Can You Actually Make Money on Prediction Markets? The Honest Math

Yes, some people make money trading prediction markets — and the same arithmetic that makes their profits possible guarantees most people can’t. Every event contract has two sides: at settlement, one side’s dollar is the other side’s loss. Add trading fees and the game is zero-sum before costs, negative-sum after. Profit means being more right than the trader on the other side, by more than the fees both of you paid. That bar is a number, computable for every price on the board — and this guide computes it: the edge that clears the fees, where documented edges have existed, why most traders lose anyway, and what the few who don’t do differently. Education, not advice.

Why is prediction market trading zero-sum?

A prediction market contract pays $1 to the correct side and $0 to the wrong one; nothing else ever enters the pot. Stocks are positive-sum — companies earn profits, so investors collectively can gain. An event contract earns nothing, pays nothing, builds nothing. Summed across every trader, profit and loss before fees is exactly zero — after fees, negative by exactly the fees. The average dollar traded on Kalshi or Polymarket must lose. That’s accounting, not pessimism.

The fees, as of August 2026: Kalshi charges takers 0.07 × contracts × price × (1 − price), rounded up to the cent — about 7% of expected profit, peaking at 1.75¢ per contract at 50¢ — while resting maker orders trade free in most markets (a few high-volume series charge makers a quarter rate). Polymarket uses the same price × (1 − price) shape with taker rates from 0.04 to 0.07 by category — politics at 0.04, crypto at 0.07, geopolitics at zero — and makers pay nothing. Full schedules live in the Kalshi fee breakdown and the Polymarket fee breakdown.

So where can profit come from in a negative-sum game? Three places. Traders with worse information or discipline — the main source, shrinking as markets professionalize. Hedgers, who accept negative expected value because they’re buying insurance; their losses fund speculators’ wins, though hedging remains a small share of volume. And platform incentive programs — real but temporary. A durable trading profit traces to one of those three, or it doesn’t exist.

How much edge does it take to break even?

A taker who buys Yes at price P and holds to resolution needs the true probability to exceed P plus the fee:

Price paid (taker)	Kalshi fee (7%)	Breakeven probability	Polymarket politics (4%)
10¢	0.63¢	10.63%	10.36%
30¢	1.47¢	31.47%	30.84%
50¢	1.75¢	51.75%	51.00%
70¢	1.47¢	71.47%	70.84%
90¢	0.63¢	90.63%	90.36%

Read the 50¢ row out loud: a coin-flip-priced contract has to be at least a 51.75% proposition before buying it earns an expected cent on Kalshi. And that understates the bar, because the quoted price isn’t the market’s estimate — the midpoint between bid and ask is, as covered in how to read a prediction market price. Crossing a 1¢ spread costs another half-cent against the mid. Call it two-plus points of genuine disagreement with the market before a taker trade at 50¢ is even neutral.

The extremes are worse than they look. A 4¢ longshot carries a fee near 0.27¢ and breaks even around 4.3% — the event must be about 7% likelier, in relative terms, than the price says. A 96¢ favorite breaks even at 96.3%, and the payoff is lopsided: about 3.7¢ per winning contract against 96¢ per loser, so one blown favorite erases roughly twenty-six correct ones — for an under-4% return that may take months.

Now the number that quietly ends most trading careers: the round trip. Buying 200 contracts at 50¢ (a $100 position) as a taker costs about $3.50 in fees; exiting through a 1¢ spread costs another $3.50 in fees plus $2 to the spread. That’s roughly $9 — 9% of the position — to end where the trader started. Churn a bankroll through ten such round trips with no edge and about 60% of it is gone to transaction costs alone, without a single bad call.

The market doesn’t pay for being right. It pays for being more right than the trader on the other side — after both of them paid the fees.

Where have real edges actually existed?

The case that profit is possible rests on documented inefficiencies, not vibes. Three stand up to scrutiny.

The favorite-longshot bias. Across decades of betting and prediction market data, low-priced contracts have won less often than their prices implied, and heavy favorites slightly more often (Snowberg and Wolfers, NBER). A 4¢ contract has historically been a worse-than-4% proposition. The catch: per-contract profit from fading longshots is tiny, fees and spread eat much of it, one upset erases a long run of wins, and capital sits locked until resolution.

Slow repricing at long horizons. Markets months from resolution, with low attention and thin books, have historically absorbed information more slowly than markets near settlement — the calibration record, and where markets beat polls, is in the EventMarkets accuracy analysis. The edge is being early where nobody is watching; the catch is months of locked-up money and books too thin to enter or exit at size.

Cross-platform gaps. An academic review of roughly $2.4 billion in 2024 election trading across Kalshi, Polymarket, PredictIt, and the Iowa Electronic Markets (Clinton and Huang, 2025) found identical contracts trading at different prices across exchanges, with arbitrage opportunities peaking in the final two weeks. Capturing a gap requires funded accounts on both venues, fees on both legs, and — the part that turns “risk-free” into “risky” — resolution terms that actually match, which nominally identical markets often don’t.

