How long does a wall last?
On SPY and QQQ, May 2018 to December 2024, a call wall rebuilt at 09:36 was still on the same strike the next session 59% of the time, after five sessions 33%.
In plain words: nobody placed a trade and no trade is simulated on this page. It measures recorded prices, so a result here is not a trading result and no profit is implied.
In this article 12 sections
The short answer
Not long. On SPY and QQQ from May 2018 to December 2024, a call wall rebuilt at 09:36 was still on the same strike the next session 59% of the time, five sessions later 33% and 21 sessions later 11%. Put walls, and walls read at 12:00 or 15:56, gave shares within 4 points of these from five sessions on.
It was shorter in recent years: after five sessions 39% in 2018–2021, 29% in 2022 and 23% in 2023–2024. Inside one day the 09:36 wall was still on the same strike at 10:00 in 91% of cases and at 15:56 in 75%, a lower bound, because the rebuilt wall moves inside a day only when price moves.
“Moved” is not “stopped mattering”. The walls are rebuilt at each clock, not copied from the live chart, and this page measures whether a level moved, not whether price respects it.
How long a wall lasts, rebuilt at the five-minute mark
SPY and QQQ, 2018–2024
The prices are recorded history. The walls are rebuilt after the fact from stored option data as the wall in force at a five-minute mark, so they can differ from the levels a live Prism by GEXfreak chart drew. No trades are involved.
The walls in the first two views are the wall in force in a rebuild at each five-minute mark, not a record of what any member saw. The third view is the earlier version, kept as it was measured.
- Rebuilt wall
- Earlier nearest expiry
Table, method and data source
| Sessions later | Start days | Rebuilt wall | Earlier nearest-expiry wall | Line that follows price | Wall minus line, points (95% range) | Median move of the wall |
|---|---|---|---|---|---|---|
| 1 | 3,336 | 59.4% | 33.2% | – | – | 0.0% |
| 5 | 3,328 | 32.9% | 12.8% | – | – | 1.3% |
| 10 | 3,318 | 20.8% | 7.7% | – | – | 2.3% |
| 21 | 3,296 | 11.2% | 5.3% | – | – | 3.4% |
- The wall in force at a clock is the wall in the latest five-minute rebuild usable at that clock (a rebuild is usable one minute after its mark: 09:36 reads the 09:35 rebuild, 12:00 the 11:55 rebuild, 15:56 the 15:55 rebuild). Move = |wall later - wall now| / the market price when the wall was read. Within a band = the share of walls that moved no further than 0.5% (or 1%) of price. Same strike = the wall did not change strike. Sessions are counted on the trading calendar; if the later session or clock has no wall (a missing rebuild, or an early-close afternoon), that start day is left out.
- Price-following line: on the same start day and side, a line at the wall's own distance from price that moves with price, so its move = (wall / price) x |price later / price now - 1|. Same distance, same days, same bands. A fixed random line at the same distance never moves, so it is not informative and is not used. For the 5-session wall-versus-market table the placebo pairs the wall change with the market return of a random other period (a circular shift of 21 to 1,600 sessions, 2,000 shifts).
- Old = the earlier version's nearest-expiry (N) book, read at the open from the prior close, price = the prior close. New = the wall in force at the clock, price = the market price at that clock. Both on the start days where both arms have a wall at the start and at the end. new_prior_close_* recomputes the new arm with the prior close as the price.
- 1,675 of 1,679 SPY sessions and 1,675 of 1,679 QQQ sessions have rebuilt walls.
- The levels are tier-0 rebuilds: each five-minute mark uses that morning's open interest and the prior close's option prices, re-evaluated at the market price at that mark. They are rebuilt for this study, not a record of what any member saw and not a copy of the live chart. How closely tier-0 matches a true intraday map was checked on SPY only, from October 2025 to August 2026 (after this window); QQQ accuracy is unmeasured, and the check found the most error late in the day and in quarterly months.
- Inside a session the open interest and option prices are held fixed, so a rebuilt wall can only move here because the market price moved. The within-day columns are a lower bound on how much a live wall could move, not a measure of it. Between sessions a wall also moves with each new session's open interest.
