Backtesting Divergence on NASDAQ 100 — What 5 Years of Data Shows

A systematic backtest of a multi-indicator divergence strategy across NASDAQ 100 stocks over five years — win rates, profit factors, drawdown, and the structural lessons that change how you think about divergence trading.

Why Backtest Divergence?

Divergence patterns are everywhere in trading content — YouTube, forums, social media. The problem is that most of what is shared is cherry-picked. A chart is shown after the fact, the divergence is circled, and the reversal looks obvious. Real trading doesn’t work that way. To understand whether divergence generates a genuine systematic edge, you need to run it forward through historical data without looking at the outcome first.

This post documents what a systematic divergence backtest on NASDAQ 100 stocks actually shows — not cherry-picked examples, but aggregate results across hundreds of signals over five years.

The System Being Tested

The system tested combines three elements:

Signal detection: Regular divergence using swing pivot detection on daily bars. A bullish divergence requires price to make a lower low while RSI, Stochastic, and MACD all independently show a higher low on the same pivot date. All three indicators must agree — single-indicator divergence is ignored.

Trend filter: Bearish divergence signals are suppressed on stocks where the close is above the 50-period EMA, and the 50-period EMA is above the 200-period EMA. In an established uptrend, bearish divergence signals have a poor track record (consistent with Bulkowski’s research).

Entry and exit: Entry is confirmed by price direction on the following session. Stop is placed at the prior session’s low (for long trades) or high (for short trades) — the most recent structural level, not an arbitrary ATR multiple. Target is 3× the ATR from entry. An ATR-based trailing stop activates if the trade moves 2× ATR in the intended direction, locking in 1× ATR of profit.

Capital and position sizing:

Key Results

Across 54 NASDAQ 100 stocks, five-year backtest:

A few things stand out immediately.

The win rate is low by most traders’ standards. 37% means you lose more trades than you win — 6 losses for every 4 wins. Most people would find this psychologically challenging to trade, especially during the inevitable losing streaks (13 consecutive losses occurred in this backtest). Yet the system is solidly profitable.

The profit factor of 3.3× explains why. A profit factor of 3.3× means for every dollar lost across all losing trades, the system earns $3.30 across all winning trades. This asymmetry is the core of the strategy — losses are cut quickly, winners are held to large targets.

8% maximum drawdown is manageable. With 1% capital risked per trade and a maximum of 5 concurrent positions, the system never risked more than 5% of capital at any one time. The 8% drawdown reflects compounding of sequential losses during the worst period.

What Drives the Edge

Exit breakdown matters more than entry

Across the 197 trades, the exits split as:

The 7% of trades that hit the full target and the 8% that hit the trailing stop are doing the heavy lifting on the profit side. This is structurally similar to how trend-following systems work — many small losses, a few large wins that define the overall P&L.

The trailing stop is significant. Adding a trailing stop mechanism (activates at +2× ATR, locks in 1× ATR of profit) produced a measurable improvement: maximum drawdown reduced substantially while P&L improved. The trailing stop converts some would-be full-target winners into smaller but guaranteed wins, and prevents some winners from reversing into breakeven or losses.

Pivot sensitivity is critical

One of the most counter-intuitive findings: using a more sensitive pivot detection (detecting pivots that only need to be the highest/lowest point within 1 bar on each side) significantly outperforms using a more conservative pivot detection (requiring 5 bars on each side). The reason is that waiting for a “major” pivot means waiting for confirmation that is already 5 bars stale. By then, the reversal has often already moved a substantial distance from the ideal entry point.

This runs counter to the common advice that “only trade off major swing points.” In a daily-bar, systematic context, faster pivot detection combined with confirmation filters (indicator confluence + entry direction) produces better results than waiting for the largest structural pivots.

Time stops reveal important structural information

When a trade has neither hit its stop nor its target after a fixed number of bars (tested at 5 bars), a time stop exits the position. This group of trades is instructive: the majority of time-stop exits were profitable at exit. The move happened, but slowly — not fast enough to hit the 3× ATR target in the allotted time. When this exit group was allowed to run longer (tested at 10 bars), performance actually worsened: some trades that had been profitable at bar 5 reversed and hit the stop loss. The bar-5 time stop was capturing real, fleeting profit.

The Structural Lessons

1. Win rate is the wrong success metric for divergence. If you evaluate a divergence strategy by its win rate, you will always be disappointed and tempted to add more filters. The correct metrics are profit factor and average R per trade. A 37.6% win rate with +0.954R average and 3.34× profit factor is a sound, profitable system — the numbers prove it.

2. Market context matters more than signal quality. The backtest covered a predominantly bull-market period (2019–2024). Bullish divergence signals drove the majority of profit. Bearish divergence signals in the same period were a drag. In a bear market, the opposite would likely be true. Divergence strategies need to be deployed with awareness of the broader market regime, not run blindly regardless of conditions.

3. The stop placement source matters enormously. Using the most recent structural low or high (the prior session’s candle extreme) as the stop source produces better results than a fixed ATR multiple, because it is anchored to actual price structure — the point where the thesis is definitively wrong. An ATR-based stop placed without reference to structure can be too tight (triggering on normal volatility) or too wide (accepting larger losses than necessary).

4. More filters do not mean better results. Adding more confluence requirements, additional RSI gates, or extra condition checks does not automatically improve performance. In this backtest, several seemingly logical additions (requiring RSI to be in extreme territory, requiring minimum pivot spacing) either had no measurable effect or slightly reduced performance by removing productive signals along with unproductive ones. Each proposed filter should be tested independently with a full backtest before assuming it helps.

5. Drawdown control comes from position sizing, not from the signal system. The 8% maximum drawdown was achieved through consistent 1% capital risk per trade and a 5-position maximum, not through having a high win rate or tight stops. The signal system determines when to trade; position sizing determines how bad the inevitable losing streaks feel. Get position sizing right first.

What This Means for Manual Traders

If you are trading divergence manually rather than systematically, the backtesting research suggests a practical framework:

Be selective about market regime. In a bull market, focus primarily on bullish divergence setups. Bearish divergence in a bull market has a poor probability-weighted outcome.

Apply the RSI zone filter strictly. Only consider setups where RSI reached a genuine extreme (below 35 or above 65) at the first pivot. Divergence where RSI stayed in the 40–60 zone is consistently weaker.

Require at least two indicators to confirm. RSI divergence alone can be a false positive. Requiring Stochastic or MACD to independently show the same divergence on the same pivot removes a large portion of noise signals.

Cut losses at the structural level, not based on a percentage. The stop goes at the candle low (for longs) or candle high (for shorts) of the divergence pivot — the price level where the pattern is unambiguously wrong.

Let winners run to a meaningful target. The strategy’s edge depends on asymmetric reward. If you take profit at 1:1 or 1.5:1, you undermine the profit factor that makes the win rate tolerable.

This post is for educational purposes only and does not constitute financial advice.