Inside Kalshi's Crypto Binary Options: $60M/Day in Five-Minute Bets

Last Updated: March 23, 2026

Update — March 23, 2026: Polymarket announced fees on all market categories today, with crypto taker rates peaking at 1.80% (effective March 30). The fee comparison sections below have been updated. The “free Polymarket” era is over — fees now eat the edge on both platforms.

Kalshi’s crypto binary options have become the fastest-growing product in prediction markets. The platform now processes over $60 million in daily crypto volume, with five-minute contracts accounting for the majority of activity. We analyzed 877,606 settled contracts and 72.1 million individual trades from the Becker (2026) dataset to understand how these markets actually work, who trades them, and where the inefficiencies live.

Key Findings:

• Five-minute contracts drive 73% of volume — reviving a product from $2K/event lows to $439K/event medians
• TWAP settlement reduces volatility by 0.554x — most pricing models miss this, overestimating move probabilities
• Market makers earn +1.12% while takers lose -1.12% — NO-side depth is 40x deeper than YES
• The order book has a structural retail optimism tax — contracts at 30 cents settle YES only 26-28% of the time
• Model-based edge is +1.2 cents per signal — real but thin, with 68% of contracts showing zero exit liquidity
• Both Kalshi and Polymarket now charge explicit fees — the “free” era is over

How Did Kalshi Crypto Volume Get to $60M/Day?

Kalshi’s crypto volume trajectory tells a story of hype, hangover, and product-driven recovery across 18 months. The path from launch to $60 million in daily volume was neither linear nor inevitable — it required a fundamental product rethink after the initial excitement faded.

The Launch Boom (October-November 2024)

When Kalshi launched crypto range contracts in late 2024, median event volume ranged from $28,000 to $82,000. The early adopters were a mix of crypto-native traders curious about a regulated alternative to offshore platforms and traditional options traders intrigued by the binary format. BTC contracts dominated, accounting for roughly 78% of early volume. ETH represented another 18%, with DOGE, XRP, and SOL splitting the remaining 4%.

Hourly contracts attracted 62% of the launch-phase volume, as traders gravitated toward faster feedback compared to daily expiry. Strike ladders were narrower then — roughly 120 strikes per event versus the 188 we see today — and spreads were wide, often 4-6 cents on mid-range contracts.

The Hangover (December 2024 - January 2025)

By early 2025, median event volume had collapsed to approximately $2,000. Three factors drove the decline. First, the novelty premium evaporated as early adopters discovered the fee drag on short-term speculation. Second, BTC volatility dropped 34% from its Q4 2024 levels, shrinking the perceived opportunity in range contracts. Third, market makers hadn’t yet built the infrastructure to keep tight quotes on 188 strikes simultaneously, so retail traders faced 5-8 cent spreads that made casual trading unattractive.

This period was brutal for Kalshi’s crypto ambitions. The $2,000 median meant many individual events attracted fewer than 50 trades total. Order books were sparse, with resting liquidity concentrated on only 8-12 of the 188 available strikes per event. The product looked like it might join the long list of financial products that launched with fanfare and died quietly.

The Five-Minute Revolution (Q2 2025 - Present)

The recovery was driven by a single product innovation: five-minute contracts. Launched in mid-2025, these ultra-short-duration contracts changed the economics of participation. By March 2026, volume had recovered to $439,000 median per event, with daily aggregate throughput exceeding $60 million.

Five-minute contracts succeeded because they aligned the product with how crypto actually trades. Volatility in a 5-minute BTC window is roughly $150-300 at current price levels. With $100 strike spacing, this means 2-4 strikes are competitive on any given event — enough for a liquid market but not so many that liquidity fragments across an unmanageable ladder. The number of 5-minute events per day (288) also multiplied the available contract supply, giving market makers a vastly larger surface area to deploy capital.

Our analysis of the volume recovery shows that 73% of current daily volume comes from 5-minute contracts, 19% from hourly, and 8% from daily. The 5-minute product didn’t cannibalize existing flow — hourly and daily volumes have held steady while 5-minute contracts added an entirely new layer of activity. This pattern mirrors what we documented in our weather convergence research, where shorter-duration contracts also attracted incremental liquidity rather than redistributing existing volume.

