The default ADX period is 14. That is the starting point, not the final answer. Here is how to choose the right period for your timeframe and market.
Choosing the best ADX settings starts with 14, Wilder’s original specification for daily commodity charts in 1978, and then understanding what actually changes when you deviate from it. The default works across most markets. Whether it is the right choice for your specific timeframe, asset, and strategy requires deliberate testing rather than assumption.
The period setting determines how many bars are included in the smoothing calculation. ADX is not a simple moving average of price. It is derived from +DM and -DM, which measure directional movement bar by bar, then smoothed using Wilder’s moving average over the specified period. The smoothed directional movement values are divided by smoothed True Range to produce +DI and -DI, and ADX is calculated from those.
A longer period means more bars contribute to each smoothed value. The result is a slower, more stable line that filters short-term noise but takes longer to respond when trend conditions change. A shorter period means fewer bars, producing a faster, more reactive line that responds quickly to new directional movement but is more prone to false signals.
The period affects three things simultaneously: sensitivity (shorter periods detect trend changes faster), smoothness (longer periods produce fewer false threshold crossings), and lag (longer periods take more time to confirm a trend has started or ended).
There is no period setting that optimizes all three. Every choice trades one against the others, and the right tradeoff depends on the strategy.
One thing the period does not change: the fundamental nature of what ADX measures. A 7-period ADX and a 21-period ADX both measure directional movement strength. They disagree on timing, not on concept. The 7-period will confirm the trend earlier and lose it earlier. The 21-period will confirm it later and hold it longer. Neither is more accurate in any absolute sense.
Wilder developed the Directional Movement System in 1978, published in “New Concepts in Technical Trading Systems.” His design was grounded in the commodity markets of that era, specifically daily charts of markets like soybeans, cattle, and gold.
His observation was that most significant price cycles in those markets ran approximately 28 days. He selected a half-cycle lookback of 14, reasoning that a half-cycle period captured enough directional movement to be statistically meaningful while avoiding the excessive lag of a full-cycle period.
This half-cycle heuristic has generalized reasonably well across other markets and timeframes over the decades, which explains why 14 has remained the default across platforms and instruments that Wilder never intended the indicator for. But it was empirically derived from a specific market in a specific era.
Two implications follow. First, 14 may not be the optimal period for crypto, equities, or intraday charts. Second, no alternative period is automatically better — replacing 14 with another number without testing is just substituting one assumption for another. The value of understanding why Wilder chose 14 is that it gives a principled basis for deviation: if the market being traded has a different dominant cycle length, the half-cycle of that market’s cycle is the logical starting point for testing.
Shorter periods (7 to 10) react faster to emerging trends. They cross above threshold levels like 20 and 25 earlier in a trend’s development, giving earlier entry signals. ADX peaks are sharper and faster. The costs are higher false positive rates — ADX crossings that fire on brief directional bursts that do not develop into sustained trends — and noisier readings that can produce misleading regime classifications.
The default period (14) sits in the middle of the practical range. It represents Wilder’s calibrated balance for a specific context, and it works well enough across many different contexts to be a reliable default.
Longer periods (20 to 30) require more sustained directional movement before ADX reaches confirmation thresholds. The result is fewer signals overall, but higher conviction per signal. A period-21 ADX that crosses 25 has been measuring directional movement across 21 bars and smoothing out shorter-term noise in a way that period-14 does not. The tradeoff is later entries and exits, which means a smaller portion of each trend is captured.
The noise tradeoff is non-linear. Going from 14 to 10 is a modest change in sensitivity. Going from 14 to 7 produces dramatically more noise. Below period 7, ADX readings typically become too erratic to serve as a reliable regime classifier. They react to individual bars rather than sustained directional conditions.
Going in the other direction, periods above 30 on daily or intraday charts produce ADX readings that lag so far behind current conditions that their utility as a regime filter diminishes. By the time a period-30 ADX confirms a trend, the trend may already be mature or exhausted.
The practical range for most systematic traders is 7 to 21, with 14 as the default starting point.
