Do site walls beat a mirror line over a month?

The short answer

No difference was detected in these months. On SPY and QQQ, the site’s call and put walls in force at 09:36 on each month’s first session, held for 21 sessions, were reached 79.6% of the time on average and mirror lines (lines the same distance from the open, on the other side) 78.2%. Closing beyond the line: 42.9% for both.

The gap was +1.4 percentage points; gaps from −3.9 to 6.8 fit these months, so a small pull cannot be ruled out. The test could reliably have seen only a shift of about 13 to 14 points or more, so this is not evidence that walls have no pull.

Lines near the open were reached far more often than distant ones, wall or not.

The walls are rebuilt from recorded data, not levels a member saw. An earlier version used a different wall and also found no difference.

What to do with this

• Treat a site wall as a reference line, not a prediction, until you have about 50 rows in your own log (steps below). This page measured whether price got to a wall within a month, not whether it turned there.
• Check distance first. Divide the wall’s distance from the open by the 14-session ATR (average daily range) on your broker chart, then read the table below. A line over 2 ATR away was reached about half the time, one within half an ATR about nine times in ten.
• When someone quotes how often a wall held, ask what a line the same distance away on the other side did over the same days.
• A log of 20 rows can only show a gap of 11 points or more, and 50 rows about 7. Read no pull or barrier into a smaller log.

How we tested

Touched asks whether price reached a line at all. Closed beyond asks whether the window ended past it, which a barrier would lower. Turning at a wall was not measured. On each of 80 first sessions we read the site’s call and put wall for SPY and QQQ from the 09:35 rebuild (usable from 09:36) and held them for 21 sessions. Each got a mirror line, 2 × the open − the wall; a wall on the wrong side of the open was dropped with its mirror line (65 of 320). A month’s gap runs from −100 to +100 (0 means wall and mirror line behaved alike); “points” means percentage points, not price.

A month’s score averages its usable lines, call and put equally, and a month counts when it has both: 70 of 80. “No difference detected” means the 95% range includes zero; it comes from re-drawing months in runs of three, because 27 windows overlap the next month’s.

Drift makes a mirror line imperfect: when the market rises, the wall above the open is nearer the move than its mirror line below. By side, touched ran +19.3 points for calls and −16.4 for puts, and closed beyond +39.3 and −39.3: they mirror each other by construction of drift and are not a finding, and the headline’s average of the two cancels that part. A second check, the wall against how often any line that far out on the same side was reached in all other windows, gave −0.2 points on touched and +0.4 on closed beyond; it has no range or test here.

What changed from the earlier version

The earlier version tested a different wall: the one from options expiring within 30 days, built once from the prior close’s data and usable from 09:30. This version tests the site’s own call and put wall in force in the rebuild at 09:36. They are different constructs, not a correction: on the first session they share a strike for 69.7% of 320 walls, and an earlier-version wall held for the window matches the site wall in force at 09:36 on only 30.9%, 19.7% and 8.1% of walls by sessions 5, 10 and 21.

Both versions detected no difference. Over the 59 months in both, rebuilt minus earlier was +0.8 (−2.5 to 4.2) on touched and −2.5 (−6.8 to 1.7) on closed beyond; both ranges include zero, so the two level sources cannot be told apart here, which does not mean the levels agree. The earlier version’s one uncorrected lean (closed beyond after 28 March 2023) did not repeat.

Check it yourself with Prism and a spreadsheet

Your log answers a narrower question: what did the walls on your Prism chart do next? This page cannot confirm that a wall on your chart equals the wall rebuilt here, so write down which wall and expiry setting you read, and the time (column M; if Prism offers one setting, write default), and compare only rows that used the same one. Read the walls once, shortly after 09:36 on the first session. Log one ETF, not both, since they move together: SPY, with 51 usable months of 80 against 44 for QQQ.

One row per month, headers in row 1. Use regular-hours daily prices with dividend adjustment turned off where your broker offers it. Read column K, touched: closed beyond (L) moves more with drift, and reading both raises the chance of a false alarm.

Steps, and what a small log can say

• 1. On day 1 fill A to E and M. Worked row, the rebuilt 1 November 2024 SPY month: A 2024-11-01, B SPY, C 571.32, D 580.00, E 570.00.
• 2. Mirror lines: I2 =2*C2-D2 and J2 =2*C2-E2. Worked row: 562.64 and 572.64.
• 3. After session 21 copy the 21 daily highs and lows onto a second sheet (dates A2:A22, highs B2:B22, lows C2:C22), then F2 =MAX(Sheet2!B2:B22) and G2 =MIN(Sheet2!C2:C22), pasted as values; H is the day-21 close. Worked row, 1 November 2024 to 2 December 2024: F 604.32, G 567.89, H 603.65.
• 4. Score the row. K2 =IF(AND(D2>C2,E2<C2),100*((F2>=D2)+(G2<=E2)-(G2<=I2)-(F2>=J2))/2,"") and L2 =IF(AND(D2>C2,E2<C2),100*((H2>D2)+(H2<E2)-(H2<I2)-(H2>J2))/2,""). Each adds 1 for a wall reached (or closed beyond), subtracts 1 for a mirror line, and averages the two sides. A wall on the wrong side of the open leaves the row blank, stricter than the study. Worked row: K +50.0, L 0.0.
• 5. Read the log. Average =AVERAGE(K2:K200) and count =COUNT(K2:K200); rough 95% range, low end =AVERAGE(K2:K200)-2*STDEV.S(K2:K200)/SQRT(COUNT(K2:K200)) and high end =AVERAGE(K2:K200)+2*STDEV.S(K2:K200)/SQRT(COUNT(K2:K200)), rougher than the figure’s because it treats months as independent. Check: the rebuilt months (average +1.4, standard deviation 23.3, 70 months) give −4.1 to 7.0, within about 0.2 of the figure’s range.
• 6. Count what you need: events for a gap of g points is about (2 × your standard deviation ÷ g)². SPY’s spread implies a standard deviation near 25 points, so a 10-point gap just clears zero at about 24 events, and about double that finds a real 10-point gap most of the time. Until you have 20 rows, use SPY’s figure. A range that leaves out zero at 50 rows is a lead: log 50 more before you rely on it.

Limits

• The walls are rebuilt after the fact, not what Prism by GEXfreak drew or a member saw. Each is the site engine’s wall in force in the 09:35 rebuild: not confirmed to match a wall Prism draws, not the earlier version’s 30-day wall, and not carried to another expiry window.
• 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. That check began at 10:00 and ran on SPY from October 2025 to August 2026, so it did not test the 09:36 walls here.
• 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. 59 of the 80 months fall before that date.
• This measures reaching a line within about a month, not turning at a wall, other horizons or large stocks.
• The data is SPY and QQQ, 70 usable months to 31 December 2024; nothing later was read. Months overlap, and other monthly studies here share the same 80. The wall is read six minutes after the open: 10 of 255 walls were already through the 09:35 price, and scoring touched from session 2 on also gives a range that includes zero.
• The mirror line controls for distance, not for everything; random lines on the same side and days were not run. The study’s sizing found shifts of about 13 to 14 points on touched and 11 to 12 on closed beyond in 80% of runs on simulated paths with no drift (a false detection in 0% to 1% with none planted), copied from its simulation, not re-run here; real prices may need a larger shift, and noise in rebuilt wall positions pulls a real effect toward no difference.
• About 89 range-only lines in the study, including the slices above, are uncorrected for how many were tried; none is a result. No month-long hold trade was tested, and nothing here is a trade recommendation.