Research: Why Retail Loses
Why Retail Loses
Section titled “Why Retail Loses”The core finding: 70% of prediction market trading addresses lose money. Fewer than 30% earn net positive returns across 124 million trades (Reichenbach & Walther). This is not bad luck — it’s structural.
The same study shows that sophisticated users do fine: traders transacting $500K+ median ROI is +2.6%. Traders transacting under $100 median ROI is −26.8%. The platform is a wealth transfer machine from unsophisticated to sophisticated participants.
Post-election 2024: Polymarket MAU dropped 57% from January to August 2025. Volume held at $7.9B. The casual users left; power users stayed. Monthly retention: 35%. That outperforms 85%+ of crypto protocols, but it means 65% of acquired users are gone within 30 days.
The question: What causes this? And what would fix it?
The Seven Adoption Barriers
Section titled “The Seven Adoption Barriers”| # | Barrier | Severity | What’s actually happening |
|---|---|---|---|
| 1 | Loss aversion & churn | ★★★★★ | Users exit after first losing position. The pain of loss > pleasure of win at 2:1. No mechanism to recover emotionally or financially. |
| 2 | Capital lockup | ★★★★ | Winning positions sit locked until resolution — sometimes weeks. Users can’t redeploy capital. Power users in particular feel this acutely. |
| 3 | No skill recognition | ★★★★ | Good forecasters earn no credential. There’s no track record. Being right repeatedly has no compounding value — each market resets. |
| 4 | Market integrity doubt | ★★★½ | Resolution failures (Zelensky suit, Oscar viewership, Venezuela) create distrust. Users who’ve been wrongly resolved exit and don’t return. |
| 5 | Onboarding friction | ★★★ | Crypto wallet + fiat onramp + UI learning curve = 3 barriers before the first bet. Most casual users don’t clear all three. |
| 6 | Liquidity gaps | ★★★ | 63% of active Polymarket markets have zero 24-hour volume. 67.7% of all markets have lifecycle under 7 days. Thin markets = bad execution = losses. |
| 7 | Information asymmetry | ★★½ | Power traders and AI agents have information advantages. Retail users are providing liquidity to better-informed counterparties. |
The Bet Size Distribution
Section titled “The Bet Size Distribution”Prediction market participation is heavily right-skewed:
| Percentile | Typical bet size (Polymarket) |
|---|---|
| Median | $30–55 |
| P75 | $200–400 |
| P90 | $1,000–2,000 |
| P99 | $20,000–50,000 |
| Mean (2024 Trump contract) | $210 |
28% of all bets are $10–50. These small-bet users are the first to churn — the $30 loss has high emotional salience relative to their budget, and the $30 win has almost no measurable financial impact.
The protocol’s 1% power users generate 70% of volume. Retention of this cohort is the survival metric.
What Loss Aversion Actually Does
Section titled “What Loss Aversion Actually Does”Prospect theory predicts users weight losses 2× heavier than equivalent gains. For a prediction market:
- A user who wins a $100 position resolving at $1.00 feels moderate satisfaction
- A user who loses a $100 position to zero feels intense distress — approximately 2× the emotional magnitude
Standard prediction market design makes this worse:
- Binary resolution (win everything or lose everything)
- No mechanism to partially exit a losing position
- No insurance product to cap downside
- No recovery path — the loss is total and final
Churn after a losing experience is not irrational. It’s the rational response to a product that offers no loss mitigation.
TAM Impact of Solving These Barriers
Section titled “TAM Impact of Solving These Barriers”| Barrier fixed | Mechanism | Expected retention improvement |
|---|---|---|
| Loss aversion | REWIND insurance | 15–30% improvement in 30-day retention |
| Capital lockup | Fast-Forward early exit | 20–40% improvement in capital utilization / redeployment |
| No skill recognition | Chrono Score | 30–50% improvement in long-term retention (6–12 months) |
| Market integrity doubt | AI resolution + FREEZE dispute layer | 5–15% improvement in institutional participation |
| Onboarding friction | PRECOGNITION guided onboarding | 5–10% improvement in activation rate |
Conservative estimate: addressing the top 3 barriers improves total addressable market by 15–30%. This is comparable to the effect Compound/Aave had on DeFi lending by adding yield to idle capital — it unlocked a category of users who wouldn’t participate otherwise.
The Non-Obvious Implication
Section titled “The Non-Obvious Implication”The 70% loss rate is not just a user welfare problem. It’s a liquidity problem.
Prediction markets need a continuous supply of “incorrect” participants to function — someone has to be on the other side of the smart money. When retail users lose and churn, the market becomes dominated by sophisticated traders playing zero-sum games against each other. Volume concentrates in fewer markets, bid-ask spreads widen, and markets on long-tail topics disappear entirely.
Insurance doesn’t just protect retail users — it makes them willing to return to the market after losses. That’s the demand-side flywheel: more insured users → more diverse market participation → more liquid markets → better information → more user trust → more participation.
The protocol that solves retail retention doesn’t just earn insurance premiums. It earns the entire TAM expansion that follows.
Related:
- Rewind — the insurance mechanism
- Fast-Forward — the capital unlock mechanism
- Chrono Score — the skill recognition layer
- Market Microstructure — volume distribution and platform dynamics