How Do Kalshi’s Crypto Range Contracts Actually Work?

Kalshi offers binary contracts on whether a cryptocurrency’s price will finish above or below a specific strike price at the end of a defined time window. A contract pays $1 if the condition is met, $0 if it doesn’t. You can buy YES (betting the condition will be met) or NO (betting against it).

The product lineup spans three time horizons: five-minute contracts (the most popular), hourly contracts, and daily contracts. Each time window produces a strike ladder of roughly 188 contracts per event, spanning from deep out-of-the-money to deep in-the-money. For a BTC hourly event, strikes are spaced $100 apart. ETH uses $40 spacing.

The critical detail most traders miss is how settlement works. Kalshi doesn’t use a spot price snapshot. Instead, CF Benchmarks calculates a TWAP (time-weighted average price) by sampling 60 one-second readings during the final minute of each contract window. Then it applies a 20% trimmed mean, discarding the highest 12 and lowest 12 readings.

How Does TWAP Settlement Actually Work in Practice?

The TWAP mechanism sounds abstract, so here’s a concrete example. Suppose BTC is hovering around $95,100 as the 2:59 PM settlement window opens. CF Benchmarks begins sampling at 2:59:00 and records one reading per second through 2:59:59, collecting 60 individual price observations.

In a typical minute, BTC might oscillate within a $130 range. The 60 readings might span from $95,050 to $95,180, with most clustered around $95,100-$95,120. The trimming step discards the 12 highest readings (say, $95,155 through $95,180) and the 12 lowest ($95,050 through $95,078). The remaining 36 readings are averaged, producing a settlement price of, say, $95,112.

The contract with the bracket containing $95,112 settles YES at $1.00. All other contracts settle at $0. If a trader held a YES contract on the “$95,100 to $95,200” bracket, they collect $1. If they held YES on “$95,200 to $95,300,” they get zero.

The trimming is specifically designed to defeat manipulation. A trader who dumps $5 million of spot BTC at 2:59:58 to push the price below a key strike only affects 2 of 60 readings — and those 2 readings fall within the 12 that get discarded. The manipulator spends real money moving the spot market but gains nothing on the binary contract. This is why the trimmed mean reduces effective volatility by a factor of 0.554 compared to a single-point settlement.

This TWAP adjustment has a concrete mathematical effect on model pricing. The variance reduction factor comes from the formula R(N) = (N+1)(2N+1) / (6N^2) for N=60 samples, combined with the trimming. Our TWAP-adjusted Black-Scholes model accounts for this with a linear transition: full adjustment below 1 minute to expiry, no adjustment beyond 5 minutes. Models that ignore this adjustment systematically overestimate the probability of large moves near expiry — a mistake we documented extensively in our backtest writeup.

For traders accustomed to traditional options, the key difference is the all-or-nothing payoff. There’s no delta hedging. A contract at 65 cents doesn’t mean you’re long 0.65 deltas of the underlying — it means you’ve paid 65 cents for a binary bet that pays $1 or $0.

The Fee Structure

Kalshi’s fee formula is deceptively simple: ceil(0.07 * P * (1-P) * 100) cents per contract for takers. Maker fees are 25% of taker fees.

The P*(1-P) structure creates a parabolic fee curve that peaks at 50/50 prices (1.75 cents taker) and falls toward zero at extreme prices. This seems fair until you realize that out-of-the-money contracts already have thin margins. A contract at 5 cents has a taker fee of roughly 0.34 cents — that’s 6.8% of the contract price. A contract at 50 cents pays 1.75 cents, just 3.5% of the price.

This fee asymmetry systematically disadvantages OTM traders, who are disproportionately retail participants making directional bets. The fee structure effectively subsidizes market-making activity at the expense of speculative flow.

As of March 30, 2026, Polymarket charges explicit taker fees on all market categories. Crypto markets carry the highest rate: fee = C × p × 0.072 × (p × (1-p)), peaking at 1.80% for 50/50 contracts. Makers pay nothing and receive a 20% rebate from the taker fee pool, distributed daily in USDC. The dynamic formula means both platforms now penalize mid-range contracts most heavily — where retail traders concentrate their activity. Our fee calculator can help you compare the all-in cost across platforms for specific trade sizes.

