Author: Analyst Lu Mingyu

If you have established an advance rule that a gold ATR trailing stop should only adjust in a favorable direction, then when ATR expands and the indicator reference line pulls back, an existing long stop loss must not be mechanically lowered alongside it. The ATR measures historical volatility distance, while the reference formula produces a candidate price level; whether to modify an active order depends entirely on your predefined trade management rules. If a rule permits widening stops, that represents an entirely different framework requiring fresh recalculation of position sizing and risk, rather than assuming original protection remains intact.

ATR Calculates Volatility Distance; Stop Rules Decide Implementation

The standard reference formula for the Chandelier Exit sets the long candidate level = 22-period High − 3 × ATR(22), and the short candidate level = 22-period Low + 3 × ATR(22). The StockCharts Chandelier Exit guide explicitly points out that raw long reference levels naturally fluctuate upward and downward alongside changes in the window's highest high and ATR. The formula itself carries no inherent guarantee that it will only move upward.

A higher high pushes the long candidate level upward, while an expanding ATR widens the deducted distance. As a rolling window advances and older highs drop out, the window's highest high can also decrease. Consequently, when price fails to print new highs, a downward shift in the indicator line does not indicate a calculation error. A 22-period setting and a 3x multiplier merely serve as benchmark parameters here and do not necessarily represent optimal values for every gold or silver instrument.

Prior to computation, you must also define whether to use high prices or closing prices, a rolling window or the highest high since trade entry, and which timeframe and smoothing method apply. Because variations in these inputs alter results, they must not be conflated during trade reviews. The ATR inherits price quotation units, so a distance in USD/oz must never be mistaken for a percentage. For unit conventions, refer to ATR Units and Cross-Asset Comparison.

A rolling 22-period window continually discards data falling outside its timeframe, whereas the highest high since entry preserves the peak reached after initiating the position. While a long trade remains active, the latter cannot decrease merely due to the passage of time, though the subtracted ATR distance can still expand. Defining the highest high is fundamentally distinct from imposing a rule that an existing stop loss cannot be lowered; changing high-price definitions does not guarantee that orders will move strictly upward.

When Volatility Expands, the Reference Line Can Drift Downward

Below is a pedagogical walkthrough on a hypothetical 1-hour gold chart across a single trading session, with all prices and ATR values in USD/oz (Beijing Time). Assume an active long position is held at 10:00 with a prior stop loss at 2305. At 10:00, the bar closes at 2335, the 22-period high is 2340, and the Wilder-smoothed ATR(22) is 10. Prior historical data is omitted; these values establish initial baseline states rather than representing live market or backtest records.

The subsequent two hourly candles close at 11:00 and 12:00. The 11:00 candle records a high of 2343, a low of 2324.2, and closes at 2338. The 12:00 candle records a high of 2342, a low of 2322.8, and closes at 2332. In both instances, preceding closes fall within each bar's range, producing true ranges of 18.8 and 19.2 respectively. Extending the 22-period calculation using Wilder's ATR smoothing method yields ATR values of 10.4 and 10.8, while the window's highest high remains 2343.

The exact calculations are: 11:00 ATR = (21 × 10 + 18.8) ÷ 22 = 10.4; 12:00 ATR = (21 × 10.4 + 19.2) ÷ 22 = 10.8. This represents an ongoing smoothing process continuing from previous values, rather than treating a single bar's range as the ATR. If an alternative smoothing algorithm is used, subsequent candidate levels must be recalculated accordingly, rather than retaining these numbers under a different algorithmic label.

Bar Close (Beijing Time)	22-Period High	ATR(22)	Raw Candidate: High − 3×ATR	Adopted Stop (Upward-Only Rule)
10:00	2340	10	2310	2310
11:00	2343	10.4	2311.8	2311.8
12:00	2343	10.8	2310.6	2311.8

At 12:00, the raw candidate level = 2343 − 3 × 10.8 = 2310.6, which is 1.2 lower than at 11:00. If an order mechanically mirrored the candidate line, the stop loss would be lowered by 1.2. However, under an upward-only convention, the stop remains at 2311.8. This difference originates strictly from trade management rules rather than an erroneous ATR calculation.

This example assumes bid-based charting, stops evaluated against bid prices, and successful order modifications after each close. Because the lows of both subsequent bars remained above active stop levels, no premature liquidation occurred. Tick-by-tick sequences and fills are not simulated here; if live prices had previously breached the stop condition, subsequent order adjustment tables would no longer apply.

Moving Only in a Favorable Direction Requires Explicit Rule Logic

This model introduces an explicit constraint: New Long Stop = max(Existing Stop, Current Candidate Level). Conversely, for short positions following the same tightening principle, the new stop takes the minimum of the two. This is an added trade management constraint; one cannot assume that every Chandelier indicator or third-party script implements it automatically. Furthermore, distinguish clearly between existing stops, initial entry placements, and re-entries: a new trade cannot unconditionally inherit the trailing stop state of a prior position.

Traders must also define re-evaluation frequency, whether to use closed bars or real-time tick values, and when modifications take effect. Here, recalculations occur on hourly closes and are assumed to take effect immediately. Live execution requires verification via confirmed order amendment logs. Prior to recalculation, the original stop remains exposed to market triggers; one cannot retroactively alter past exit conditions based on newer indicator readings.

Plotting a Line Does Not Mean Broker Orders Have Been Modified

An indicator candidate line on a chart, a planned price derived from rules, and the active stop price resting on an exchange server can be three entirely different figures. When an indicator line updates on screen, verify whether a modification request was transmitted and confirmed before reading back the order's actual stop price. Symbol-specific minimum stop distances and quotation conventions remain governed by exchange and broker rules. Refer to Chart Quotes vs. Stop-Loss Trigger Quotes to reconcile pricing conventions.

The MetaTrader 5 order execution manual distinguishes between server-side stop losses and client-terminal trailing stops. Built-in terminal trailing stops operate strictly on fixed-point distances; they do not automatically parse ATR or execute the custom constraints outlined here. If relying on terminal-based scripts, application shutdowns, network disconnects, or submission failures can disrupt updates. Server-side stops and local trailing logic must be audited independently.

Holding Stops Unchanged Does Not Cap All Portfolio Risk

Restricting stop adjustments to favorable directions helps maintain established loss limits, but heightened volatility can trigger earlier exits before the market resumes its original trend. It neither guarantees superior profitability nor precludes other exit strategies. Altering parameter values shifts exit criteria; this three-line illustration cannot substitute for complete strategy validation accounting for transaction costs and execution constraints.

Stop price distance must be scaled by position size to evaluate account risk, as detailed in Stop-Loss Distance and Account Risk. Even when an order rests at 2311.8, price gaps, spread widening, or thin liquidity can yield fills substantially different from expectations; the lessons in Difference Between Stop Price and Execution Price cannot be overlooked. Insufficient margin may also trigger forced liquidation, risking partial or total capital loss. Rigorous post-trade reviews should log candidate levels, execution rules, and actual broker order records separately, as retrospective chart overlays alone cannot confirm whether risk controls functioned effectively.