Every Kalshi 15-minute market looks the same on the surface: a strike, a clock, a yes price and a no price. But they don't behave the same. We ran DeskBot with real money on 14 of them from Oct 8 to Oct 11, 2026, and logged every result. Here is what 336 settled live trades show when you split them by market.
Short version: the bot won 317 and lost 19 (94.3%). After $23.75 in fees it was up $64.06. But three markets lost money even though they mostly won: one with plenty of trades and two with only a handful. A few losses at these prices can wipe out a lot of wins.
Why a 91% win rate can still lose money
DeskBot buys near-certain sides, usually between 90¢ and 95¢. A win at 92¢ pays about 8¢ per contract. A loss costs the full 92¢. So one loss erases roughly eleven wins, and fees push the line a bit higher. The number to watch isn't the win rate on its own. It's the win rate compared with the win rate you need to break even at the prices you actually paid. Work it out for any price with the Kalshi break-even calculator.
Prefer the command line? Our fee and break-even calculator is free and open source on GitHub: kalshi-fee-calc (Python, no dependencies, reproduces Kalshi's published fee table).
Results by market (live, real money)
| Market | Trades | Record | Win rate | Needed to break even | Net after fees |
|---|---|---|---|---|---|
| XRP | 73 | 69-4 | 94.5% | 92.8% | +$14.61 |
| Ethereum | 67 | 65-2 | 97.0% | 92.5% | +$44.95 |
| Solana | 67 | 61-6 | 91.0% | 93.0% | −$23.22 |
| Bitcoin | 49 | 47-2 | 95.9% | 92.1% | +$28.25 |
| Zcash | 27 | 26-1 | 96.3% | 92.7% | +$10.37 |
| NEAR | 23 | 22-1 | 95.7% | 92.3% | +$5.82 |
| HYPE | 10 | 8-2 | 80.0% | 92.5% | −$20.76 |
| Dogecoin | 8 | 7-1 | 87.5% | 93.3% | −$7.65 |
| BNB | 4 | 4-0 | 100% | 91.7% | +$3.83 |
| Others (EUR/USD, nat gas, oil, silver, copper) | 8 | 8-0 | 100% | about 93% | +$7.87 |
"Needed to break even" is the average price paid plus fees, per contract, for that market's trades.
What stands out
Ethereum and Bitcoin were the cleanest
ETH went 65-2 and BTC 47-2, both well above their break-even lines. They are the most heavily traded crypto assets on the board, and in this sample their prices near the close tracked the final result closely.
Solana won 91% of the time and still lost money
This is the clearest example of the break-even trap. SOL went 61-6, which sounds great. But at an average cost of about 93¢, it needed 93% just to stay even. Six losses at those prices cost more than 61 wins earned. SOL moves fast late in the window, and a 90-something-cent side can flip in the last minute.
The small, volatile markets need a bigger cushion
HYPE (8-2) and DOGE (7-1) had too few trades to draw firm conclusions, but both lost money. A single late swing in a thin market is enough. If you trade these, demand a wider gap between your expected win rate and break-even, or skip them.
Entry price mattered
Trades bought at 95¢ went 29-0 and trades at 94¢ went 63-2. At 89¢ and below the record was 8-5. A cheaper price means the market itself is less sure, and these markets usually price that doubt correctly.
The honest caveats
- This covers less than four days. 336 trades is a real sample in total but a small one per market. Several markets have fewer than 30 trades.
- Results depend on market conditions. A quiet weekend and a news-heavy weekday can look very different.
- Past results don't predict future results. Trading can lose money, including more than a day's gains in a single trade.
Watch it happen live
DeskBot streams its own account 24/7 on Kick: the countdown to each close, today's record, the win rate versus break-even, and every call, wins and losses. Watch it on Kick or follow every settled call in the DeskBot Telegram channel.
Data: DeskBot's own trade ledger, live (real-money) trades only, Oct 8 8:39 AM to Oct 11 2:49 PM ET, 2026. Not financial advice. Not affiliated with Kalshi.