Who Actually Trades Kalshi Crypto?

The participant mix on Kalshi crypto has evolved substantially since launch. Based on order book analysis and the Becker (2026) trade-level dataset of 72.1 million trades, we can identify four distinct participant classes, each with measurably different behavior, P&L characteristics, and order book footprints.

Retail Directional Traders

Retail directional traders represent the largest group by headcount but a minority of volume. These participants typically buy YES contracts on short-term crypto moves, often at prices between 20-50 cents. Our data shows a pronounced favorite-longshot bias: retail traders systematically overpay for low-probability outcomes, particularly on 5-minute contracts where the action is fastest.

The average retail YES buyer trades 3-8 contracts per event, with a median position size of $14. Repeat trading frequency is low — roughly 68% of retail accounts trade fewer than 10 events per month. This pattern suggests most retail participants treat Kalshi crypto as entertainment-adjacent rather than as a serious trading venue. Their loss rate tracks the structural -1.12% taker disadvantage documented by Becker, plus the additional drag from the favorite-longshot bias — combining for an estimated -2.5% to -3.5% per-event return.

Algorithmic Market Makers

Algorithmic market makers dominate the order book. The NO side of Kalshi’s crypto books is consistently 40x deeper than the YES side. To put that asymmetry in concrete terms: a typical BTC 5-minute contract at the 45-cent strike might show 12 contracts bid on the YES side at 44 cents, versus 480 contracts offered on the NO side at 56 cents.

This creates a structural environment where YES buyers face immediate fills but wide spreads, NO buyers get better prices but may need to wait, and large YES orders move the market more than equivalent NO orders. The asymmetry is most pronounced on 5-minute contracts, where the speed of price discovery attracts aggressive YES-side retail flow.

The market-making operation on Kalshi crypto is sophisticated. Top-tier makers quote across the full 188-strike ladder simultaneously, dynamically adjusting based on real-time spot price feeds, implied volatility surfaces, and order flow toxicity scores. They earn the +1.12% average documented by Becker but face tail risk on high-volatility events where 5-10 strikes can flip from deep OTM to near-ATM in seconds.

Model-Based Traders

Model-based traders occupy a middle ground between retail and market makers. These are quantitative participants running pricing models — Black-Scholes variants, jump-diffusion models, or proprietary approaches — and trading when market prices diverge from model fair value.

Our own backtesting research found that model-based signals generate roughly +1.2 cents per trade after taker fees across 28,496 qualifying signals. The edge is real but thin: a Kou jump-diffusion model with fat-tail parameters calibrated on 5,152 hourly returns (kurtosis 32.9) produced slightly better results at +1.4 cents, but the improvement didn’t justify the added complexity. The market already prices in fat tails.

The critical constraint for model-based traders is capacity. Of our 28,496 signals, 68% of the target contracts showed zero exit liquidity. The correlation between model-identified edge and market volume was 0.849 — meaning the largest mispricings exist in the least liquid contracts. Realistic capital deployment is capped at roughly $5,000-25,000 before market impact destroys the edge.

Cross-Platform Arbitrageurs

Cross-platform arbitrageurs exploit pricing differences between Kalshi and Polymarket. Our analysis found 12 actionable arb opportunities with an average net profit of 7.3 cents across 37 cross-platform pairs. But the opportunity set is structurally limited: only 108 multi-source markets out of 2,205 total active markets overlap across platforms — a mere 4.9% of the addressable market.

The arb window is also short. When a genuine cross-platform discrepancy opens, it typically closes within 90-180 seconds as automated systems on both sides converge. Manual arb execution is essentially impossible at these speeds. You can monitor cross-platform pricing efficiency in real time on our dashboard.

Becker (2026) quantified the maker-taker dynamic across 72.1 million trades: market makers earn an average of +1.12% per contract while takers lose -1.12%. This is consistent with other binary options markets and reflects the cost of immediacy.

