Loss Chasing, Near-Misses, and the Design of Persistent Play in Electronic Gambling and Loot-Box Systems

Abstract

Electronic gambling machines, loot boxes, and gacha reward systems do more than deliver random outcomes; they package uncertainty in sound, light, timing, visual suspense, partial payouts, and a standing invitation to play again. This paper takes up two mechanisms that live inside that packaging and tend to amplify one another: loss chasing and the near-miss effect. Loss chasing is the continuation or intensification of play after losses, typically driven by the goal of recovering money already spent. A near-miss is a losing outcome dressed up to look like a close success. Near-misses carry no monetary value in a game of pure chance, yet they have been shown to raise the motivation to keep playing and to recruit reward-related neural circuitry usually associated with winning (Clark et al., 2009). Drawing on the behavioral, cognitive, and clinical literatures on near-misses, losses disguised as wins, variable-ratio reinforcement, gambling-related cognitive distortions, virtual currencies, and loot-box monetization, the paper then develops a formal model of session persistence and runs an illustrative simulation comparing four conditions: ordinary losses, near-miss-heavy play, losses disguised as wins, and a combined high-risk design. The results favor a structural reading of persistence. Continued play is not reducible to individual weakness or statistical naivety; it emerges from the interaction between a player’s vulnerability and the way a game is built.

Keywords: loss chasing; near miss; electronic gambling machines; loot boxes; gacha; gambling disorder; losses disguised as wins; cognitive distortions; responsible gambling; product safety.

1. Introduction

We all recognize the scene. A player sits in front of a screen, gaze locked, finger moving almost on its own from one spin or purchase to the next. From the outside, the conduct can look senseless: money is draining away, the odds are understood to be unfavorable, and still the play continues. The standard explanations—it’s entertainment, the player lacks discipline, the player doesn’t grasp probability—each capture a fragment of the behavior while missing the machinery that generates it.

Electronic gaming environments are organized around repeated contact with uncertainty, and the player is reacting to far more than the tally of wins and losses. The relevant stimuli are a stream of cues: reels decelerating, symbols drifting into near-alignment, music swelling ahead of a result, a trivial payout dressed as a triumph, an account balance recoded as gems or credits, and the next attempt sitting one tap away. Gambling behavior responds not only to the monetary result but to the form in which that result arrives.

Two mechanisms do most of the work. The first, loss chasing, is the impulse to keep gambling after losses in order to win back what has gone. The second, the near-miss effect, is the experience of a loss as something that came close to a win. In a genuine skill task the near-miss is informative—a tennis serve that clips just outside the line tells the server to adjust by a hair. On an electronic gambling machine or in a loot-box draw, the same inference is almost always wrong. Proximity is an illusion: the next outcome is drawn from the identical probability structure no matter how close the previous animation appeared to land.

My argument is that loss chasing turns most corrosive when near-misses, celebrated net losses, and frictionless payment run together. If that is right, the locus of the problem shifts. It does not sit only inside the player; a good part of it sits in the design itself.

2. Literature Review

2.1 Variable-ratio reinforcement and repeated play

Variable-ratio reinforcement is the behavioral bedrock of electronic gambling. On this schedule, reward arrives after an unpredictable number of responses, and the resulting behavior is famously durable because any given failure can be read as the response that almost reached a payout. Gambling translates the schedule into a simple internal refrain: the last spin lost, but the next one could win.

Machines and loot-box interfaces sharpen that refrain by raising the event rate. A table game is paced by conversation, dealing, chip handling, and the gaps between decisions. A slot or a box draw collapses stake, suspense, result, and re-entry into a few seconds, so the player is not merely gambling but cycling through uncertainty at speed.

The speed is what lets small reinforcements pile up. An occasional win, a near-miss, a fraction of the stake returned, the chime of coins, the spectacle of an almost-jackpot—each feeds the reinforcement environment, and together they can supply enough stimulation to sustain play even as the balance erodes.

