“Hold it for a few days” is not an exit rule. A swing-trading hypothesis needs to state what information triggers an entry, what invalidates the position, how long it may stay open and what costs accumulate while a thesis is still active. Those details make it possible to review the idea when price and opinion move.
The term “swing” has no single exchange-defined duration. Describe the holding window you intend to study in hours or completed daily bars and UTC timestamps. If you compare two assets or two tests, use the same conventions or explain why they differ.
Specify the decision horizon
Name the price series and source: for example, a particular BTC perpetual contract on a specific venue, with hourly bars. State which bar completes before the entry condition can be known and which later price represents a simulated entry. Do not treat one venue’s spot chart, another venue’s perpetual funding and a third venue’s execution depth as if they were a single market history.
Write an upper bound for the planned hold and an explicit exit or invalidation condition. “The trend no longer looks good” is not an observable rule. An exit could be tied to a documented level or a number of completed bars, provided you state when it becomes knowable and how it would be executed. A paper stop is still a model assumption, not a guaranteed fill. Hyperliquid’s current TP/SL documentation says its triggers use mark price and its triggered market or limit order follows different execution rules; a stop-limit may not fill after a gap. Model the chosen venue’s mechanics rather than assuming one stop behavior across crypto markets.
Work a holding window through a risk budget
Consider a wholly fictional spot example, included to show the arithmetic rather than propose a setup. A rule becomes knowable after a completed four-hour bar; the model enters at the next available quote of $102. The prewritten invalidation is $98, the maximum modeled loss including both assumed fees is $50, the fee assumption is 0.10% per side, and order quantity rounds down to 0.1 units. If no earlier exit rule fires, the position closes at the first available quote after 72 hours. This makes the stop, clock and decision timestamp inspectable.
For quantity q, modeled loss at the stop is q × ($102 − $98) + 0.001 × q × $102 + 0.001 × q × $98. The per-unit amount is $4.20, so q ≤ $50 ÷ $4.20 = 11.9047…; rounding down gives 11.9 units. Entry notional is $1,213.80. A modeled full fill at $98 gives $47.60 gross price loss plus $1.21 entry fee and $1.17 exit fee, or $49.98 total. The arithmetic assumes a 0.10% fee on each side and excludes spread, slippage, gaps and borrowing. The $50 is a research budget under these assumptions, not a guaranteed maximum loss: a gap or a worse fill can exceed it.
An opposite-signal exit and a time exit are different rules. The example closes at the first available quote after 72 hours only if no earlier invalidation applies; it does not keep extending the window because a position is losing. Record the actual modeled exit time and price, and include a benchmark and flat periods when comparing variants.
Cost a position through its whole expected life
A position that stays open over several sessions can cross venue funding settlements. A current funding quote does not promise the rate for the next settlement, let alone a week of payments. For each historical position, use rates with an explicit sign, interval and timestamp; check that the rate applies at the position’s settlement snapshot. Keep fees, bid–ask spread, slippage and funding separate in the ledger so each source of modeled cost remains inspectable. The worked example above is spot, so it has no perpetual funding line; switching its instrument to a perpetual contract requires adding the applicable settlement cash flows and margin assumptions.
Venue mechanics are specific. Hyperliquid’s documentation, checked 27 September 2026, describes hourly funding for its perpetuals; a positive rate means longs pay shorts on that venue. It uses a particular premium formula and oracle notional, so do not transplant its sign, timing or calculation to another instrument. Use the Hyperliquid funding documentation for that venue and check the exact market separately elsewhere. For a funding history with a missing value, keep the interval marked as unavailable instead of recording zero cost.
For a simple directional linear contract, plan the quantity from a stated risk budget and entry-to-invalidation distance. Then inspect the venue’s current margin and liquidation conditions for that exact position. A multi-day holding rule can encounter collateral needs or exits that a daily close-to-close bar never shows.
Test the claim the holding period implies
A multi-day trade often claims that an observed market behavior lasts beyond one day’s noise. Test that claim with completed, timestamped observations and a benchmark that answers the same question. If the user checks a chart every hour, assess the rule at that cadence; a strategy evaluated only once per day is a different process. Do not extrapolate unobserved intraday stop behavior from daily candles.
Reserve a later chronological period before tuning parameters. Next, rerun with higher costs and examine losing sequences, maximum holding time, drawdown and individual-trade contribution. If two profitable exits account for most of the total result, say so. Record when the holdout influenced another design choice; that segment is no longer untouched.
Preserve the thesis through the review
A pre-trade note can record the mechanism you expect, the expected time window and evidence that would change the view. An after-action note should then separate planned entry and exit from observed prices, hypothetical model fills or a documented real fill. Keep a “stay flat” decision too. Never rewrite the initial thesis to look prescient in hindsight.
At each scheduled review, log the completed-bar signal, open quantity, current invalidation level, elapsed holding time and—on a perpetual—the funding rates actually settled. Do not change the exit rule just because the chart looks uncomfortable; if the hypothesis changes, close or classify the current record under its original rule, then version a new test. A review calendar is part of the strategy because a rule checked once per day is not the same as one monitored continuously.
The SuperQuant research application is a local workflow for rules, hourly historical tests and paper practice; its example histories are synthetic and its simplified engine does not reproduce every venue’s tiered liquidation or fill behavior. A market observation and a backtest ledger are not verified exchange fills.
Use the swing and day-trading comparison to make the holding window explicit, then include each carry assumption from funding rates in the trading plan template. Read the exact instrument and current margin rules before applying any example. Multi-day exposure can lose money; this guide is educational, not an individualized recommendation.