Polymarket’s Speed Advantage

Polymarket’s market makers operate with a structural advantage that Kalshi can’t easily replicate. Polymarket’s CLOB (Central Limit Order Book) runs on Polygon with approximately 250-millisecond block times, allowing market makers to update quotes faster than on Kalshi’s infrastructure. Live prices come from clob.polymarket.com, not from the slower Gamma API — a detail that trips up many data consumers.

This speed differential matters most during high-volatility events. When BTC drops $500 in a minute, Polymarket market makers adjust in sub-second timeframes while Kalshi quotes may lag. The result: Polymarket typically shows tighter spreads on crypto markets, though Kalshi’s regulated status and USD settlement provide their own advantages.

What Does the Order Book Actually Look Like?

The microstructure of Kalshi’s crypto order book reveals patterns that aren’t visible from price data alone. Across 877,606 settled contracts, the structural imbalances are consistent and measurable.

NO-Side Depth Asymmetry

The NO side consistently carried 40x more resting liquidity than the YES side. Here is what a representative order book snapshot looks like for a BTC 5-minute contract at the $95,100 strike with 3 minutes remaining:

Total YES-side resting depth: 29 contracts. Total NO-side resting depth: 735 contracts. Ratio: 25:1 in this snapshot, consistent with the dataset-wide 40:1 average. The NO side is a wall; the YES side is a screen door.

This creates a structural environment where:

1. YES buyers face immediate fills but wide spreads
2. NO buyers get better prices but may need to wait
3. Large YES orders move the market more than equivalent NO orders

The asymmetry is most pronounced on 5-minute contracts, where the speed of price discovery attracts aggressive YES-side retail flow.

The Retail Optimism Tax

Our dataset reveals what we call the “retail optimism tax” — a systematic overpricing of YES contracts driven by directional retail flow. The pattern is clearest in the 20-50 cent price range, where retail traders concentrate their activity.

Contracts priced at 30 cents (implying a 30% probability) settle YES approximately 26-28% of the time. The 2-4 percentage point gap represents the optimism premium that flows from retail to market makers. On low-probability contracts (under 10 cents), the bias is even more pronounced.

This is the favorite-longshot bias well-documented in sports betting and horse racing, now playing out in five-minute crypto bets. Retail traders overestimate the probability of dramatic short-term moves, and market makers systematically collect the difference. We found a strikingly similar pattern in Kalshi weather markets, where low-probability temperature brackets also settle YES less often than their prices imply — suggesting this bias is structural to prediction markets, not specific to crypto.

Calibration Across the Probability Spectrum

Our model calibration across 5.48 million observations shows where market prices are most and least efficient:

Markets are most efficient in the 20-40% probability zone. Extreme prices (sub-10% and above-90%) show systematic overconfidence — models (and markets) treat these events as more certain than they actually are.

ETH markets show more severe miscalibration than BTC, likely reflecting lower liquidity and less sophisticated market-making. For ETH, even mid-range probabilities (60-70%) show +1.68 percentage points of overconfidence.

What Should Traders Actually Look For?

The data points to several concrete conclusions for anyone trading or considering trading Kalshi crypto contracts. Our analysis draws on 28,496 signals across 877,606 settled contracts, combined with the full 72.1 million trade records.

When Markets Are Most Efficient

Efficiency varies by time-to-expiry and contract type. Our research found:

• 5-minute contracts: Most efficient in the final 60 seconds, least efficient in the first 2 minutes of the window. The TWAP adjustment hasn’t fully priced in during early trading.
• Hourly contracts: Peak inefficiency occurs 30-40 minutes before expiry, when the market is transitioning between “anything could happen” and “settlement is approaching.”
• Daily contracts: Least liquid overall, with the widest spreads (6-10 cents on mid-range strikes) but the highest model-identified edge per signal.

