What does the overnight gap do to a stop?
A stop below the market does not guarantee an exit at that price when the instrument gaps past it — the fill happens at the reopening price, which can be considerably worse.
This is the single most consequential difference from spot forex, where continuous trading through the week means stops are usually reached rather than jumped. Indices close and reopen, and the reopening price reflects everything that happened while they were shut.
A backtest that fills every stop at the stop price is understating the tail of the loss distribution, and it is understating it precisely on the days that matter — the ones with overnight news. The maximum drawdown figure inherits the understatement.
The honest model is to fill at the first available price after the gap rather than at the stop level. It is a small code change and it can meaningfully alter the risk profile of any strategy holding overnight.
Cash session or futures session?
Index instruments trade across a longer window than the underlying exchange is open, and the character of the price action inside and outside the cash session is genuinely different.
During the cash session the index reflects actual trading in its constituents. Outside it the instrument is being priced from futures and from whatever is moving, with thinner liquidity, wider spreads and a greater tendency toward moves that reverse at the open.
A strategy tested across all hours is being tested on two different markets averaged together. Splitting the results by session frequently shows the edge lives in one of them, and the split costs nothing beyond acknowledging it as a trial.
The related trap is the open itself. The first minutes of the cash session carry enormous volume and volatility, and a strategy whose backtest fills instantly at printed prices during that window is modelling an execution quality that is not available.
What about dividends and index adjustments?
The underlying index falls when its constituents go ex-dividend, and how your instrument handles that determines whether the drop is a real price move or a bookkeeping event your strategy will trade.
Cash index CFDs typically apply a dividend adjustment to positions rather than letting the drop hit the price unmodelled, which means the price series contains a mechanical decline that no strategy should be reacting to. A backtest reading the raw series sees a downward move that never represented anything.
Constituent changes work the same way at a larger scale. When an index rebalances, the level shifts for reasons entirely unrelated to any signal, and a rule sensitive to exact levels will react to it.
The practical requirement is to know which series your data provider supplies — adjusted or raw — and to check it against a known ex-dividend date the way you would check an equity series across a split.
Do financing costs matter for index positions?
For anything held more than a day or two, considerably — overnight financing on a leveraged index position accrues daily and compounds across the longest holds, which are usually the most profitable trades.
The charge is a function of position value rather than of risk, so it scales with leverage rather than with the stop distance. A strategy sized to risk 1% but holding a large notional pays financing on the notional.
The asymmetry matters too. Financing is typically charged on longs and may be charged or credited on shorts depending on the rate environment, so a long-biased index strategy carries a persistent cost that a short-biased one does not.
Modelling it takes the daily rate and the number of nights held. Omitting it flatters exactly the swing and position strategies whose edge depends on holding through the moves that justify them.
Triple-charge nights are the detail that catches people out. Most brokers apply three days of financing on one weekday to cover the weekend, so a strategy that systematically holds across that day pays considerably more than a flat daily rate suggests — and which weekday it falls on varies by instrument and by broker, which makes it worth checking rather than assuming.