2.2 Near-misses as motivational events

A near-miss is an objectively losing outcome that wears the costume of a win. On a slot it might be two jackpot symbols on the payline with the third sitting one position high; in a loot-box reveal it might be an animation that flirts with a rare drop before settling on something ordinary. Either way the result is a loss, framed as nearness.

Its force comes from the word “almost.” Clark and colleagues (2009) reported that near-misses felt unpleasant relative to wins and yet pushed up the motivation to continue while engaging brain circuitry tied to reward. The wrinkle is worth dwelling on: near-misses do not generally feel good. More often they irritate. But irritation can be its own engine, producing the sense that one more go is warranted. Where probability sees a loss, the player reads a signal.

2.3 Losses disguised as wins

The second mechanism is the loss disguised as a win (LDW), which appears when a player gets back less than was wagered while the machine throws a celebration. Stake five dollars, receive two back, and the screen erupts in flashing light and triumphant sound. The ledger records a three-dollar loss; the senses record a win.

Work on multiline slots has found that LDWs can generate arousal closer to the response to wins than to the response to plain losses (Dixon et al., 2010). The feedback system, in effect, teaches the player how to score the session—and if net losses keep getting dressed as victories, the player’s read on how the night is going drifts away from the balance sheet. LDWs also thicken the event stream. A run of quiet losses feels nothing like a run of net losses studded with fanfare; the balance may fall identically in both, but the second feels busier, more hopeful, more rewarding.

2.4 Cognitive distortions and the misreading of chance

Gambling-related cognitive distortions are well documented. The best known is the gambler’s fallacy—the conviction that independent random events correct themselves over a short run, so that after a losing streak a win is somehow “due.” The belief is false for independent outcomes, yet it feels natural, since people expect randomness to even out where they can see it.

The illusion of control runs alongside it. Let a player stop the reels, pick the box, time the swipe, or choose among visually distinct options, and the outcome starts to feel partly authored by the player’s own hand. Langer (1975) showed that skill-like cues nudge people toward overestimating their control over chance, and contemporary gambling and gaming interfaces are stocked with exactly those cues.

Loss aversion and the sunk-cost trap then tighten the grip. A loss weighs more heavily than an equivalent gain pleases, so once a player is down, walking away means ratifying the loss as permanent, whereas continuing keeps alive the slim statistical hope of repair. By that stage the motive has often changed. The player is no longer chasing pleasure but trying to undo the discomfort the gambling itself produced.

2.5 Loss chasing and gambling disorder

Few markers of harmful gambling are as legible as loss chasing. Returning after losses to “get even” sits among the DSM-5 criteria for gambling disorder (American Psychiatric Association, 2022), and it matters clinically because chasing rewrites the emotional purpose of play. The session stops being recreation and becomes corrective, defensive, urgent.

It helps to separate two varieties. Within-session chasing keeps the player at the machine while behind, often with faster play or larger stakes meant to recover before the session ends. Between-session chasing brings the player back later—days or weeks on—with recovery as the explicit aim. Both do damage, but within-session chasing is the more design-sensitive of the two, because rapid play, near-misses, and one-tap continuation can all act on the player in real time.

2.6 Loot boxes, gacha systems, and virtual currency

The same architecture has migrated out of the casino. Loot boxes and gacha systems routinely combine real-money payment, randomized rewards, scarce prizes, unlimited buying opportunities, and animated suspense. Zendle and Cairns (2018) found loot-box spending to be associated with problem-gambling severity. That association does not on its own establish that loot boxes cause gambling disorder, but it does point to gambling-style monetization holding particular appeal for vulnerable users.

Virtual currency compounds the worry. Convert dollars into gems, coins, crystals, or points and the act of spending drifts out of plain financial language: fifty dollars becomes five thousand gems, and the real cost loses its edge while repeat purchases get easier to justify. Casino chips, ticket-in/ticket-out systems, and account credits do comparable work. None removes money from the transaction; each softens the player’s contact with it.

