What it is
Short holds, tight stops, frequent entries on momentum bursts, order-book imbalances or mean reversion to a short moving average. Edge per trade is tiny, so execution quality and fee tier matter more than the signal.
scalping
Scalping: the strategy type that fees kill first.
Scalping takes many small profits on 1-minute to 15-minute charts. A 0.04% taker fee on both sides is 0.08% per round trip, so a strategy that averages 0.1% per trade before costs keeps almost nothing. The backtest has to charge every fee.
Official Binance broker · Non-custodial · Live only when you confirm · Updated
What the backtest must get right
Taker fees on every fill, the exchange's minimum order size, fills inside the bar on 1-minute data, and funding if you scalp perpetuals and hold through a funding timestamp. OptAlgo replays all of them and the honesty grade says which were real.
A prompt that builds one
Describe entries, exits and risk in one sentence; the assistant writes the code and asks before spending backtest units.
“Scalp BTCUSDT.P on 5m: enter long when RSI(7) crosses above 30 and price is above VWAP, exit at +0.4% or −0.25%, taker fees, from 2024.”What usually survives
Most scalping ideas grade well on honesty and fail the overfit check or the walk-forward, because the edge was fees-sized noise. The ones that survive trade less often than their first version and use maker orders where the engine allows it.
faq
Scalping strategy FAQ
Is scalping crypto profitable with a bot?
Only when the edge per trade is a multiple of the round-trip fee. The backtest with real fees and a walk-forward answers it per strategy; most fail there, which is cheaper than failing live.
What timeframe does the engine support for scalping?
Strategies run at bar close on any timeframe from 1 minute up; fills are replayed on 1-minute data regardless of the strategy timeframe.
Do I need to code this?
No. Describe the rules to Claude or ChatGPT through the OptAlgo connector and the assistant writes the strategy against the OptAlgo SDK, validates it for free and backtests it. You can read and download the code.
How do I know the backtest is not overfit?
Every run returns an overfit verdict based on how many variants you tried. Then ask for a walk-forward or a holdout: the strategy must keep its edge on data it never saw before it is worth paper trading.