Scalping sells itself: dozens of trades a day, small targets that hit quickly, compounding on fast-forward. The mechanical reading we tested is the honest core of most momentum-scalp tutorials — a 1-minute candle whose body exceeds 2×ATR(14), closing beyond the prior 20-minute high or low. That is a genuine momentum burst by any definition: an objectively outsized bar breaking to fresh short-term extremes. Stop 1.5×ATR, target 2R, in and out within minutes, both directions, around the clock.
We ran it across 12 months of Binance 1-minute data on BTC and ETH — 525,600 candles per symbol, every signal, 1% risk, costs on. And we ran it knowing what we were really measuring: not whether 1-minute momentum exists, but what 0.05%-per-side commission does to trades whose entire lifespan is a few 1-minute bars. The answer is the most quotable finding in this research series: both accounts finished at zero, and on BTC the engine paid $10,850 in commission — more than the entire $10,000 it started with. The autopsy below is the point of the page.
Verified Result
No edge: net negative after costs across 2 markets.
| Market | TF | Trades | Win | PF | Max DD | Net |
|---|---|---|---|---|---|---|
| BTC | 1m | 15,653 | 15.2% | 0.05 | 100.0% | -100.0% |
| ETH | 1m | 11,133 | 28.0% | 0.27 | 100.0% | -100.0% |
How the SVS 20 breaks down ▾
12 months of real 1-minute data, fees on (0.05%/side), $10k start, 1% risk. How the score works →
The exact rules we tested
- Work on 1-minute candles (both directions, no session filter — crypto trades 24/7).
- Displacement: the candle’s body must exceed 2×ATR(14) — an outsized momentum bar, not drift.
- Breakout: the same candle must close beyond the prior 20-minute (20-bar) high or low.
- Enter at that candle’s close in the breakout direction.
- Stop 1.5×ATR(14) from entry; target 2R.
- Same-bar convention: when one 1-minute bar touches BOTH the stop and the target, the engine books the stop (worst-case). At this timescale that convention materially shapes the result — a fact about 1-minute backtests generally, not just this one.
- Risk 1% of equity per trade; 0.05% commission per side; 10× max notional leverage — fees on, which is the point of the run.
Results
Binance spot 1-minute klines (data-api.binance.vision), aggregated per strategy timeframe · starting balance $10,000 · risk 1%/trade · Commission 0.05% per side; no spread/slippage modeled (BTC/ETH spot spreads are sub-basis-point); position size capped at 10× notional leverage. Generated 2026-06-12 by the Secuora ai-strategy deterministic runner (same engine as the in-app AI backtester).
| Month | Trades | Win rate | Net P&L |
|---|---|---|---|
| 2025-06 | 1533 | 12% | −$10,000 |
| 2025-07 | 1593 | 10% | −$0 |
| 2025-08 | 1387 | 11% | −$0 |
| 2025-09 | 1680 | 6% | −$0 |
| 2025-10 | 1130 | 19% | −$0 |
| 2025-11 | 847 | 26% | −$0 |
| 2025-12 | 1355 | 14% | −$0 |
| 2026-01 | 1581 | 13% | −$0 |
| 2026-02 | 779 | 32% | −$0 |
| 2026-03 | 1016 | 29% | −$0 |
| 2026-04 | 1201 | 19% | −$0 |
| 2026-05 | 1551 | 12% | −$0 |
| Month | Trades | Win rate | Net P&L |
|---|---|---|---|
| 2025-06 | 850 | 30% | −$9,986 |
| 2025-07 | 851 | 30% | −$14 |
| 2025-08 | 677 | 31% | −$0 |
| 2025-09 | 811 | 27% | −$0 |
| 2025-10 | 823 | 31% | −$0 |
| 2025-11 | 778 | 35% | −$0 |
| 2025-12 | 1082 | 24% | −$0 |
| 2026-01 | 1185 | 23% | −$0 |
| 2026-02 | 820 | 33% | −$0 |
| 2026-03 | 964 | 32% | −$0 |
| 2026-04 | 1063 | 27% | −$0 |
| 2026-05 | 1229 | 21% | −$0 |
Assumptions (how loose terms were pinned down)
- Scalp read mechanically: a 1m candle with body > 2×ATR(14) closing beyond the prior 20-minute high/low; 1.5×ATR stop, 2R target
- When one 1-minute bar touches BOTH stop and target, the engine counts the stop (worst-case convention) — at this timescale that convention materially shapes the result, which is itself a lesson about 1m backtests
- This run also measures what FEES do to 1-minute scalping — watch the fees column
The fee-drag finding: commission outgrew the entire account
Here is the geometry that decided this run before a single candle printed. ATR(14) on a 1-minute BTC chart is usually a few hundredths of a percent of price, so even this run’s 2R target (two times a 1.5×ATR stop) is a move measured in basis points — while the round trip costs 0.10% of notional (0.05% per side). Risking 1% of equity at 1-minute stop distances wants more position than the 10× leverage cap allows, and at the cap the commission per round trip eats a material slice of any winner. A scalp that executes perfectly can still finish nearly flat after costs.
