Beat the naive trade first

The short answer

In these back-tested sessions neither wall rule averaged more than a naive trade: the same stock, side, 15:45 entry and five-session hold on every eligible session, with no signal. The put-wall hold long averaged −0.010R over 516 trades against +0.091R, a gap of −0.101R (range −0.188R to −0.012R). The call-wall breakout long averaged +0.114R over 312 against +0.151R, a gap of −0.037R (range −0.273R to +0.198R, which includes zero).

Beta-hedged (the market’s move taken off both), the gaps are −0.039R and +0.033R, and both ranges include zero. That describes these sessions and does not say either rule loses. The levels are walls rebuilt at the 5-minute mark, not copies of a live chart.

What to do with this

• Before you credit a signal, run the same trade with no signal on every eligible session in the same stocks. The gap between the two averages is what the signal added, if anything.
• Read the gap against its 95% range (where the true gap probably sits) and against the beta-hedged gap. A range that includes zero means no gap was detected, not that the signal does nothing. The put-wall hold long’s range sat below zero, but its beta-hedged range (−0.108R to +0.032R) included zero.
• Drop the stock that adds most to a total and recompute: for the call-wall breakout long that is INTC, and without it the gap per trade is +0.006R.

What the naive trade copies

A rule’s naive trade copies its stock, side, 15:45 entry and five-session hold and drops only the signal, with stop and target distances borrowed from a random real trade of the rule in that stock, in implied moves (about how far the options market expects price to move in a day). Its average is the stock’s own over all eligible sessions, so it shares none of the rule’s look-back: a comparison with “similar weeks” would carry the very condition the rule trades. The put-wall hold’s naive trade keeps one condition, dealers net long gamma, because it is a state of the rebuilt map, not a look-back. The two rules:

• Put-wall hold long. At 15:45, the rebuilt map shows dealers net long gamma, the prior close is at or above the put wall, price is within one implied move above the put wall and below the call wall, and the session low never went more than a quarter implied move below the put wall. Stop half an implied move below the put wall; target the lower of the call wall and price plus the five-session implied move; time exit at the close of the 5th session.
• Call-wall breakout long. At 15:45, the prior close was at or below the call wall in the rebuilt map and price is above it by up to one implied move. Stop half an implied move below the call wall; target price plus the five-session implied move; time exit at the close of the 5th session.

Why a naive long earns something

The stocks mostly rose, which is why a naive long earns something. Across 10,876 five-session windows on the nine rebuilt stocks (in the years each passes the coverage checks), 55% ended higher, for a mean of +0.57% a window; in 2022 only 45% did, for −0.61%. TSLA and NVDA carry much of that: without them the mean is +0.26%. Price only, no dividends; overlapping windows overstate the sample.

Two rules, two gaps

Each rule beside its naive trade, weighted to the stocks the rule traded, cut by option-price era (defined in Limits) and month type; 15:45 is the only clock, reading the wall in force in the 15:40 rebuild. The put-wall hold’s range sat wholly below zero overall (adjusted p 0.987 across the three rules scored together: no sign of a beat), in the fallback-price era and in quarterly months, when the accuracy check on SPY found the most error. The call-wall breakout’s range included zero in every row (adjusted p 0.987), and so did every beta-hedged range. The slices are uncorrected cuts of the same trades: read them as description.

The range re-draws the sessions in blocks of 20, 10,000 times, so neighbouring days are not counted as independent. Its half-width is the smallest gap the test could tell from zero: about ±0.09R and ±0.24R.

What you match on moves the gap

The naive trade depends on how it is weighted. The study weights it to the stocks the rule traded; the table shows four other ways. The put-wall hold’s gap runs from −0.104R to −0.087R across the five, the call-wall breakout’s from −0.088R to +0.005R, against its own range of ±0.24R. Choose the match before you look at the result, and report the spread.

Where the gap came from

Best and worst stock are named only among stocks with 20 or more rule trades. The put-wall hold beat its naive trade in two of nine stocks (best INTC +0.108R, worst PFE −0.233R); MSFT adds the most, −17.0R of a −52.2R total gap, and without it the gap per trade is −0.082R. The call-wall breakout’s six such stocks run from NKE +0.055R to INTC −0.463R. INTC alone adds −13.4R to a total gap of −11.6R; without its 29 trades the gap per trade over the other 283 is +0.006R.