- The five expiry-window columns of the earlier version (nearest expiry, 7, 14, 30 and 90 days) cannot be rebuilt: the snapshot store holds one wall set per five-minute mark. That table is kept below, marked as the earlier version. The earlier walls are kernel-bucket walls rebuilt once a session from the prior close, not the wall set drawn on the chart, and the two sets differ: the rebuilt walls sit on a multiple of $5 in 90% to 95% of sessions at 09:36, the earlier nearest-expiry walls in 60% to 66%.
- Through 28 March 2023 the option prices are an older fallback estimate; from 29 March 2023 they are native. The persistence, parity and within-day results for SPY and QQQ together are shown for the two periods separately, and for quarterly months (March, June, September, December) against the other months. The wall-versus-market table, the market-move table and the strike-spacing figures are not split this way.
- A window starts every session, so windows overlap and the counts are not independent trials. Bands are 95% ranges from a bootstrap over calendar months.
- The price-following line is a line at the wall's own distance from price that moves with price. It has the same distance and the same days as the wall, so it shows how far a level would move if it simply tracked the market. It is not matched to the strike grid: a wall sits on a strike, usually a multiple of $5, and the line can end anywhere. As a check run after the study (post hoc, not part of the frozen protocol), the line was also snapped to the nearest $5 at the later date. Call wall read at 09:36, SPY and QQQ together, share of start days that moved no further than 0.5% and 1% of price: after 5 sessions 35.3% and 40.1% for the wall, 16.4% and 33.1% for the unsnapped line, 24.3% and 28.3% for the line snapped to the nearest $5; after 1 session 62.2% and 67.9% for the wall, 40.0% and 67.2% for the unsnapped line, 51.7% and 57.1% for the line snapped to the nearest $5. The full snapped table is stored with these results, without bands.
- The earlier version compared strikes with exact equality, and about one in ten of its stored strikes carry rounding noise near 6e-14. Recomputing its same-strike shares with a tolerance of 1e-9 gives the same numbers (the strike-noise check stored with these results), so nothing in the earlier table moves; comparisons here use the tolerance.
- This measures how far a rebuilt wall moves, not whether it acted as support or resistance, and it says nothing about any trade.
Data receipt: 16 sources, built 2026-09-29
Each source is fingerprinted with SHA-256, so a changed file changes the fingerprint. A single file shows its own hash; a source of several files shows one hash of their names and hashes.
- SPY five-minute level rebuilds, 2018 to 2024 (7 files) ·
a8f3316c7a050109… - SPY levels, nearest expiry (earlier version) (8 files) ·
79e54806a25c2dd2… - SPY levels, expiring within 7 days (earlier version) (8 files) ·
78e0b100b0c87ebe… - SPY levels, expiring within 14 days (earlier version) (8 files) ·
574a6235889a467c… - SPY levels, expiring within 30 days (earlier version) (8 files) ·
4a0140085ed79dc9… - SPY levels, expiring within 90 days (earlier version) (8 files) ·
6c300d08252d3575… - SPY daily prices, regular hours (the trading calendar) ·
b07df140c15ec92f… - QQQ five-minute level rebuilds, 2018 to 2024 (7 files) ·
e5b3b18463aafb81… - QQQ levels, nearest expiry (earlier version) (8 files) ·
61635b7513cc309d… - QQQ levels, expiring within 7 days (earlier version) (8 files) ·
4da191f589de3036… - QQQ levels, expiring within 14 days (earlier version) (8 files) ·
4f05121639e420f7… - QQQ levels, expiring within 30 days (earlier version) (8 files) ·
01b7f086ddf58c52… - QQQ levels, expiring within 90 days (earlier version) (8 files) ·
4e321a062f26cad9… - QQQ daily prices, regular hours (the trading calendar) ·
98b03f4add721f50… - Earlier version of this evidence ·
48adc73c02be21fb… - The level reader every rebuilt wall comes through ·
ea82eeb097740870…
Descriptive history, not a forecast. “Same strike” means the wall did not change strike; a price-following line sits at the wall’s own distance from price and moves with it.
