Gaps, Overnight Levels and the Risk You Cannot Stop

Why gaps happen, and what they create

A gap is a price jump between one session’s close and the next session’s range, and it happens because information arrives while the market is shut: an earnings release, a guidance change, a regulatory decision, a competitor’s surprise, a central bank statement. Nothing about the market is different at the open except that yesterday’s prices are stale, so the first trades happen at a level where buyers and sellers actually agree — which is usually not the level of the last close. Gaps are therefore not anomalies to be ignored; they are the visible mechanism by which information is repriced, and for a holder they are the moment when a plan’s assumptions are tested. Two properties matter for risk. The first is that the gap cannot be traded: there is no price between the close and the open at which you could have exited, so a stop on the far side of the gap is an instruction that was overtaken rather than one that failed. The second is that gaps are not symmetric in their consequences for a plan, because they arrive with the position already on. A trader who is flat has no exposure to the event; a trader who is short premium, or holding a leveraged position, has exposure that was chosen before the news existed. That asymmetry is why the calendar belongs in the trade plan next to the stop rather than in a separate habit: the event changes the distribution of the outcome, so the position should be sized for the distribution rather than for the average day. Where a gap leaves a level, it also creates one. A gap up that holds tends to be defended at its upper edge by the participants who bought it, and the unfilled portion of the gap acts as a reference for the market for weeks — which is the same level logic as TR13, with the gap edge as the zone. A gap that is filled quickly is the opposite statement: the news did not require repricing, and the price returns to where it started. The distinction is worth recording, because it is one of the few cases where a single session’s structure has consequences measured in months. Planned risk against realised risk — Entry $74.00, stop $71.55, budget 1.5% of $120,000: 734 shares — the plan’s advertised risk · Open after the report: $69.10: The stop becomes a market order and fills there · Loss per share: $4.90 against a planned $2.45: Double the stop distance, from one overnight event ← · Realised loss: $3,600 = 3.0% of the account: Budget 1.5%, realised 3.0% ← The honest maximum loss in a plan is the budget plus the largest gap the instrument has delivered in the last year or two. Quoting only the stop distance is how a plan that looks like 1% per trade turns into a series of 3% losses in a volatile quarter.

Sizing for the event, and the limits of order types

The arithmetic of the previous section gives the response, and it is a sizing response. If the scheduled event can move the instrument six percent and the plan’s budget is 1.5%, then the position that survives the event unpleasantly is a fraction of the position the stop distance alone would justify — roughly a quarter, in this example — and the alternative is simply not to hold through it. Both answers are acceptable; the unacceptable version is holding full size and calling the risk fixed. That is why the trading calendar is reviewed with the same discipline as the levels: scheduled events are known in advance, and a position that ignores them was sized with a stale distribution. The order-type question is a red herring, and it is worth understanding exactly why. A stop order becomes a market order when the price is reached, so through a gap it fills at the open — there is no mechanism to fill at a price that did not trade. A stop-limit order carries a limit price, so it will not fill worse than the limit; through a gap it simply does not fill, and the position remains open while the news is already public and the price is below the limit. Neither instrument can protect a price through a discontinuity, because the price did not exist. That leaves three honest tools: reduce size, reduce exposure before the event, or accept and state the larger loss. A fourth possibility — hedge with options — transfers the risk at a cost, and it is the only one that keeps the position on through the event, which is why event-driven traders use it. What follows for the plan is a single sentence that should appear in writing: the maximum loss on this position is the stop distance plus the largest gap this instrument has delivered since I started tracking it. That sentence does two things. It makes the risk of an event legible before the event, so that a 3% loss is a planned outcome rather than a surprise, and it makes the size arithmetic honest, because the number that gets divided into the risk budget is the gap-inclusive distance rather than the tidy one. • A stop is a market order through a gap: it fills where the market is, not where the level was. • A stop-limit protects the price by not filling, which leaves the position open through the news. • Gaps are how information gets repriced, so they belong in the plan rather than in the excuse column. • Sizing for the event means a fraction of the position, or no position, and both are defensible. • A held gap edge becomes a level; a filled gap says the news did not require repricing. • State the loss as stop distance plus the largest gap, and divide the budget into that. A specific trap: using a wider stop to justify holding an event position. The stop is not the thing that protects you through a gap, so widening it converts a defined plan into an undefined one while looking more prudent. The levers are shares and exposure, and neither is the stop.