These are starting points derived from the half-cycle logic and common practice. Validate against the specific asset and strategy before committing.
Daily charts: 14. Wilder’s default was designed for this timeframe. A 14-period daily ADX covers roughly two calendar weeks of price action. For most assets, this remains a sensible default. Consider testing 10 if you find that 14 is consistently late in confirming trends you can already see developing in price.
4-hour charts: 14 is widely used. A 14-period ADX on a 4H chart covers 56 hours, approximately 2.3 days of 24/7 crypto trading. This is a shorter window than Wilder’s daily-chart intent. Some traders use 10 to 12 on 4H crypto charts to compensate. Test whether 14 or 10 better captures the dominant trend cycles in the specific pairs being traded.
1-hour charts: 10 to 14. At 14 periods, 1H ADX covers only 14 hours of data. This represents less than one full trading day in traditional markets, or 14 hours of continuous crypto price action. Whether this window is sufficient depends on the intended holding period.
15-minute charts: 7 to 14. At 14 periods, 15-minute ADX covers 3.5 hours of price data. Shorter periods (7 to 10) are common at this resolution because the full-cycle logic breaks down for intraday timeframes. The 25 threshold may also need recalibration — brief intraday directional moves can produce ADX readings above 25 that do not represent genuine trend conditions in the way a daily chart reading does.
Weekly charts: 14 works well. Some traders use 20 to 21 for cleaner, less noisy readings on higher timeframes, particularly for long-duration position systems where early entry is less important than conviction.
Crypto markets have structural differences from the traditional markets Wilder designed for. These differences affect ADX behavior in ways that inform period selection.
24/7 trading with no session gaps. Traditional market daily bars contain overnight gaps. These gaps contribute to True Range and affect the ADX calculation. Crypto daily bars represent uninterrupted 24-hour price action with no gaps. True Range on a crypto daily bar is purely the intraday range. This makes crypto ADX readings more comparable to intraday ADX readings in traditional markets than to daily readings.
Higher structural volatility. Crypto assets exhibit higher volatility than most traditional assets at equivalent timeframes. A trend that produces an ADX reading of 30 on a daily equity chart might produce a reading of 45 on a daily crypto chart because the True Range denominator is larger. The 25 threshold still functions as a reasonable trend confirmation level, but ADX values in confirmed trending conditions tend to run higher in crypto than in equities.
Faster cycle dynamics. Crypto trends can develop and exhaust faster than traditional market trends, particularly in altcoins. This favors slightly shorter periods on intraday timeframes, where the goal is to capture trends that may last hours rather than days.
Across the live scanner running BTC, ETH, SOL, BNB, and XRP on the 4-hour timeframe since May 2026, the 14-period default has produced more reliable regime classifications than shorter variants tested against it. The noise increase from period 10 outweighed the sensitivity gain for the system’s holding periods. The 20 and 25 threshold levels held consistently across pairs. What changed between periods was not whether the thresholds worked, but how quickly ADX reached them.
The correct approach is to test across a range, measure what matters, and commit to a setting rather than continuously adjusting.
Run your strategy across periods 7, 10, 12, 14, 18, and 21 on a representative historical dataset. For each period, measure three outcomes.
Signal quality by regime. What is the win rate of trend-following signals when ADX is above threshold at each period setting? What is the win rate when ADX is below threshold? The larger the gap between the two, the better the period is at separating tradeable trend conditions from noise.
False positive rate. How often does ADX cross above the threshold and then fall back below it within three to five bars without a meaningful trend developing? A high false positive rate at a given period suggests the period is too short for the asset and timeframe being tested.
Trend capture percentage. For each identified trend in the historical data, what percentage of the trend’s total move is captured between ADX confirmation and ADX exit? A period that confirms late and exits late may capture a similar percentage as one that confirms early. What matters is whether the captured portion is the higher-probability middle of the move.
Document results across all periods before choosing. Then treat the selected period as a fixed system parameter, not a dial to adjust when the system hits a losing streak.
