- Order the criteria first: jurisdiction → market availability → real total cost → resolution handling → funding. Fees are fourth, not first, despite dominating most comparisons.
- Polymarket: deepest liquidity on large event markets, 2% on net winnings only, non-custodial USDC settlement. Does not serve US users.
- Kalshi: CFTC-regulated, US-accessible, fiat settlement, per-side volume fees. Fewer markets, shallower book on exotics.
- Manifold / Metaculus: play-money forecasting venues. Excellent for calibration, not for capital-weighted signals.
- PredictIt: historically expensive and capped; verify current terms before considering it.
1. Order the Criteria Before the Platforms
Most "best prediction market" articles start with a fee table. That is the wrong starting point, because fees are rarely the binding constraint. Work through these in order and the platform choice usually makes itself:
| # | Criterion | Why it comes at this position |
|---|---|---|
| 1 | Jurisdiction | If you cannot legally use it, no other metric matters. This eliminates platforms before any comparison begins. |
| 2 | Market availability & liquidity | A cheap platform with no liquidity in your markets is worse than an expensive one with a deep book. |
| 3 | Real total cost per trade | Spread plus fee. Posted fees alone routinely mislead by a factor of two or more. |
| 4 | Resolution handling | How disputes are resolved determines whether your edge survives contact with reality. |
| 5 | Funding & withdrawal | Affects friction and cost, but is usually solvable — unlike the four above. |
The reordering matters. Two traders comparing the same fee tables will reach opposite conclusions if one is a US resident and the other is not, because criterion one already decided it. Any comparison that does not state its jurisdictional assumption is not usable.
2. The Scored Comparison
Scores are relative assessments across the platforms discussed, not absolute quality judgements. Higher is better within each row.
| Criterion | Polymarket | Kalshi | Manifold | Metaculus | PredictIt |
|---|---|---|---|---|---|
| Liquidity (large events) | Excellent | Good | Low | N/A | Moderate |
| Liquidity (niche markets) | Moderate | Moderate | Low | N/A | Low |
| Fee structure (winners) | 2% net winnings | Per-side volume | None | None | High |
| Fee on losing trades | $0 | Charged per side | N/A | N/A | Charged |
| US access | Not served | Yes | Yes | Yes | Restricted |
| Non-US access | Broad | Limited | Yes | Yes | Limited |
| Settlement asset | USDC (Polygon) | USD (fiat) | Play money | Play money | USD |
| Custody model | Non-custodial | Custodial | N/A | N/A | Custodial |
| Market breadth | Very broad | Broad (regulated list) | Community-created | Curated | Narrow, capped |
| Best for | Liquidity + low net fees | US regulated access | Forecasting practice | Calibration scoring | Legacy use |
Read this table as conditional. Every "excellent" is scoped — Polymarket's liquidity advantage is concentrated in large event markets, not uniformly across all of them. Kalshi's access advantage only helps if you are in a jurisdiction it serves. Terms change; verify current specifics on each platform before acting.
3. Polymarket — Where It Wins
Polymarket's advantages are structural, and three of them matter a great deal.
Liquidity depth on major events
On high-profile markets — elections, major crypto thresholds, significant geopolitical events — Polymarket's order books are deep enough that a mid-sized position can be entered and exited without moving the price materially. That depth is the single most valuable property a prediction market can have, because it directly reduces your cost. Detail in liquidity and spread.
Fee structure that rewards being right
The 2%-on-net-winnings model means losing trades cost nothing in platform fees. For a trader who wins most positions, that structure is materially cheaper than per-side volume fees, which are charged regardless of outcome. Full breakdown in fees explained.
Non-custodial settlement
Your funds sit in your own wallet. There is no exchange balance sheet, no withdrawal queue, and no account freeze in the custodial sense. This removes the largest systemic risk in crypto, though it shifts responsibility for security onto you — discussed at length in is Polymarket safe.
The trade-offs: Polymarket does not serve US-based users, so criterion one eliminates it for that audience. Niche markets can be thin, and the wide market breadth means quality varies enormously between markets. Thin, ambiguous, or novel markets are where most losses occur, and their existence is a byproduct of having such broad coverage.
4. Kalshi — Where It Wins
Kalshi's advantage is regulatory legitimacy, and for a US resident that advantage is decisive.
| Property | Why it matters |
|---|---|
| CFTC-regulated exchange | Operates inside a US regulatory framework rather than around it |
| US residents can trade | Accessible where Polymarket is not — usually the deciding factor |
| Fiat settlement | Funds via bank transfer, no stablecoin or wallet required |
| Curated market list | Fewer ambiguous, exotic markets — less resolution risk by construction |
| Predictable fee model | Volume-based fees are easy to forecast before entering |
The trade-off is breadth and depth. A curated regulated list means fewer markets, and outside the headline ones the order book can be thinner than Polymarket's. For a US-based trader comparing a deep Kalshi book against a Polymarket book they cannot legally access, the decision is straightforward. This is covered in more detail in our Polymarket vs Kalshi comparison.
