Pipe top and bottom patterns: two-bar extremes and confirmation

Long price bars do not require long candle bodies

Bulkowski describes overlapping, conspicuous spikes on weekly charts. His pipe-top guidelines explicitly allow candle shapes from doji to Marubozu. Requiring two long bodies or opposite candle colors would therefore add a filter to that description. Similar extreme prices matter, but exact equality is unnecessary; the source supplies no universal percentage tolerance for overlap.

• The original research uses weekly observations. Applying the name to two hourly candles creates a different sample; historical weekly results cannot establish an intraday edge.
• A small decline into a bottom or a small rally into a top can occur inside a larger trend. Record both the local approach and the broader context rather than assuming a major trend reversal.

A repeatable review needs explicit tolerances

The following is an illustrative review specification, not a standard definition or a tested detector. Use completed weekly OHLC bars with positive prices. Call the pair A and B, and calculate each range as high minus low. Freeze the median range M of the 52 completed weeks before A; reject insufficient history or M = 0. Compare the selected extremes using their absolute difference divided by the lower of the two prices.

• Save a screenshot at B close. Future right-side clearance cannot be required at that time without delaying recognition. Later visual review may reject a crowded shape, but must record when that information arrived.
• The 52-week median, 2× range, 1% alignment and 80% overlap are modeling choices. Different values change which pairs qualify; none has been optimized or recommended here.

Recalculate a bottom and its mirrored top

All prices below are invented. Assume M = 2 and the approach and clearance checks pass. For the bottom, A has a range of 54 - 50 = 4; B has a range of 54.20 - 50.20 = 4. The shared range is 54 - 50.20 = 3.80, so overlap is 95%. The low difference is 0.20 / 50 = 0.4%. Both bars pass the chosen length and alignment tests despite their small bodies.

• The bottom bodies are 0.20 and 0.10, only 5% and 2.5% of their four-unit ranges. A body-size filter would reject this deliberately chosen example.
• The top uses price = 100 minus the corresponding bottom price, with high and low swapped. Its ranges remain 4, overlap remains 95%, and high alignment is 0.20 / 49.80, about 0.402%.
• A later bottom bar with high 54.60 and close 54.10 has not confirmed. A close at 54.20 also fails the strict rule; 54.30 passes. For the top, a close at 45.80 fails and 45.70 passes.

Separate a lookalike from a failed confirmation

For this review, allow confirmation on the next four completed weeks after B. Before checking confirmation, invalidate a bottom candidate if that week trades below the saved lower extreme; invalidate a top candidate above the saved upper extreme. This conservative wick-based invalidation and four-week expiry are custom choices, not claims about Bulkowski's universal rules.

• An unfinished weekly bar is provisional even if its current price is beyond the boundary. Store candidate and confirmation timestamps separately; never move an alert back onto A or B.
• A return inside the pair is an observation, not a calculated trading loss. An entry model, protective orders, costs and holding period are still needed. Gaps can prevent execution at a chosen price.

Why two pipes are not a multiweek double bottom

The useful distinction is the sequence of tests. Adjacent pipe bars have no intervening rebound or pullback. Ordinary double bottoms and double tops contain separate swing areas with a reaction between them, often spread over weeks or months. That reaction supplies their neckline. Merely zooming out does not turn two neighboring candles into two independent swing tests.

Mark the anchors before asking for an alert draft

AI Chart Analysis accepts a chart image and returns chart-pattern commentary, price levels and uncertainties. Upload a screenshot with the symbol, weekly timeframe, A and B, surrounding bars and frozen levels visible. Check its interpretation against the actual OHLC data. A cropped image can hide the clearance problem, and image interpretation does not prove that a numerical detector has passed.

• For Pine Script AI Coding Agent, provide the complete review specification: baseline window, range and alignment formulas, completed-bar timing, four-week expiry and invalidation precedence. Ask for separate candidate, confirmation, invalidation and expiry events.
• Keep manual judgments about prominence and broader trend in the review log. Only convert explicitly defined conditions into alerts; do not describe a script as reproducing every visual judgment.
• Request a syntax check on the final generated source, then compare its events with the hypothetical cases above in TradingView. This page supplies a review worksheet, not a pipe scanner or a backtest.