Which timeframe works for crypto trading? 31,407 backtests after costs
We ran the same sixteen indicators on eight bar lengths, from 1 minute to 4 hours. The shorter the bar, the less money survived Upbit's fee and one tick of slippage.
Share of setups that made money after costs, by bar length
5%
6%
3%
6%
34%
55%
74%
81%
1m
3m
5m
10m
15m
30m
1h
4h
The percentage above each column is the share of setups above water. Vertical scale: yearly return in %.
Return per year after costs, %
Sea level (0) is break-even after costs: blue is under water, green to brown is land.
By bar length
| Bars | Setups that traded | Above water, fee only | Above water, fee and one tick | Passed the two-period test | Median return a year | Holding the same coins |
|---|---|---|---|---|---|---|
| 1-minute8 coins, Apr–Oct 2026 | 883 | 36.0% | 4.9% | 6 | −100.0% | +513.0% |
| 3-minute8 coins, Apr–Oct 2026 | 4,064 | 58.3% | 6.2% | 24 | −94.2% | +154.2% |
| 5-minute12 coins, Jan 2025 – Oct 2026 | 4,100 | 35.7% | 2.7% | 0 | −44.4% | −9.5% |
| 10-minute12 coins, Jan 2025 – Oct 2026 | 4,179 | 50.5% | 6.3% | 1 | −30.5% | −20.3% |
| 15-minute12 coins, Jan 2024 – Oct 2026 | 4,095 | 81.5% | 34.2% | 0 | −5.7% | +17.5% |
| 30-minute12 coins, Jan 2024 – Oct 2026 | 4,073 | 80.8% | 54.6% | 305 | +1.5% | +13.8% |
| 1-hour12 coins, Jan 2024 – Oct 2026 | 4,161 | 89.4% | 73.6% | 810 | +7.0% | +11.9% |
| 4-hour12 coins, Jan 2024 – Oct 2026 | 3,754 | 83.4% | 80.6% | 1,267 | +22.3% | +10.6% |
Why short bars lose
Every trade pays Upbit's 0.05% fee, once when you buy and again when you sell. A market order also crosses the gap between the best buy and sell prices, which is at least one price tick. On a coin priced at 128 won, one tick is 1 won: 0.78% of the price.
Those costs are the same whether you hold for a minute or a week. Price moves are not. Moves grow roughly with the square root of time, so a 4-hour bar moves about fifteen times as much as a 1-minute bar while each trade still costs the same.
Short bars also trade far more. The median setup made 17,128 trades in six months on 1-minute bars, and 479 in almost three years on 4-hour bars. Each of those trades paid the fee twice.
One tick decides the short bars
With the fee alone, short bars looked fine: 58% of 3-minute setups made money. Charging one price tick on each trade left 6.2%. On 4-hour bars the same tick moved the share only from 83% to 81%, because each trade there is chasing a much bigger move.
The fee-only numbers on 1-minute bars were not just optimistic but impossible: the median Williams %R setup showed more than +50,000% a year. A bar's closing price lands on the buy side or the sell side of the order book more or less at random, and a backtest that buys at one and sells at the other books the spread as profit. How the one-minute trap works.
1 and 3 minutes: the market outran every rule
April to October 2026 was a strong window for the 8 coins in the 1- and 3-minute test. Holding them returned +513% a year on 1-minute bars (six months, turned into a yearly rate). Not one setup on either bar length beat simply holding, and 80% of 1-minute setups lost at least 95% a year.
5 and 10 minutes: almost nothing above water
Only 2.7% of 5-minute and 6.3% of 10-minute setups made money. One 10-minute setup passed the two-period test, with +10.5% a year: RSI(21) buy above 65, sell below 40 + 4-hour chart above EMA 200 + RSI(14) below 70.
15 minutes: profitable at first, not in the second period
A third of 15-minute setups made money over the whole period, but none passed the two-period test. The best one, SMA 50 / 200 cross + close above EMA 200 + Bitcoin 4-hour above EMA 100, made +33.1% a year overall: +63.3% in the first period and only +4.8% in the second.
30 minutes to 4 hours: where rules survived
From 30-minute bars up, most setups stayed above water and hundreds passed the two-period test: 305 on 30-minute bars, 810 on 1-hour bars and 1,267 on 4-hour bars. The best setups that passed:
| Bars | Best setup that passed | Return a year | Worst fall | First period | Second period | Trades |
|---|---|---|---|---|---|---|
| 30-minute | SMA 50 / 200 cross + RSI(14) above 50 + Bitcoin daily above EMA 100 | +33.1% | −22.5% | +35.0% | +30.9% | 666 |
| 1-hour | Donchian 100/55: buy 100-bar high, sell 55-bar low + volume above 1.5× its 20-bar average + stop 2% | +44.4% | −27.4% | +56.1% | +31.7% | 1,018 |
| 4-hour | BB(30, 2.5σ) buy above upper band, sell below middle + close above EMA 200 + stop 2% | +64.3% | −11.3% | +70.5% | +57.4% | 517 |
Holding the same 12 coins over the same period returned +10.6% a year and was down 74% at its worst. The second period, from July 2025, was a falling market for these coins: holding them returned −13.7% a year. That makes it a hard test: the rules that passed had to make money while the coins themselves were falling.
What this does not tell you
- The 12 coins were chosen in 2026 among large coins listed for at least two years. Coins that crashed or were delisted are missing, which flatters every rule. How much that can matter.
- One exchange, one stretch of time per bar length. The 1- and 3-minute tests cover only six months.
- Slippage is one tick per trade. Fast markets and thin order books cost more.
- We tested thousands of setups, so the best of them are partly lucky. The two-period test reduces that, it does not remove it.
- Spot buying only: no shorting, leverage, funding or taxes.
Questions
What is the best timeframe for crypto day trading?
In our tests, 4-hour bars, followed by 1-hour bars. Below 15 minutes, no more than 6% of setups made money after costs on any bar length.
Is 1-minute scalping with indicators profitable?
Not with market orders on Upbit in our tests: 4.9% of 883 setups made money after costs and none beat holding the coins. Scalping that works usually relies on cheaper execution, such as limit orders and fee rebates, rather than indicators.
Why does one tick of slippage change the result so much?
Because short-bar strategies trade constantly for small moves. If a setup aims to make 0.2% per trade and each round trip costs 0.1% in fees plus 0.2% in ticks, every trade loses on average however often it wins. Check your own numbers with the bot vs. fees calculator and the Upbit tick size tool.
Browse every setup by bar length in the explorer.