Optimal Holding Periods and Exit Timing

Our backtests reveal a clear hierarchy of holding period performance. The data comes from modeling various exit strategies across the 28,496 qualifying signals:

• 5-minute exits: Average P&L of -1.0 cents per contract. Exits this fast incur round-trip fees twice (entry + exit) on a position that hasn’t had time to converge toward fair value.
• 10-minute exits: Average P&L of -0.8 cents. Slightly better than 5-minute, but the improvement doesn’t cover double fee drag.
• 30-minute exits: Best risk-adjusted returns among all holding periods tested. The 30-minute window gives enough time for price convergence while avoiding overnight or multi-hour holding risk. Average P&L turns positive at +0.3 cents when combined with model-based entry signals.
• Hold to settlement: Average P&L of +1.2 cents for model-selected signals. The highest raw return but carries full settlement risk — the binary outcome is fully realized, with no ability to cut losses if the market moves against you.

The 30-minute result aligns with what we found in our weather convergence research, where 804,248 trade-out scenarios showed that winners drift from a median of 43.8 cents to 77 cents in the final hours, with most of the price convergence happening in predictable windows. Intermediate holding periods capture a portion of the convergence while preserving optionality.

Time-of-Day Efficiency Patterns

Not all hours are created equal. Our analysis of trade timestamps reveals systematic differences in market efficiency across the trading day:

• US market hours (9:30 AM - 4:00 PM ET): Tightest spreads (1.5-2.5 cents), deepest books, and the most efficient pricing. Model-identified edge is smallest during these hours — roughly +0.8 cents per signal. This is when the most sophisticated participants are active.
• Asian market hours (8:00 PM - 4:00 AM ET): Wider spreads (3-5 cents) and thinner books. Model edge increases to approximately +1.8 cents, but fill rates drop significantly due to lower resting liquidity.
• Off-peak hours (4:00 AM - 9:30 AM ET): Widest spreads, lowest volume. The edge per signal is highest (+2.1 cents average) but practical execution is difficult — many strikes show zero resting orders.

The implication for model-based traders: US hours offer the most reliable execution but the thinnest edge. Overnight sessions offer larger mispricings but the worst fill quality. The capacity-efficiency tradeoff is inescapable.

Fee Comparison With Worked Examples

Consider a trader buying a BTC hourly contract at 35 cents:

On Kalshi:

• Taker fee: ceil(0.07 * 0.35 * 0.65 * 100) = ceil(1.5925) = 2 cents
• All-in cost: 37 cents for $1 payoff potential
• Breakeven: needs 37% probability (not 35%)

On Polymarket (post-March 30, 2026):

• Taker fee: 1 × 0.35 × 0.072 × (0.35 × 0.65) = 0.574 cents (~0.57 cents)
• Plus typical spread: 1-2 cents
• All-in cost: approximately 36.6-37.6 cents
• Breakeven: approximately 36.6-37.6%

At 35 cents, the platforms are now nearly identical on all-in cost. Kalshi’s fee is higher in absolute cents (2 vs. 0.57) but Polymarket’s wider spreads close the gap. The real difference shows at 50/50 prices: Kalshi charges 1.75 cents (3.5%) while Polymarket charges 1.80% — effectively the same drag on mid-range contracts.

For makers, Polymarket is now clearly cheaper: zero maker fees plus a 20% rebate from the taker pool. Kalshi charges makers 25% of the taker fee. This maker incentive explains why Polymarket’s order books remain deeper despite the new taker fees.

The Honest Bottom Line

Our backtest of three pricing models across 28,496 qualified signals found average profits of +1.2 to +1.4 cents per contract after taker fees. That’s a real edge, but it’s thin.

At taker fees, the edge survives. At maker rates, it roughly doubles to +1.8-2.0 cents. But the capacity constraint is severe: 68% of contracts showed zero exit liquidity for trade-out strategies, and the correlation between model-identified edge and market volume is 0.849. The biggest opportunities exist in the least liquid contracts.

For retail traders buying YES contracts on directional moves, the math is unfavorable. Fees, favorite-longshot bias, and the 40x NO-side depth advantage all work against you. With Polymarket’s March 2026 fee rollout, the “fee-free alternative” no longer exists — crypto takers now pay 1.80% peak on Polymarket versus 3.5% peak on Kalshi, but both platforms extract enough to make casual speculation negative-EV. The honest assessment: crypto binary options are a negative-sum game for directional retail traders on both platforms, and a slightly positive-sum game for model-based participants willing to accept thin margins and capacity constraints. This conclusion is consistent with every efficiency test we’ve run — our weather model research across 1,506 events reached the same verdict on a completely different asset class.