What changed from the earlier version
The earlier version measured a different wall. It built five expiry windows once a session from the prior close and read each at the open; the comparison below uses its nearest-expiry wall. This version measures the wall in force in the rebuild at each clock: the wall in the latest five-minute rebuild usable at that time. The two are different constructs (two definitions of the wall), not a correction: the earlier table measured what it said it did, and rerunning its code reproduces all 2,098 of its published numbers. The five windows cannot be rebuilt (the store holds one wall set per mark), so their table stays below, marked as the earlier version.
| Sessions later | Start days | Same strike | Within 0.5% | Within 1% |
|---|---|---|---|---|
| 1 | 3,331 | 33.2% → 59.4% | 47.8% → 62.2% | 65.3% → 68.0% |
| 5 | 3,316 | 12.8% → 33.0% | 21.3% → 35.3% | 34.0% → 40.1% |
| 10 | 3,306 | 7.7% → 20.8% | 13.0% → 22.6% | 22.3% → 25.6% |
| 21 | 3,284 | 5.3% → 11.2% | 8.0% → 11.8% | 12.2% → 13.5% |
How the two versions compare
The rebuilt wall stays on its strike far more often: after five sessions 33.0% against 12.8%, a gap of +20.2 points (95% range 17.4 to 23.0). Of the 72 comparisons (three clocks, two sides, four holds, three measures), 66 differ by a rule set in advance (the range excludes zero and the gap is 2 points or more); the 6 that do not are all within-1% cells, and after five sessions in the native-price years (from 29 March 2023; before that, an older fallback estimate of option prices) the rebuilt call wall was behind the earlier one at that band (34.6% against 39.0%). After five sessions the same-strike gap was +22.3 points in the fallback-price years, +14.4 in the native-price years, +20.9 in quarterly months and +19.9 in others.
The earlier wall is read at the open and measured against the prior close, the rebuilt wall at 09:36 against the price then; using the prior close instead moves any rebuilt share by at most 0.24 points.
How the walls were rebuilt and counted
At each five-minute mark the rebuild holds one call wall and one put wall. The wall in force at a clock is the wall from the latest rebuild usable then (a rebuild is usable one minute after its mark, so 09:36 reads the 09:35 rebuild). “Sessions later” compares the wall with the wall at the same clock that many trading sessions on; within 0.5% (or 1%) means it moved no more than that share of the price it was read at.
A window starts every session, so the 3,328 start days at five sessions overlap and are far fewer independent trials; the 95% ranges come from redrawing whole calendar months at random. A price-following line sits at the wall’s own distance from price and moves with price exactly.
How long the rebuilt wall lasts
Read at 09:36, 59% of call walls were on the same strike the next session, 33% after five, 21% after ten and 11% after 21. The typical (median) wall had not moved after one session, because more than half kept their strike, and had moved 1.3% after five sessions and 3.4% after 21. Across the three read times, 33% to 35% were on the same strike after five sessions.
| Sessions later | Start days at 09:36 | Same strike, 09:36 | Same strike, 12:00 | Same strike, 15:56 | Within 0.5%, 09:36 | Within 1%, 09:36 | Median move, 09:36 |
|---|---|---|---|---|---|---|---|
| 1 | 3,336 | 59.4% | 63.2% | 62.7% | 62.2% | 67.9% | 0.0% |
| 5 | 3,328 | 32.9% | 34.6% | 33.9% | 35.3% | 40.1% | 1.3% |
| 10 | 3,318 | 20.8% | 21.5% | 22.1% | 22.6% | 25.7% | 2.3% |
| 21 | 3,296 | 11.2% | 11.9% | 12.3% | 11.8% | 13.5% | 3.4% |
By year, price era, month and time of day
Persistence fell across the years: 39% in 2018–2021, 29% in 2022 and 23% in 2023–2024 (call wall on the same strike after five sessions). The two price eras split the same way (37% while option prices were the fallback estimate, 22% after), but the native era begins in March 2023, so year and era are one comparison, not two, and this page does not say why persistence fell. Quarterly months (March, June, September, December) and the others were 33% and 33%.