5. Manifold & Metaculus — Play-Money Value
These platforms are frequently dismissed in trading-focused comparisons, and that is a mistake — they serve a genuinely different and useful purpose.
| Platform | What it does well | What it cannot do |
|---|---|---|
| Manifold | Community-created markets, fast experimentation, zero financial risk | Cannot produce capital-weighted signals — no money at stake |
| Metaculus | Scored forecasting over long horizons, excellent for calibration training | Not designed for trading; no profit mechanism |
The underrated use: use play-money platforms to build calibration and only then move to real money. Tracking your probability estimates against outcomes over dozens of questions is the fastest way to learn whether your judgement is genuinely good or merely confident. That skill transfers directly to real-money trading — see probability and odds.
6. Real Cost Per Trade, Not Posted Fees
The posted fee schedule is only one component of your cost. This is where platform comparisons most often mislead.
| Cost component | How to measure it | Typical magnitude |
|---|---|---|
| Platform fee | Read the schedule for your platform | 0% on losses to ~2–7% |
| Bid-ask spread | Compare best bid to best ask at your intended size | 0.5–15%+ — usually the largest |
| Slippage | Difference between expected and filled price on market orders | Larger with size, worse in thin markets |
| Funding / withdrawal | Cost to get money in and out | Small but non-zero |
| Gas | Network cost per transaction on-chain | Negligible on Polygon |
The critical insight: a platform advertising a 2% fee with deep liquidity can be far cheaper in total than one advertising 0% with a 6% spread. Compare net cost per completed trade, at the size you actually trade, in the markets you actually use. Anything else is comparing marketing copy.
7. Resolution Quality as a Differentiator
This is the most under-compared criterion and, in practice, one of the most consequential.
| Dimension | What good looks like | What poor looks like |
|---|---|---|
| Criteria clarity | Unambiguous trigger, named source, explicit deadline | Headline-only questions with undefined sources |
| Dispute process | Documented path, defined timeframe, published reasoning | Opaque, slow, or effectively unappealable |
| Precedent | Similar markets resolved consistently over time | Same wording resolved differently on different occasions |
| Market curation | Ambiguous ideas rejected before listing | Anything listed, ambiguity left to traders |
A platform with excellent liquidity and sloppy resolution will cost you more over time than one with a thinner book and rigorous criteria, because ambiguity losses are total and uninsurable. See resolution explained for how disputes actually arise.
8. Choosing for Your Situation
| Your situation | Reasonable first choice | Reasoning |
|---|---|---|
| Outside the US, crypto-native, want liquidity | Polymarket | Deepest books on major events, 2% on winnings only, non-custodial |
| US resident, wants regulated access | Kalshi | The jurisdiction constraint outranks every other criterion |
| Learning to forecast, no capital at risk | Metaculus / Manifold | Build calibration first; it transfers directly |
| Wants niche, unusual markets | Compare case by case | Thin liquidity and ambiguous criteria are the norm here — verify each market individually |
| Wants the lowest possible cost | Measure, do not assume | Spread usually dominates fee; compute net cost at your size |
Five checks before you commit capital
- Confirm your legal position. Verify the rules where you actually live. This page is not legal advice.
- Check liquidity in your specific markets, not the platform's headline volume.
- Measure the spread at your intended size before pricing any edge.
- Read a resolution dispute from that platform, if any is public. It tells you more than any feature list.
- Test the full funding and withdrawal path with a small amount before committing meaningfully.
Frequently Asked Questions
What is the best prediction market?
There is no universal answer, because the binding constraint differs by person. Polymarket leads on liquidity and net fee structure for non-US crypto-native traders. Kalshi leads on US regulated access. Manifold and Metaculus serve forecasting and calibration rather than trading. Rank jurisdiction, market availability, net cost, resolution handling and funding in that order — the platform choice usually follows.
Is Kalshi better than Polymarket?
They optimise for different constraints. Kalshi is CFTC-regulated, accessible to US residents, settles in fiat, and charges per-side volume fees. Polymarket is non-custodial, settles in USDC on Polygon, charges 2% on net winnings only, and offers deeper liquidity on large event markets but does not serve US users. Jurisdiction decides this question before any fee comparison does.
Which prediction market has the lowest fees?
For a trader who wins most positions, Polymarket's structure is typically cheapest because losing trades incur no platform fee. But posted fees are not total cost — the bid-ask spread usually dominates and can be five to ten times the platform fee. Compare net cost per completed trade at your actual size, not advertised schedules.
Do I need identity verification?
Regulated venues such as Kalshi require identity verification because they operate under financial regulation and settle in fiat. Non-custodial platforms do not have a traditional account to verify in the same way, since you transact through your own wallet — though jurisdictional access restrictions still apply. Confirm your own legal position rather than assuming access equals permission.
Are play-money prediction markets worth using?
Yes, for calibration. Metaculus scores forecasters over time and Manifold supports rapid experimentation, both without financial risk. What they lack is capital-weighting, since nobody loses money for being wrong. Use them to build and measure your forecasting skill, then apply it where real money makes the aggregate price meaningful.
What is the most important thing to check before choosing?
Your jurisdiction, first and decisively. If you cannot legally use a platform, its liquidity and fees are irrelevant. After that, check whether the markets you actually want to trade have sufficient liquidity there, then measure real net cost per trade, then examine how resolution disputes are handled. Funding logistics come last because they are usually solvable.
Can I use more than one prediction market platform?
Yes, and there are legitimate reasons to — different platforms list different markets, and pricing can diverge between them. That also creates arbitrage opportunities, though you must account for fees, spreads, funding costs, and settlement differences on both sides before assuming a gap is tradeable. Our arbitrage guide covers the mechanics.