For a deeper look at our backtesting methodology and the specific models we tested, see our full backtest writeup. To monitor current cross-platform pricing efficiency in real time, check the SIGNAL index on our dashboard.

Frequently Asked Questions

How do Kalshi crypto binary options work?

Kalshi offers binary contracts on whether BTC, ETH, and other crypto assets will finish above or below a specific price at the end of a 5-minute, hourly, or daily window. Each event produces approximately 188 contracts across a strike ladder. Contracts settle at $1 if the condition is met, $0 if not. Settlement uses CF Benchmarks TWAP pricing, which samples 60 one-second readings and applies a 20% trimmed mean to resist manipulation.

What are the fees on Kalshi crypto contracts?

Kalshi charges taker fees using the formula ceil(0.07 * P * (1-P) * 100) cents per contract. Fees peak at 1.75 cents for 50/50-priced contracts and decline toward zero at extreme prices. Maker fees are 25% of taker fees. The P*(1-P) structure means OTM contracts (under 10 cents) pay a higher fee as a percentage of contract price — 6.8% at 5 cents versus 3.5% at 50 cents. Our fee calculator can model exact costs for any position size.

Are Kalshi crypto options profitable?

For model-based traders, our backtest of 28,496 signals found +1.2 cents average profit per signal after taker fees. The edge is real but thin, and 68% of target contracts showed zero exit liquidity. For retail directional traders buying YES contracts, the combination of taker fees (-1.12%), favorite-longshot bias (-2 to 4 percentage points), and the 40x NO-side depth asymmetry makes consistent profitability unlikely. Market makers earn +1.12% on average across 72.1 million trades (Becker 2026).

How does CF Benchmarks TWAP settlement prevent manipulation?

CF Benchmarks samples 60 one-second prices during the final minute, then discards the 12 highest and 12 lowest readings (20% trimmed mean). A manipulator who spikes the spot price at 2:59:58 only affects 2 of 60 readings, and those readings are likely discarded by the trimming. The cost of moving spot crypto markets by $200+ for a sustained 10+ seconds makes manipulation economically irrational for the $1 binary payoff. The tradeoff: the trimming reduces effective volatility by a factor of 0.554, meaning models must adjust or they will systematically overestimate move probabilities.

Is Kalshi better than Polymarket for crypto trading?

Neither platform is universally superior. Both now charge explicit taker fees on crypto — Kalshi peaks at 3.5% (1.75 cents) at 50/50, Polymarket peaks at 1.80% at 50/50. Polymarket’s lower taker rate is partially offset by its maker rebate program (20% of taker fees returned to liquidity providers), which keeps its order books deeper. Kalshi offers CFTC-regulated contracts and USD settlement. Polymarket offers tighter spreads (1-2 cents versus 3-5 cents on comparable crypto contracts) and faster quote updates via 250-millisecond Polygon block times. For a 35-cent contract, all-in costs are nearly identical on both platforms.

Key Takeaways

• Kalshi processes $60M+/day in crypto volume, driven by 5-minute contracts that revived a product from $2K/event lows to $439K/event medians
• CF Benchmarks TWAP settlement reduces effective volatility by 0.554x — a detail most pricing models miss, causing systematic overestimation of move probabilities
• Market makers earn +1.12% on average while takers lose -1.12% (Becker 2026, 72.1M trades)
• NO-side depth is 40x deeper than YES, reflecting institutional market-making against retail directional flow
• Model-based trading produces +1.2 cents per signal after fees — real but thin, with severe capacity limits (68% of contracts show zero exit liquidity)
• 30-minute holding periods produced the best risk-adjusted returns; 5-minute exits averaged -1.0 cents P&L
• US trading hours show tightest spreads and smallest edge; overnight sessions offer larger mispricings but worse fill quality

For the full academic treatment of these findings, see our research paper on crypto binary option efficiency.