Double top pattern: confirmation beyond two highs

Identify the two highs and the intervening low

A double top chart pattern resembles an M after a rise. Follow the time order: first high H1, intervening low N, then second high H2. The neckline is horizontal at N. A small second-peak overshoot can fit a tolerance rule, while two adjacent high candles without a distinct pullback do not satisfy the rule set below.

An illustrative rule set with explicit tolerances

These settings are a worked example, not recommended parameters or a replica of the TradingView detector. Assume a separately specified prior-rise filter has passed. Use standard price candles and local highs with three completed bars on each side; reject tied highs in that seven-bar window. Retain the latest two confirmed highs only when the geometry below passes.

• Freeze both highs and N when the candidate is created. Evaluate one active candidate at a time, and consume it after one confirmation.
• Before confirmation, invalidate the candidate if a completed bar has a high above max(H1, H2) + 0.10 × D. In this example that boundary is 100.6; check invalidation before the breakdown rule on the same bar.
• The valley-depth rule limits shallow pullbacks. It does not prove buying pressure has weakened or that a reversal will follow.

Compare an unfinished M, a continued rally and a failed break

All three hypothetical paths start with the same highs at 100 and 99.6 and the same neckline at 94. Read the following as schematic price paths, not market charts or backtest results. Each arrow is a later event; the intervals are not to scale.

• Path A may later confirm, invalidate or expire. Calling it a failed trade before an entry exists mixes two different outcomes.
• Path B can develop into a different structure. Do not redraw the old peaks after the rally just to preserve a bearish label.
• Path C belongs in the breakdown sample even if the detector later removes its drawing. Omitting it biases a historical review.

Shared shape:  rise -> H1 100 -> valley 94 -> H2 99.6
Neckline:       94, frozen for this candidate

A. Unfinished:  H2 -> close 94.4 -> close 95.2
   No completed close below 94; no breakdown signal.

B. Rally:       H2 -> close 98.0 -> high 101.0
   High exceeds 100.6 before confirmation; candidate invalid.

C. Failed:      H2 -> close 94.3 -> close 93.8 -> close 94.5
   Breakdown confirmed, then neckline reclaimed.
   The original breakdown remains in the record.

Put the signal on the bar where it becomes knowable

A peak at bar 30 with three right-side confirmation bars is first known at the close of bar 33. A marker drawn back on bar 30 describes geometry, not an opportunity available then. TradingView documents this distinction for pivot-based scripts; its own Double Top indicator uses 5/5 pivots and can also show emerging structures with In Progress enabled.

• If price crossed below 94 on bar 31, this example cannot issue a confirmed-pivot signal on bar 31. Its fresh-crossing rule requires a later eligible crossing, including on bar 33 if the conditions hold.
• A wick below 94 followed by a close above it does not pass the close-based rule. An intrabar alert would describe a different event.
• Store pivot time, pivot confirmation time and neckline event time independently. Review the second peak bar by bar before trusting a historical marker.
• Waiting for the chart bar to close addresses intrabar changes. It does not remove every possible source of repainting or make a pattern predictive.

Separate breakdown failure, retest entries and trade exits

For the example, define an early failed-break event as the first later close above the frozen neckline. This is a custom rule. TradingView documents a different Failed status: price moves above the second top before reaching the target. Name the chosen definition in a report so these outcomes are not counted as though they were identical.

• A double top trading pattern is not a complete strategy until position size, stop orders, costs and maximum holding time are specified. A stop beyond a peak and a close-based exit have different risks.
• If a stop and target are both touched within one historical candle, their sequence may be ambiguous from OHLC alone. Check the broker-emulator assumptions and available intrabar data.
• Keep breakdown-entry and retest-entry results separate. Combining their best outcomes after the fact does not test either rule.

Keep a candidate log alongside the trade report

A trade export contains executed trades, so it cannot show every rejected or missed pattern. Keep a separate candidate log with the two peaks, neckline, settings, eligibility time and outcome. Label visually misidentified structures separately from correctly detected breakdowns that later lost money.

• Record all eligible candidates and classify them as expired, invalid before confirmation, confirmed, or failed after confirmation. Preserve the failed paths rather than only saving target hits.
• To measure missed patterns, review a fixed sample independently under the same written rules. A log of detected candidates alone cannot establish the miss rate.
• Give Pineify AI Coding Agent the full specification, including tie handling and event timestamps, when requesting a detector draft. Review the output and its syntax before testing it in TradingView.
• If you add trade rules, export the resulting trades for Backtest Report analysis. Use the separate candidate log to investigate false detections; the report does not automatically audit chart-pattern recognition.
• Compare fixed rules on a separate date range and include commissions and slippage. No measured win rate or preferred timeframe is established by this example.