A wall that moves usually moves by $5 or more (72% of SPY’s one-session call-wall changes, 85% of QQQ’s), and rebuilt walls sit on a multiple of $5 in 90% (SPY) to 95% (QQQ) of sessions. Inside a day, the 09:36 wall was on the same strike at 15:56 in 68% (Native prices) to 80% (2018–2021) of cases by period. These same-day shares are lower bounds on how often a live wall changes (see Limits).
| Period | Start days, 5 sessions | After 5 sessions, 09:36 | 12:00 | 15:56 | Same day, at 10:00 | 12:00 | 15:56 |
|---|---|---|---|---|---|---|---|
| All years | 3,328 | 32.9% | 34.6% | 33.9% | 90.6% | 82.0% | 75.1% |
| 2018–2021 | 1,836 | 39.1% | 40.8% | 40.7% | 92.2% | 86.2% | 80.0% |
| 2022 | 502 | 29.5% | 29.3% | 27.5% | 89.8% | 78.5% | 69.8% |
| 2023–2024 | 990 | 23.1% | 25.8% | 24.5% | 88.2% | 75.9% | 68.8% |
| Fallback prices (to 28 Mar 2023) | 2,456 | 36.7% | 38.3% | 37.8% | 91.4% | 83.9% | 77.8% |
| Native prices (from 29 Mar 2023) | 872 | 22.1% | 24.0% | 22.8% | 88.6% | 76.5% | 67.7% |
| Quarterly months | 1,108 | 32.8% | 34.0% | 34.4% | 91.5% | 82.7% | 75.3% |
| Other months | 2,220 | 33.0% | 34.9% | 33.7% | 90.2% | 81.6% | 75.1% |
Does a wall move like price?
Against a line that follows price at the same distance, the rebuilt wall stayed within 0.5% of its start more often: after five sessions 35.3% against 16.4% (+18.9 points, 95% range 16.2 to 21.6) and, at 1%, 40.1% against 33.1% (+7.0 points, range 4.0 to 9.9). At 1% after one session the two were not clearly apart (+0.8 points, range −2.4 to 3.8), and within a day the wall was behind the line at that band at 10:00 and 10:30 and ahead of it from 15:30 on.
This comparison is not matched to the strike grid: a wall sits on a strike and the line does not. In a check run after the study (not planned beforehand, with no ranges), the line was rounded to the nearest $5 at the later date. That moved the five-session line from 16.4% to 24.3% at 0.5% and from 33.1% to 28.3% at 1%, so the wall’s lead became +11.1 and +11.8 points. For the put wall the lead went from +16.2 and +3.1 points to +8.0 and +7.9. The lead depends on how the line is built: read it as a range. This page does not say it comes from the strike grid, or that it does not.
When the call wall changed strike over five sessions, it moved the same way as price in 94% of SPY windows and 94% of QQQ windows; paired with the price move of a random other period, they matched 53% and 52%. From five sessions on, the median wall moved 86% to 113% as far as price did over the same windows.
| Sessions later | Wall, 0.5% | Line, 0.5% | Snapped line, 0.5% | Wall, 1% | Line, 1% | Snapped line, 1% |
|---|---|---|---|---|---|---|
| 1 | 62.2% | 40.0% | 51.7% | 67.9% | 67.2% | 57.1% |
| 5 | 35.3% | 16.4% | 24.3% | 40.1% | 33.1% | 28.3% |
| 10 | 22.6% | 9.8% | 14.7% | 25.7% | 20.9% | 17.4% |
| 21 | 11.8% | 5.7% | 8.5% | 13.5% | 11.3% | 9.8% |
The earlier version, kept: five expiry windows
As it was measured, longer windows stayed on the same strike more often: after one session 70% of SPY’s and 72% of QQQ’s 90-day walls, against 31% and 35% of their nearest-expiry walls.
| Expiry window | Same strike after 1 session | Same strike after 5 sessions | Within 0.5% after 5 sessions | Within 1% after 5 sessions |
|---|---|---|---|---|
| Nearest expiry | 31/35% | 12/13% | 25/18% | 39/29% |
| Expiring within 7 days | 55/54% | 20/20% | 29/22% | 41/30% |
| Expiring within 14 days | 61/60% | 28/28% | 35/30% | 44/36% |
| Expiring within 30 days | 67/67% | 34/37% | 38/37% | 46/40% |
| Expiring within 90 days | 70/72% | 39/44% | 42/44% | 48/45% |
Track a wall you wrote down
Prism by GEXfreak (membership required) draws a call wall and a put wall; another tool’s walls will differ somewhat from this page’s rebuild.