3. Conceptual Definitions

For clarity, the paper uses the following definitions.

Near-miss: A losing outcome that resembles a winning one closely enough to create a perception of proximity or almost-success.

Loss chasing: Continued or intensified gambling after losses, motivated by the wish to recover earlier losses.

Loss disguised as a win: A net-negative outcome presented through celebratory audiovisual feedback as though it were a success.

Payment friction: The psychological or practical resistance involved in spending. Cash carries relatively high friction; stored credit, virtual currency, autoplay, and one-click purchasing lower it.

Session persistence: The duration, number of plays, or amount wagered before voluntary stopping.

Structural risk: Harm potential created by the design of the gambling or gaming environment rather than by individual vulnerability alone.

4. Theoretical Framework

A player’s choice to go again after each spin, purchase, or reveal can be treated as a stopping problem. At every event the player weighs the recent outcome, the current emotional state, the size of the loss so far, the funds remaining, and how easy it is to continue. The weighing is far from fully rational; arousal, distorted interpretation, and the framing of outcomes all bend it.

A simplified continuation model:

P(Continue_t = 1) = σ( α
                        + β₁·NearMiss_t
                        + β₂·LDW_t
                        + β₃·CumulativeLoss_t
                        + β₄·LowFriction_t
                        + β₅·Vulnerability_i
                        + β₆·Arousal_t )

where σ is the logistic function.

The expected directions are mostly intuitive. Near-misses, LDWs, low payment friction, vulnerability, and arousal should all push the probability of continuing upward. Cumulative loss is the awkward term. In a purely rational model larger losses should raise the probability of stopping; in a chasing model they may instead raise continuation, because stopping forces the player to accept the loss as final.

Arousal can be carried as a state variable:

Arousal_t = ρ·Arousal_{t-1}
            + λ₁·NearMiss_t
            + λ₂·LDW_t
            + λ₃·Win_t

with 0 < ρ < 1. Emotionally charged events do not vanish on the next frame; a near-miss can bleed into the following decision and make immediate continuation more likely.

5. Hypotheses

The framework yields five hypotheses.

H1: Near-miss-enriched environments produce longer sessions than ordinary-loss environments.

H2: Environments with frequent losses disguised as wins produce longer sessions than ordinary-loss environments.

H3: A combined environment containing near-misses, LDWs, and low-friction payment produces the highest session persistence.

H4: Player vulnerability predicts persistence and chasing across all conditions.

H5: Payment friction moderates the loss–continuation relationship; under low friction, players are more likely to continue after losses.

6. Illustrative Simulation Study

6.1 Purpose

Commercial behavioral data from electronic gambling machines and loot-box systems are hard to come by. Operators guard such records as proprietary, and platform-level logs rarely reach independent researchers. The paper therefore reports an illustrative simulation rather than a claim about any product on the market. The aim is modest and methodological: to show how structural design variables can be modeled, estimated, and compared, not to fix a harm rate to a particular game.

6.2 Sample generation

A synthetic sample of 3,200 players was generated and assigned at random to one of four conditions, 800 players each:

Every simulated player started with a bankroll of 100 units. Players in the combined condition could top up more easily, standing in for stored payment details, virtual currency, or rapid credit continuation. Individual vulnerability was drawn from a normal distribution and represented susceptibility to arousal, distorted interpretation, and chasing.

6.3 Stopping rule

Stopping probability after each event was generated from:

P(Stop_t = 1) = σ( γ₀
                   + γ₁·SpinCount_t
                   + γ₂·CumulativeLoss_t
                   − γ₃·NearMissCondition_i
                   − γ₄·LDWCondition_i
                   − γ₅·CombinedCondition_i
                   − γ₆·Vulnerability_i
                   − γ₇·Arousal_t )

The negative signs on the near-miss, LDW, combined-condition, vulnerability, and arousal terms mean those variables lower the probability of stopping.

