Day trading is a holding-period label, not a setup. For a crypto market that trades across the clock, “the session” is not self-evident. Here, intraday means that the research plan opens and closes its exposure inside a declared observation window; state separately whether a position may carry beyond it. A useful plan names the timezone and interval it studies, the exact decision time, the end of its observation window and a condition for standing aside.
Write the process before looking at today’s chart. Otherwise a plan can quietly expand: one rejected entry becomes a later entry, a wider stop, or an unrelated market. A visible session boundary also makes it possible to compare like with like in a backtest and a journal.
Define a repeatable window
Choose a clock convention such as UTC, and state what a session open and close mean in a continuous market. Daylight-saving changes may alter a local session; UTC does not change when a country shifts its clocks. If you are researching activity around a named regional opening, compare at least that exact local-time conversion and the corresponding UTC dates. Check both days that change daylight saving and days that do not.
The boundary is a way to classify historical evidence. It does not mean prices follow reliable daily rhythms. To investigate such a claim, define the event first, collect enough coverage across a stated period and compare a baseline. A busy time observed during one volatile event cannot establish a stable intraday edge.
Decide what is known before the order
Specify whether a signal uses a completed candle or an observation still forming. A rule based on the final four-hour close cannot assume it was known at that period’s start. Name the next valid decision time and entry convention. Define whether the sample includes overnight intervals, weekends and any period in which the user would not monitor.
For example, suppose a replay rule checks a one-hour candle that opens at 14:00 UTC and closes at 15:00. If the condition depends on that candle’s final close, it becomes knowable only after the interval ends. Binance’s Spot API identifies each kline by open time and returns a separate close time; other data providers may label bars differently, so record the source’s convention. A backtest can choose the next bar’s open as its entry proxy, but that is a model assumption. It does not prove an order could fill there. Never use the completed close to make a decision and then assign the fill to a price from before that close existed.
Keep the experiment narrow: one instrument, one timeframe, one entry rule and one exit. Record the level that invalidates the hypothesis and how to leave the test. If two indicators change in one run, the reason for any result becomes hard to identify. Include a deliberate stay-flat case so that no signal does not become a pressure to create one.
Budget the event-by-event costs
For an intraday perpetual, write down entry and exit fees, bid–ask spread, adverse slippage and the venue’s settled funding that overlaps the actual holding interval. Record units, source and date. A stop order may trigger without executing at its selected level. Short holding periods can make a small price movement particularly sensitive to fees, so calculate a round trip and rerun the same rule with higher costs.
Order behavior is product-specific. As one current example, Coinbase’s Advanced Trade documentation says its bracket and take-profit/stop-loss downside protection may not execute during extreme volatility. Check the current order specification for the exact venue and market you study; do not treat one platform’s protection logic as universal.
Do not confuse “intraday” with “not leveraged” or “does not pay funding.” Those conclusions depend on notional exposure, contract and the actual time at which an eligible position crosses a funding settlement. Check the signed settlements and include them individually in the test where the data supports it. Leave unverified intervals unknown.
A session-plan card
| Before the window | After the window |
|---|---|
| Named instrument, venue and exact UTC interval | Decisions and quotes recorded at their source time |
| One rule and the bar when it can first be known | Order status, modeled or actual fill clearly separated |
| A calculated size and complete assumed cost | Fees, spread, funding and missing observations |
| A stop or condition to stay flat | Which planned rule changed, and why |
| A fixed session end and review date | Whether the research question still makes sense |
Record written decisions before showing the next market segment in a replay. Otherwise, memory supplies information the trader did not have at the time.
At the end of the window, close or explicitly carry any remaining position according to the frozen rule. Reconcile the full record: no-signal periods, order requests, rejected or partial orders, modeled versus actual fills, spread, fees, slippage and any funding settlement that applied. Compare rule-following separately from the outcome. A flat session is a valid result when the setup did not occur; it should not be rewritten as a missed trade after seeing what price did later.
Try to falsify the session rule
After an initial historical run, split the period chronologically and preserve a later segment before revising rules. Compare the stated window with a broader same-market baseline, net of the same costs. Vary cost assumptions and inspect drawdown, losing sequences, trade concentration and no-trade periods. If the apparent result vanishes when the sample shifts by an hour, report that boundary sensitivity.
SuperQuant supports research and manual paper-practice workflows on hourly histories; the example histories are synthetic. Its hosted workspace is not a real exchange. It cannot validate second-by-second execution or a scalping fill. Match the research question to the actual data resolution and record that limit. For broader historical session evidence and the difference between activity measures and liquidity, see crypto market sessions; that research does not identify an ideal trading hour.
Use the crypto trading plan template to write a time boundary and cost assumptions before running. Read the broader trading-style comparison and swing-trading guide before deciding which holding horizon your data can support.
Intraday results may vary and can lose money. This is general education, not a signal, schedule recommendation or individualized advice.