What changed from the earlier version

The earlier version tested four rules on twelve stocks, with walls and a gamma sign read once a session from a book built at the prior close and held all day. This one reads the wall and sign in force in the 15:45 rebuild, on nine stocks and 45 of 84 possible stock-years, for the two rules whose inputs the site draws (the other two read an overnight wall move that the site does not draw). These are different constructs, not a correction: on the 6,601 sessions both cover, the earlier wall is the same strike as the rebuilt wall for 60.8% of walls, so the rules fire on different sessions (166 of the put-wall hold’s 516 trades and 140 of the call-wall breakout’s 312 are shared) and the naive trade changes with them.

The middle column runs the earlier rule on the earlier levels over the sessions the rebuild uses. Rebuilt minus earlier levels, the gap moved −0.089R (−0.215R to +0.036R) and −0.054R (−0.262R to +0.160R), both including zero; of 20 such bands, one left zero out (the call-wall breakout’s gap in quarterly months, −0.383R), about what chance gives. Earlier, TSLA supplied 81% of the call-wall breakout’s total gap (+0.381R a trade over 77 trades); rebuilt, its 49 trades gave +0.017R.

Build your own naive trade

Seven spreadsheet steps for any rule: the rule’s trades on sheet Rule, the naive trades on sheet Naive, each with A ticker, B date, C R. Worked numbers: the call-wall breakout’s 1,492 naive trades against 312.

• 1. List the rule’s instrument, side, entry time, stop, target and longest hold. Split its conditions into the signal (the event you trade on) and filters (such as skipping ex-dividend days). The naive trade keeps the filters, drops the signal and never keeps a look-back.
• 2. On Naive list every eligible session, signal or not, one trade open at a time: after an entry skip the next five sessions. Give each a stop and target from one real trade of the rule in that stock, as a percent of entry (=ABS(entry-stop)/entry), picked with =INDEX(list,RANDBETWEEN(1,COUNT(list))) and pasted as values.
• 3. Find each naive exit (minute bars if you can; on daily bars count the stop first and score your own trades the same way). Work out R: long =(exit-entry)/(entry-stop), short =(entry-exit)/(stop-entry), less costs in R (=cost_per_share/ABS(entry-stop)). Use every eligible day, not “matching” days beside your signal days: they share the market’s move.
• 4. On a third sheet, one row per ticker: E ticker, F naive average =AVERAGEIF(Naive!A:A,E2,Naive!C:C), G your trade count =COUNTIF(Rule!A:A,E2). The weighted naive average is =SUMPRODUCT(F2:F10,G2:G10)/SUM(G2:G10) and the gap is =AVERAGE(Rule!C:C) minus it: +0.114R − 0.151R = −0.037R.
• 5. Band: =1.96*SQRT(STDEV(Rule!C:C)^2/COUNT(Rule!C:C)+STDEV(Naive!C:C)^2/COUNT(Naive!C:C)), here ±0.23R.
• 6. For a beta-hedged view, take beta × the market’s return over the hold off every trade and naive trade, in R (hedged R = R − beta × market return ÷ (ABS(entry-stop)/entry), beta =SLOPE(stock returns, SPY returns) over the 60 prior sessions, sign flipped for a short), then repeat step 4.
• 7. Read it. A gap wider than its band, with no one stock carrying it (drop the biggest contributor), can be told from zero; inside the band it is not detected. A 0.10R gap with these swings needs about =COUNT(Rule!C:C)*(band/0.10)^2 = 1,600 trades, 34 years at this study’s 47 a year.

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. 15:45 is a late-day clock, when that check matched least often (call wall 74% at 15:00, 57% at 16:00).
• 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. BAC, PFE and DVN sit in the fallback era only, so the native era rests on six stocks.
• Nine stocks and 45 of 84 stock-years, chosen by GEXfreak and all still listed when the study was built, so failed companies are missing. IBM, VZ and GPRO could not be rebuilt.
• The same 2018–2024 sessions as the earlier version, so not independent evidence; it adds no test to the many-tries count. Nothing after 31 December 2024 was read, and earlier GEXfreak research had looked at 2025 for these rule types.
• Both averages include modeled commissions, slippage and dividends; a stop can be gapped through, so a loss can exceed 1R. A back-test on rebuilt levels is not a signal or a promise of a future return.