6.4 Outcome variables

Four outcomes were recorded: the number of plays before stopping, the total amount wagered, the final net loss, and whether the player entered a chasing state. A chasing state was coded when the player continued substantially after crossing a loss threshold, escalated bet size following losses, or replenished funds after depletion.

7. Simulated Results

7.1 Descriptive results

The ordering runs as predicted. Ordinary losses produced the shortest sessions and the lowest chasing rate. Near-misses and LDWs each lifted persistence on their own. The combined condition produced the sharpest escalation, with average play length running to more than twice that of the ordinary-loss condition.

7.2 Regression model: session length

An ordinary least squares model was estimated with log session length as the dependent variable, using ordinary losses as the reference category.

In the simulation, the near-miss and LDW conditions each add roughly a quarter to session length. The combined condition does far more than the sum of its parts—consistent with structural risk that compounds once several persistence-enhancing features sit in the same product.

7.3 Logistic model: chasing behavior

A logistic regression followed, with chasing as the dependent variable.

The combined condition again dominates: players exposed to near-misses, LDWs, and low-friction continuation faced far higher odds of chasing than those in the ordinary-loss condition. Vulnerability carried a strong independent effect of its own.

8. Discussion

The results back a structural account of persistence. Near-misses and LDWs are not just garnish that makes a game more fun; they alter how a loss is experienced. A plain loss says money is gone. A near-miss says success was within reach. A celebrated net loss says something good happened, even as the balance slid downward—three different stories told over the same financial fact.

That difference is the crux. Players respond to the event as presented, not only to its accounting. A three-dollar loss wrapped in lights and sound is processed differently from a silent three-dollar loss; a losing spin that nearly bared the jackpot lodges in memory differently from a losing spin with no drama at all. In each case the design, not the math, sets the psychological meaning of the outcome.

Risk in the simulation is also cumulative. A near-miss feature lengthens play, LDWs lengthen play, low-friction payment lengthens play—and stacked together they produce something well beyond any single feature, which is precisely the configuration of modern digital systems, where speed, stored payment, virtual currency, animation, and reward scheduling can be fused without a seam.

A corollary is that return-to-player percentage cannot stand alone as a measure of harm risk. A game with a generous RTP may still be dangerous if it runs fast, fires events constantly, manufactures near-misses, celebrates net losses, and makes continuation effortless. A game with a leaner RTP but a slower pace and honest loss feedback may keep play shorter. Harm risk is a property of the whole environment, not of the house edge in isolation.

9. Clinical Implications

Loss chasing matters clinically because it signals a shift in what gambling is for. A recreational player gambles for excitement, company, or distraction; a chasing player gambles to undo damage. The money already lost will not stay in the past—it becomes psychologically live, a standing reason to keep going.

That is the source of the destructiveness. As losses mount the player’s mood sinks, yet the nearest tool for easing that mood is more of the same gambling, so the activity that inflicts the pain doubles as the escape from it. The loop mirrors other addictive patterns, where short-run relief props up long-run harm.

Cognitive Behavioral Therapy speaks directly to the distorted beliefs about randomness, control, and recovery that hold the loop together. A player may need to learn that a near-miss is not progress, that past losses do nothing to make a future win more likely, and that stopping does not turn irrational simply because money is already gone. Even so, treatment should not be asked to shoulder the whole load. When a product is engineered to intensify persistence, the clinician is working against the architecture of the game.

10. Regulatory and Product-Safety Implications

A serious regulatory stance has to reach past the narrow question of whether a product technically satisfies the legal definition of gambling. The more useful question is whether the product borrows gambling-like mechanisms to push spending, persistence, and impaired control. Several measures follow from the evidence.

First, net losses should read as losses. Stake five dollars, get two back, and the interface should not throw a celebration; the financial result belongs on the screen.

Second, near-miss frequencies should be auditable. Where near-misses are generated or amplified beyond what chance alone would produce, regulators should classify them as a structural risk feature.

