What the House Is

The House Architecture has four primary components, each independently advantageous and collectively decisive over any realistic time horizon.

The House Architecture is not unique to casinos. It is a general design pattern for systems that extract value from participant engagement. The specific implementation varies by context, but the structural logic is identical.

The House Architecture does not merely exploit mathematical asymmetry. It exploits a stack of cognitive biases that make players systematically miscalculate their actual position. The cognitive exploitation stack is documented in behavioral economics, neuroscience, and the internal design documents of gaming companies (several of which have been subject to regulatory disclosure or litigation discovery).

The gambler's fallacy produces the belief that past losses make future wins more likely — that a streak of losses "must" end. The hot hand fallacy produces the belief that a winning streak is likely to continue. Loss aversion produces asymmetric responses to losses and gains that make players hold losing positions longer and exit winning positions earlier than expected value would suggest. Sunk cost reasoning keeps players in losing sessions because "I've already lost so much, I might as well try to get it back." The near-miss effect produces continued play after a near-win. The availability heuristic makes the memorable large win dominate the calculation while routine small losses are forgotten.

"Time on device" originated as a slot machine design metric and became the organizing principle of social media platform design. The term was adopted explicitly by Facebook engineers and product managers, as documented in Frances Haugen's disclosure of internal Facebook research. The metric is structurally identical in both contexts: more time means more bets (in gambling) or more ad impressions (in social media), and the House Architecture is optimized accordingly.

The implications of time-on-device optimization are identical in structure across contexts. In gambling: more time means more exposure to the house edge, which means greater expected loss. In social media: more time means more exposure to algorithmically selected content optimized for engagement rather than wellbeing, which means greater expected cognitive, social, and emotional cost. The "game" in both cases is designed not to be won by the player, but to be played — as long as possible, as frequently as possible, with as little friction to continued engagement as possible.

The House Architecture's structural inevitability rests on a mathematical foundation that is explored in detail in the next paper: the Ergodicity Trap. The standard expected value calculation — which appears to show that games with small house edges are near-fair — is a statistical artifact that applies to a hypothetical player playing an infinite number of independent games simultaneously. No real player is in this position. Real players play sequentially, with finite resources, and in non-ergodic conditions where a run of losses can permanently end the game regardless of what the expected value calculation says about the long run.

The House Architecture is designed for non-ergodic conditions. The house has effectively infinite resources and can sustain any run of player wins; the player cannot sustain an equivalent run of losses. The structure guarantees that, given sufficient time, variance alone will ruin every player — even players who have positive expected value in individual bets — because the variance will eventually produce a loss sequence large enough to eliminate the player's stake before the long-run average can assert itself.