What happens after the gap

A gap is not a signal on its own; it is a change in the level from which everything else is measured. What follows divides into two behaviours, and the base rate matters far more than the shape of the candle. A **continuation gap** opens beyond a consolidation and keeps going, because the news changed the fair value and the market is repricing to it. A **common gap** opens on thin overnight flow and closes back through itself within days. The same ten-percent print can be either, and no amount of studying the bar before the open tells you which one is forming. The useful discipline is to condition on the reason rather than the pattern. Gaps on a specific, arithmetic event — an earnings surprise, a guidance change, an index inclusion, a takeover — are far more likely to hold, because the information is durable. Gaps on a sentiment shift, a sympathy move from a peer, or a thin holiday tape are far more likely to fill, because nothing fundamental changed. Empirical work on post-earnings-announcement drift finds that the direction of an earnings surprise tends to persist for weeks, but the effect is modest and it is a drift, not a jump. Two facts follow for anyone trading around an event overnight. First, the gap that hurts is the one you held *through* — once it has happened, you are choosing between a smaller loss and a larger one, not between a loss and no loss. Second, the tendency to fill is a description, not a guarantee: base rates are used by people who have not yet been stopped out by the exception. The reason a gap holds or fills is almost always identifiable in the news, not in the chart. If you cannot say what changed, you are guessing about the base rate.

Do gaps fill? The base rate behind the claim

“Gaps always fill” is one of the most repeated statements in technical analysis and one of the least quantified. As stated it is false, and the useful version of it is a conditional statement about what kind of gap you are looking at. The mechanism underneath the claim is real. A gap is a price zone where no trading occurred, so there is no population of participants holding a cost basis there, and thin zones get traversed quickly when price returns to them — the same point the volume-at-price reading makes about low-volume pockets. What the mechanism does not say is when, or whether, price comes back. The distinction that does the work is between a gap that changed the information and a gap that changed the flow. An earnings surprise, a takeover, an index inclusion, a regulatory decision — these move the consensus value of the business, and the pre-gap price is no longer a reference for anything; asking whether that gap fills is like asking whether yesterday’s estimate comes back. A gap created by a large seller in the pre-market, by an overreaction to news that turns out not to be material, or by a thin overnight move with no news at all leaves the previous price as a live reference, and those are the gaps that tend to be revisited. There is a statistical trap in how the base rate is usually presented. Measured across all gaps, a high fill rate is easy to obtain, because most gaps are small and most small moves are retraced by ordinary noise. The question worth asking of any stated fill rate is what threshold it used, in which direction, and whether the gap came with news. A claim of the form “ninety percent of gaps fill” that is computed on gaps of a fraction of a percent, in both directions, is a statement about intraday volatility rather than about gaps. Two further pieces belong beside it. The first is the cost of avoiding the exposure: there is a documented tendency for a meaningful share of equity returns to accrue outside the regular session, so the rule “never hold through the close” gives up some of the drift as the price of avoiding the gap. The second is event gaps in particular, where the option market publishes an **implied move** — what the straddle price says the market expects the stock to do on the day. Comparing the realised gap with the implied move is the fastest way to see that the distribution is fat on both sides: most events land inside the implied move, and the ones that do not can be several times it. So the working version of the rule is an ordering: treat a gap as an information event first, ask whether the reference price it left behind still means anything, and only then treat it as a chart feature. The size decision follows from the realised gap rather than the planned stop, which is what the earlier section established as the only defence that works. • A gap changes the information or the flow; only the second kind leaves a live reference price. • Aggregate fill rates are dominated by small gaps that ordinary noise retraces. • Ask any fill statistic for its threshold, direction and whether news was involved. • Avoiding overnight exposure has a cost: part of the drift accrues outside the session. A practical comparison for any position held through an event: the implied move from the option chain against the actual gap for the last four events. That single table tells you whether your sizing has been calibrated to the distribution the market prices, or to the quieter one you remember.

Sources

• Gap behaviour and continuation studiesBulkowski, “Encyclopedia of Chart Patterns” (gaps section)
• Overnight risk and event-driven position sizingStandard risk-management practice for scheduled announcements
• Stop orders and gap fillsExchange and broker documentation on order handling through a gap