One additional check: test whether the period that performs best in backtesting also performs in walk-forward analysis on data not used in the original test. A period that looks good in backtesting but fails in walk-forward is a sign of overfitting, not genuine edge.
No period eliminates ADX’s fundamental limitations, and some limitations get worse rather than better when the period is changed.
Lag is structural, not adjustable. Every smoothed indicator lags price. Shorter periods reduce the lag but do not eliminate it. At period 7, ADX still confirms a trend that is already partially developed. The lag is smaller than at period 14, but the tradeoff is a meaningful increase in false signals. There is no period that produces both low lag and high signal quality simultaneously.
The 25 threshold is not universal. Wilder’s 25 level was derived from empirical observation of daily commodity charts using a 14-period calculation. There is no mathematical relationship that makes 25 the correct threshold for every period and timeframe. If you change the period, test whether 25 remains the appropriate threshold or whether a different level better separates trending from ranging conditions in your specific configuration.
Optimization is susceptible to overfitting. A period of 11 that shows a 4% win rate improvement over period 14 on a specific historical dataset is not meaningfully better. It has likely fit to noise in that dataset. Genuine improvements from period selection are large, consistent across multiple market conditions and time periods, and hold up in walk-forward testing. Marginal gains on a single backtest are not evidence of a better parameter.
The period cannot fix bad regime classification. If a market is generating ADX readings above 25 but the trend is not behaving like a trend — choppy, low follow-through, frequent reversals — changing the period will not resolve this. ADX measures directional movement strength, not trend quality. The period setting does not change this dynamic.
14 is the most widely used period and the right starting point for daily and 4-hour charts. The best period depends on the timeframe being traded, the volatility of the asset, and the strategy’s holding period. Shorter periods (7 to 10) suit faster-moving markets and intraday strategies. Longer periods (20 to 21) suit swing trading and higher timeframes. Test before committing to any setting.
Start with 14 on daily and 4-hour charts. For intraday crypto on 1-hour and 15-minute charts, test shorter periods (10 to 12) because the 14-period lookback covers a short window at those resolutions. Crypto’s 24/7 trading and higher volatility do not change the calculation fundamentally, but they do affect how quickly trends develop and exhaust. Validate on the specific asset before finalizing.
ADX 14 is more sensitive and confirms trends earlier. ADX 20 is smoother and confirms trends later but with fewer false signals. On daily charts the difference is modest. On intraday charts, ADX 20 may lag too significantly to be useful. Start with 14, then test 20 if you are getting too many false trend confirmations. The right answer depends on your strategy’s tolerance for false signals versus late entries.
The 25 threshold was observed by Wilder using 14-period ADX on daily commodity charts. Changing the period affects how quickly ADX reaches 25 and how long it stays there, but does not mathematically invalidate the threshold. Shorter periods may produce more frequent crossings of 25, including on weak trends. If you change the period, validate whether 25 remains the right threshold for your specific configuration.
For day trading on 15-minute to 1-hour charts, periods of 7 to 14 are most common. Shorter periods (7 to 10) react faster to intraday trend development. Longer periods (12 to 14) produce smoother signals with less noise. The right choice depends on how long each trade is intended to run. Test across the 7 to 14 range on your specific asset and timeframe before settling on a period.
Run your strategy on historical data across a range of periods: typically 7, 10, 12, 14, 18, and 21. For each period, measure the win rate of trend-following signals when ADX is above threshold, the false signal rate, and the proportion of each trend captured. Look for the period where the separation between trending and ranging signal quality is largest and most consistent. Avoid choosing the period that simply produces the highest overall win rate, as this is likely overfitting.
Wilder chose 14 based on his observation that most significant price cycles in commodity markets ran approximately 28 days. A half-cycle lookback of 14 days captured enough directional movement to be meaningful without excessive lag. This half-cycle logic generalizes reasonably well across other markets and timeframes, which is why 14 has remained the default for decades despite being designed for a specific context most modern traders are not using.