two kinds of numbers: the expected losses in a bridge-crossing game and Netflix’s controversial viewing metrics. Matt Parker explains the bridge puzzle as a simple 50/50 expectation problem, while Julia Alexander critiques Netflix’s “two minutes watched” definition of a view and places Squid Game’s 140 million accounts in context.

Main Topics: Probability in Squid Game’s bridge game (Priority: 5/5): A listener question asks how many players should survive a bridge-crossing contest. Matt Parker reframes it as an 18-step, 50/50-life-loss puzzle and explains the expected number of lives lost. Expected value versus actual outcomes (Priority: 5/5): The discussion distinguishes average expected loss from real-world results, noting that the expected nine lives lost does not mean every run ends with exactly nine losses. Distribution of outcomes in the bridge puzzle (Priority: 4/5): The hosts explore how likely different outcomes are, including the chance of exactly nine survivors, fewer than nine, more than nine, or losing all 16 lives. Netflix’s 140 million account claim (Priority: 5/5): The episode examines Netflix’s headline number for Squid Game’s popularity and questions what it actually measures. How Netflix defines a 'view' (Priority: 5/5): Julia Alexander explains that Netflix counts a title as viewed after two minutes, contrasting this with older 70%/90% completion-style measures and arguing it can inflate impressions of popularity. Limits of streaming metrics (Priority: 4/5): The segment concludes that hours watched and account counts still do not directly reveal the number of actual viewers because household size and shared accounts vary widely.

Key Arguments: The bridge game is mathematically straightforward: with 18 steps and a 50% chance of losing a life at each step, the expected number of lives lost is 18 × 0.5 = 9. Expected value is an average over many repetitions; it does not predict a single playthrough exactly. The chance of losing all 16 lives on the 18-step path is extremely small, about 0.06%, because you would need to fail on nearly every step. Squid Game’s reported 140 million figure is not a count of full viewers; it reflects Netflix accounts that watched at least two minutes. Netflix changed its reporting method from roughly 70% of a title watched to a two-minute 'starter' metric, making comparisons with earlier metrics difficult. Even total hours watched cannot fully identify unique viewers because one account may represent one person or many people. Comparisons with box-office-style metrics are misleading for streaming platforms because there is no one-to-one relationship between viewing and revenue or audience size.

Data Points: Contestants in Squid Game: 456 - Number of financially desperate players in the drama Prize money: 45.6 billion Korean won / 38.6 million US dollars - Reward for the winner of Squid Game Netflix accounts watching Squid Game: more than 140 million - Netflix’s reported popularity metric for the show Bridge puzzle path length: 18 steps - Analogous probability puzzle used to avoid spoilers Lives in puzzle: 16 lives - Starting lives in Matt Parker’s simplified version Probability per step: 50% - Each step has a coin-flip chance of costing a life Expected lives lost: 9 - 18 steps × 0.5 expected loss per step Chance of no life losses across 18 steps: 1 in 262,144 (about 0.0004%) - Probability of getting the safe outcome 18 times in a row Chance of exactly nine people over: 18.5% - Probability of losing exactly nine lives in the puzzle Chance of fewer than nine survive: 40.7% - One side of the distribution around the mean Chance of more than nine survive: 40.7% - Other side of the distribution around the mean Chance of exactly losing 10 lives: 16.7% - Equivalent to having exactly eight safe choices in the original analogy Chance of losing all 16 lives: about 0.06% - Probability of total failure on the 18-step path Older Netflix completion threshold: about 70% of a title - Netflix’s pre-2020 reporting method Netflix viewing tiers: starter / watcher / completer - Julia Alexander’s explanation of Netflix’s internal categories Hours streamed for Squid Game in leaked reports: 1.4 billion hours - Referenced as a figure that could be broken down, though still not equivalent to unique viewers

Pivotal Quotes: "If you've got to do 18 steps, and each time there's a 50-50 chance that you will lose a life or not, then the expected return ... it's just 18 times a half. It's nine." — Matt Parker: Explaining the expected number of lives lost in the bridge-crossing probability puzzle "They report two minutes of an episode has been watched at least by more than 140 million accounts." — Julia Alexander: Describing how Netflix’s Squid Game viewing figure is calculated "Two minutes is not watching a show." — Host/Tim Harford: Critiquing Netflix’s definition of a view

Implications: Listeners should treat streaming “view” claims cautiously: Netflix’s headline numbers can overstate true audience size, and even hours watched don’t equal unique viewers. The episode also shows how expected-value math can simplify seemingly complex probability problems.