Third, payment friction should be restored where it can be. Players should see the real-money value of every stake and purchase, and virtual currency should not be permitted to mask it.

Fourth, session interruptions should mean something. A pop-up that dismisses on a single tap is not a break; cool-down periods, time reminders, and pre-commitment tools should be built to interrupt the automatic next play.

Fifth, randomized monetization that minors can reach warrants stricter scrutiny. Loot boxes and gacha may not be casino gambling in every legal sense, but they can reproduce much of the same psychology, and wherever real money buys a chance-based reward, consumer protection deserves to be taken seriously.

11. Limitations

The empirical section is a simulation, not field data. It cannot show that any particular machine, casino, game, platform, or publisher produces the effects reported here; its job is to lay out a testable structure. A real study would need behavioral logs, controlled experiments, player-level spending records, and ideally longitudinal follow-up.

The model also flattens the player. Real players differ in income, age, gambling history, psychiatric vulnerability, alcohol use, social setting, cultural background, and prior exposure to these products. The simulation reduces vulnerability to a single trait, where the real thing is multidimensional.

And the paper leans on structural risk without dissolving individual agency. Players do choose, but they choose inside environments engineered to steer attention, emotion, and continuation, and any complete account has to hold both at once.

12. Conclusion

Loss chasing and near-misses go a long way toward explaining why players stay in situations where leaving would be the financially rational move. A near-miss recasts a loss as near-success; a loss disguised as a win recasts a net-negative result as a celebration; virtual currency and frictionless payment drain the force out of spending; and rapid play loops the whole sequence before reflection can get a word in.

What emerges is not entertainment so much as an engineered system for keeping people playing.

The lesson, then, is not that vulnerable players are short on willpower. It is that some game environments are constructed to make self-control hardest at exactly the moment losses are climbing. Safer design need not strip out uncertainty or fun. It would strip out misleading feedback, put losses back in plain view, thin out manufactured near-misses, lengthen the road from loss to re-stake, and keep the true price of play impossible to overlook.

References

American Psychiatric Association. (2022). Diagnostic and Statistical Manual of Mental Disorders (5th ed., text rev.).

Auer, M., & Griffiths, M. D. (2013). Voluntary limit setting and player choice in most intense online gamblers: An empirical study of gambling behaviour. Journal of Gambling Studies, 29(4), 647–660.

Clark, L., Lawrence, A. J., Astley-Jones, F., & Gray, N. (2009). Gambling near-misses enhance motivation to gamble and recruit win-related brain circuitry. Neuron, 61(3), 481–490.

Dixon, M. J., Harrigan, K. A., Sandhu, R., Collins, K., & Fugelsang, J. A. (2010). Losses disguised as wins in modern multi-line video slot machines. Addiction, 105(10), 1819–1824.

Goodie, A. S., & Fortune, E. E. (2013). Measuring cognitive distortions in pathological gambling: Review and meta-analyses. Psychology of Addictive Behaviors, 27(3), 730–743.

Gooding, P., & Tarrier, N. (2009). A systematic review and meta-analysis of cognitive-behavioural interventions to reduce problem gambling: Hedging our bets? Behaviour Research and Therapy, 47(7), 592–607.

Langer, E. J. (1975). The illusion of control. Journal of Personality and Social Psychology, 32(2), 311–328.

Schull, N. D. (2012). Addiction by Design: Machine Gambling in Las Vegas. Princeton University Press.

Tversky, A., & Kahneman, D. (1991). Loss aversion in riskless choice: A reference-dependent model. Quarterly Journal of Economics, 106(4), 1039–1061.

Zendle, D., & Cairns, P. (2018). Video game loot boxes are linked to problem gambling: Results of a large-scale survey. PLOS ONE, 13(11), e0206767.