The results read like a controlled demolition: roughly 27,000 signals across BTC and ETH, both equity curves to zero, and on BTC more paid in commission ($10,850) than the $10,000 the account started with. The collapsed win-rate column — 15% on BTC, 28% on ETH — has two drivers working together. Fees: a winner that barely clears its target can still book negative after the round-trip commission. And the same-bar convention: on 1-minute bars the stop and the 2R target frequently sit inside a single bar’s range, and when both are touched the engine books the stop — the conservative convention every honest backtest at this timescale has to pick a side on. And this test models zero slippage; live execution would be strictly worse. The lesson is not that 1-minute momentum is fake — the displacement-plus-breakout pattern is real. It is that a strategy must clear its costs by a multiple, and this one tried to clear them with a rounding error.
How to backtest momentum scalping on Secuora
Both halves of this entry — displacement and the N-bar breakout — are built-in primitives of the AI backtester, so you can reproduce this run in seconds and then try to fix it.
- Open /backtest/ai and type the entry: "on 1-minute candles, enter when a candle’s body exceeds 2×ATR(14) and it closes beyond the prior 20-minute high or low; stop 1.5×ATR; target 2R."
- Run it with costs on and read the fees against net P&L before you look at anything else — that one comparison is the entire lesson of this page.
- Re-run on 5-minute candles, or with a 3R target, and watch the fee share of P&L shrink as the trade’s geometry grows relative to its costs and fewer bars span both stop and target.
- Open /backtest/demo, set the chart to 1m (a derived 30s view exists for finer execution), and replay a US-open hour placing simulated SL/TP orders to feel the speed you would need to trade at.
- Log a handful of replayed scalps in the journal with the emotion fields filled in — scalping tilt is real, and the journal is where you catch it.
Methodology, in one paragraph
Data: Binance spot 1-minute klines (data-api.binance.vision), aggregated per strategy timeframe, June 1, 2025 – June 1, 2026 (12 months). Execution: Secuora’s deterministic strategy runner (the same engine behind the in-app AI backtester) — single position at a time, entries at the close of the signal candle, commission 0.05% per side; no spread/slippage modeled (btc/eth spot spreads are sub-basis-point); position size capped at 10× notional leverage, starting balance $10,000, 1% risk per trade. Swings are confirmed fractals with no look-ahead. These are mechanical results: no discretion, every signal taken. Past performance does not predict future results; this is research, not financial advice.
Frequently asked questions
Is scalping profitable?
For most retail traders paying taker fees, the math is brutal: at 0.05% per side a round trip costs 0.10% of notional, which on 1-minute trade geometry is a material slice of even a 2R target. In our 12-month test both accounts went to zero, and the BTC run paid $10,850 in commission against a $10,000 starting balance, with zero slippage modeled. Scalpers who survive do it with maker rebates or institutional fee tiers, larger targets relative to costs, or both. Edge has to clear costs by a multiple, not by a rounding error.
What is momentum scalping?
Entering in the direction of a sudden burst of speed and exiting within minutes. The mechanical version tested here: a 1-minute candle whose body exceeds 2×ATR(14) — an objectively outsized bar — closing beyond the prior 20-minute high or low, with a 1.5×ATR stop and a 2R target.
Why is the win rate in this test so low?
Two reasons stack. Wins are booked net of fees, so a scalp that barely tags its 2R target can still close negative — and counts as a loss. And on 1-minute bars the stop and target often sit inside a single bar’s range; when both are touched, the engine books the stop (the worst-case convention). The collapsed win-rate column is fee drag plus that convention made visible — not a statement about how often the breakout direction was right.
How should I test a scalping strategy before risking money?
Run it mechanically first: at /backtest/ai the displacement and N-bar breakout primitives reproduce this exact entry, with costs on, in seconds. Then replay sessions at /backtest/demo on the 1-minute chart and place simulated SL/TP orders to feel the pace. If the mechanical run cannot beat its own fee line, no amount of screen time will.