The pattern: every documented edge is small, capacity-constrained, operationally annoying, and shrinking as sharper money arrives. Anything advertised as large, easy, and permanent is something else.

Why do most traders lose?

Start structurally: in a negative-sum game, the median participant must lose. Adjacent evidence says active retail loses faster — in the classic study of 66,465 brokerage accounts (Barber and Odean, Journal of Finance), the most active traders trailed the market by 6.5 points a year, mostly through trading costs. Prediction markets compress the dynamic: costs are proportionally larger, and there is no rising tide because the sum is zero.

The specific behaviors that transfer money from retail traders to everyone else:

• Paying the toll every time. Market orders cross the spread and pay taker fees both ways — the 9% round trip — and the other side of that toll is often a resting-order professional paying nothing.
• Overtrading. Reacting to every headline multiplies round trips. The fee formula doesn’t care that each trade felt justified.
• Correlated bets that look diversified. Five positions that all hinge on the same inflation print or the same candidate’s polling are one position at five times the size, and they fail together.
• Buying lottery tickets. The longshot bias means the most popular retail trade — cheap Yes contracts on exciting outcomes — is precisely the systematically overpriced one.
• Grading luck as skill. A 60% contract fails four times in ten. Separating a real 55% hit rate from coin-flipping takes hundreds of resolved trades; almost nobody keeps records that long before concluding they’re gifted.

What does being good actually look like?

Strip away the mystique and profitable trading is a clerical discipline. Five habits recur among traders who plausibly sustain it — observation, not prescription.

Calibration tracking. A probability estimate written down before every trade and scored against outcomes. Without a few hundred scored forecasts, “I’m good at this” is a feeling, not a finding.

Fee-adjusted breakeven math. The table above, applied before entry: if the estimate doesn’t clear the breakeven with room, there is no trade, however strong the opinion.

Bankroll sizing. The Kelly criterion caps rational stakes: a genuine five-point edge at 50¢ justifies roughly 7% of bankroll after fees, and practitioners commonly bet half that. Positions sharing a driver count as one. Staking half a roll on one market answers a math question with adrenaline.

Market selection. Edges live where attention doesn’t: specific, unglamorous questions a trader independently models — not the headline market everyone is staring at, where the other side is sharpest.

Cost engineering. Resting limit orders instead of market orders, fewer decisions instead of churn, attention to fee tiers by category. Zero maker fees aren’t free money — makers eat adverse selection, filled fastest when the other side knows something — but for patient traders they remove the largest drag.

The honest bottom line

Most people who trade prediction markets should expect to lose modest amounts of money, slowly, to fees and better-informed counterparties — that is what a negative-sum market delivers to its average participant, and no strategy article changes it. The documented edges are real, small, laborious, and increasingly contested. Meanwhile, the one thing these markets give away free is what they’re best at: calibrated probabilities, readable by anyone, no position required. Any profits that do materialize are taxable. None of this is financial advice; it’s the math to check anyone’s promises against — including ours.

Quick answers

Can you actually make money on Polymarket or Kalshi? Some traders do, consistently. But event contracts are zero-sum before fees and negative-sum after: sustained gains require other traders’ sustained losses plus the fee bill. Most participants lose.

Is prediction market trading profitable for most people? No. The average dollar traded must lose after fees, and the typical retail pattern — taker orders, frequent trading, longshot buying — loses faster than average. Round-tripping a 50¢ position as a taker costs roughly 9% in fees and spread.

What edge does it take to beat the fees? As of August 2026, a Kalshi taker buying at 50¢ breaks even at a true probability of 51.75% under the 7% fee formula; Polymarket’s 0.04 politics rate puts it at 51%, with category rates from zero (geopolitics) to 0.07 (crypto). Formulas and examples: the Kalshi fees guide and Polymarket fees guide.

What’s the best Kalshi trading strategy? There is no list of winning picks — anything published that widely stops working on publication. The documented approaches are structural — fading the favorite-longshot bias, early entry in low-attention markets, cross-platform gaps — each small, fee-sensitive, and capacity-limited. Execution discipline (maker orders, fee math, calibration records, Kelly-bounded sizing) has mattered more than any pick.

Is trading prediction markets gambling or investing? Economically it’s zero-sum speculation with a genuine information-aggregation function — closer to poker than index investing, since profit comes from other participants rather than economic growth. The legal classification is actively contested; the arguments are laid out in gambling or investing.

Do prediction market profits get taxed? Yes — winnings are taxable income, with platform-specific reporting and some genuinely unsettled IRS questions, covered in the prediction market taxes guide.