Appendix: Reproducible Simulation Code

import numpy as np
import pandas as pd
import statsmodels.formula.api as smf

np.random.seed(42617)

conditions = [
    "Ordinary losses",
    "Near-miss enriched",
    "LDW enriched",
    "Combined high-risk condition"
]

def sigmoid(x):
    return 1 / (1 + np.exp(-x))

rows = []

for ci, condition in enumerate(conditions):
    for player in range(800):
        vulnerability = np.random.normal(0, 1)
        bankroll = 100.0
        starting_bankroll = 100.0
        topups = 0
        plays = 0
        arousal = 0.0
        cumulative_loss = 0.0
        total_wagered = 0.0
        chase = 0
        bet = 1.0

        while plays < 500:
            if bankroll <= 0:
                if condition == "Combined high-risk condition" and topups < 2:
                    bankroll += 100
                    topups += 1
                    chase = 1
                else:
                    break

            plays += 1

            if condition == "Ordinary losses":
                p_win, p_ldw, p_nm = 0.08, 0.00, 0.03
            elif condition == "Near-miss enriched":
                p_win, p_ldw, p_nm = 0.08, 0.00, 0.22
            elif condition == "LDW enriched":
                p_win, p_ldw, p_nm = 0.08, 0.26, 0.03
            else:
                p_win, p_ldw, p_nm = 0.08, 0.24, 0.20

            if cumulative_loss > 45:
                escalation_probability = sigmoid(
                    -3.0 + 0.7 * vulnerability + 0.7 * (ci == 3)
                )
                if np.random.rand() < escalation_probability:
                    bet = min(5.0, bet + 0.5)
                    chase = 1

            total_wagered += bet

            u = np.random.rand()
            near_miss = 0
            ldw = 0
            win = 0

            if u < p_win:
                win = 1
                net = bet * np.random.choice(
                    [2, 5, 10, 20],
                    p=[0.55, 0.25, 0.15, 0.05]
                )
            elif u < p_win + p_ldw:
                ldw = 1
                net = -bet * np.random.uniform(0.2, 0.8)
            else:
                net = -bet
                if u < p_win + p_ldw + p_nm:
                    near_miss = 1

            bankroll += net
            cumulative_loss = max(0, starting_bankroll + 100 * topups - bankroll)

            if cumulative_loss > 60:
                chase = 1

            arousal = (
                0.82 * arousal
                + 0.55 * near_miss
                + 0.42 * ldw
                + 0.35 * win
            )

            if plays >= 20:
                near_miss_condition = int(ci in [1, 3])
                ldw_condition = int(ci in [2, 3])
                combined_condition = int(ci == 3)

                stop_linear = (
                    -4.20
                    + 0.0045 * plays
                    + 0.0095 * cumulative_loss
                    - 0.55 * near_miss_condition
                    - 0.48 * ldw_condition
                    - 0.65 * combined_condition
                    - 0.45 * vulnerability
                    - 0.28 * arousal
                )

                stop_probability = sigmoid(stop_linear)

                if np.random.rand() < stop_probability:
                    break

        rows.append({
            "condition": condition,
            "vulnerability": vulnerability,
            "plays": plays,
            "log_plays": np.log(plays),
            "net_loss": starting_bankroll + 100 * topups - bankroll,
            "total_wagered": total_wagered,
            "topups": topups,
            "chase": chase
        })

df = pd.DataFrame(rows)

summary = df.groupby("condition").agg(
    n=("plays", "size"),
    mean_plays=("plays", "mean"),
    sd_plays=("plays", "std"),
    median_plays=("plays", "median"),
    chasing_rate=("chase", "mean"),
    mean_net_loss=("net_loss", "mean"),
    mean_wagered=("total_wagered", "mean")
)

ols = smf.ols(
    'log_plays ~ C(condition, Treatment(reference="Ordinary losses")) + vulnerability',
    data=df
).fit(cov_type="HC3")

logit = smf.logit(
    'chase ~ C(condition, Treatment(reference="Ordinary losses")) + vulnerability',
    data=df
).fit()

print(summary.round(3))
print(ols.summary())
print(logit.summary())