- 1. One sheet, one row per session, filled in at the same time each day (write the time down): A date, B call wall, C put wall, D price at that time.
- 2. Five sessions later, on the row for the start day: E2 =IF(B7=B2,1,0) (B7 is the call wall five sessions on) and F2 =IF(ABS(B7-B2)/D2<=0.005,1,0); fill down. =AVERAGE(E:E) is your share still on the same strike and =AVERAGE(F:F) your share within 0.5%.
- 3. Look the wall up again each time rather than reusing yesterday’s: the rebuild found 59% on the same strike the next session and 33% after five, and a live wall can change more often. Twenty sessions give only four separate five-session windows, so a share from them can only be 0%, 25%, 50%, 75% or 100%; about 50 separate windows (250 sessions) put the margin near ±13 points around a share of 33%.
Limits
These levels are rebuilt at each five-minute mark; Prism by GEXfreak keeps no five-minute history of its levels. A tier-0 rebuild uses that morning’s open interest and the prior close’s option prices, re-evaluated at that mark’s market price. Only SPY has a true intraday map (from August 2025), used just to check accuracy. Within $1 of it, tier-0 matched the call wall 77.5% of the time, the put wall 83.3%, the daily flip 68.9%, the weekly flip 73.7% and the monthly flip 66.9%. QQQ and stock accuracy is unmeasured.
- Through 28 March 2023 an option’s price is a stored last trade, or a modeled value where none traded (the fallback-premium era); after that date prices are native. The native era has 443 SPY and 441 QQQ sessions, so its cells are the thinnest. The check matched call walls within $1 80% of the time at 10:00 and 57% at 16:00 (put walls 91% and 61%), so the 15:56 reads are the least certain.
- Between sessions a wall also moves with each new morning’s open interest, which the sessions-later columns include; inside a session the rebuild holds it fixed, so the same-day columns are lower bounds.
- Windows overlap, so start days are not independent trials; no cell has fewer than 251. Only SPY and QQQ are covered, from May 2018 to December 2024: 4 of 1,679 sessions per symbol have no rebuild, afternoon reads drop early-close sessions, and 09:36 stands in for the open because the first usable rebuild is 09:36.
- The snapped-line check ran after the study and is not part of the plan written down beforehand.
- One run: the clocks and measures were fixed before the across-session numbers were computed. The same-day shares had already appeared in an earlier check of the rebuild, so that column was not blind.
- This measures how far a wall moves, not whether price treats it as support or resistance (“How often price reaches a wall” looks at that). No trade is placed or simulated here. The window ends on 31 December 2024, and the page uses no later data.
Sources and editorial notes
Read the primary references and editorial policy
Published by GEXfreak. Prism by GEXfreak is the chart product these guides describe. Every figure comes from recorded market history: prices, option open interest and recorded sessions. Where a figure shows trade results, the trades are back-tested, not placed. This is product education, not independent investment research.
Each figure lists its data sources and their SHA-256 fingerprints under “Table, method and data source”. The references below explain the statistics and the disclosure wording used on this page. They do not endorse GEXfreak or these studies.
- The Jackknife and the Bootstrap for General Stationary Observations — H. R. Künsch, The Annals of Statistics 17(3): 1217–1241. Published 1989. Checked .
For the product’s calculation and data boundaries, read Prism’s methodology.
Check a level on your own chart.
Membership is required for live SPY, QQQ and IWM books. The figures on this page are recorded history, not a live book.
Create an account to save your layout; review current access conditions on the pricing page.