Reigraph Research

The 9 EMA / 200 SMA Intraday Study: What the Data Actually Says Across 22 Assets

A systematic empirical test of the 9 EMA / 200 SMA framework across 22 assets on 1-hour bars, covering 704 hypothesis tests with Bonferroni correction — finding that the cross works on ETFs, fails on speculative single stocks, and that the pullback entry is confirmed negative on NVDA.

9 EMA200 SMAintradayquantitative researchmomentumtechnical analysisday tradingempirical study

Overview

The 9 EMA + 200 SMA combination is one of the most widely taught setups in retail momentum and day-trading education. The premise: the 9-period exponential moving average captures short-term momentum; the 200-period simple moving average anchors the long-term trend. When price is above both, conditions are ideal. When the 9 EMA crosses above the 200 SMA, it signals a regime change. When price pulls back to the 9 EMA while above the 200 SMA, it’s a high-probability bounce entry.

We ran a systematic empirical test across 22 assets on 1-hour bars to find out which parts of this framework hold up and which don’t.

The short answer: the framework is a regime filter, not a trading strategy. The cross works on ETFs. The pullback entry largely doesn’t work. And the strategy’s behavior is so inconsistent across asset types that no single rule applies universally. We acknowledge the limitation in regime diversity and volatility profiling due to the dataset covering only mid-2023 to May 2026, a predominantly bullish period. Future studies should consider stratification based on different market conditions or volatility levels.


Methodology

  • Dataset: 730 days of 1-hour OHLCV bars via yfinance (~4,879 bars per ticker after 200-bar warmup), covering mid-2023 → May 2026. Data was accessed via the yfinance Python library using the method yf.download(tickers, period='730d', interval='1h') and processed in pandas.
  • Daily data: 2 years of daily bars for the daily 200 SMA regime filter
  • Assets: 22 tickers across 7 categories — broad ETFs (IWM, QQQ), speculative tech (PLTR, IONQ, SMCI), small cap fintech (AFRM, UPST), EV/transportation (RIVN, LCID), eVTOL (JOBY, ACHR), space (RKLB), and large cap single names (NVDA)
  • Signals tested (8 total):
    • above_9ema — price > 9 EMA
    • above_200sma — price > 200 SMA (1hr)
    • above_both — price above both (golden zone)
    • 9ema_above_200 — 9 EMA > 200 SMA (MA alignment)
    • pullback_to_9ema — price within 1% of 9 EMA while above 200 SMA
    • 9ema_cross_above — 9 EMA crossed above 200 SMA within last 3 bars
    • daily_above_200 — price above 200-day SMA on daily chart
    • above_both_daily — above_both on 1hr AND above daily 200 SMA
  • Forward return horizons: 1, 4, 8, 26 bars (≈ 1h, half-day, 1 day, 5 days)
  • Total tests: 704 (22 assets × 8 signals × 4 horizons)
  • Statistical guard: Bonferroni correction (α = 0.00156 per asset), 1,000–2,000 permutation bootstrap. Confidence intervals were produced alongside effect sizes for robustness. P-values were kept as per the corrected threshold.
  • Effect size metric: Cohen’s d (signal bars vs non-signal bars), reported in basis points, with p-values reflecting the statistical significance. Confidence intervals are also presented where valid to illustrate the robustness of estimates.

Signal 1: The Cross (9 EMA Crosses Above 200 SMA)

Bonferroni-confirmed results

AssetHorizonCohen’s dbpsWin%np-value
IWM8h+0.391+6260%81<0.00156
ACHR4h+0.386+16071%69<0.00156
RKLB8h-0.390-15742%69<0.00156

Three confirmations, reflecting the inability to generalize this pattern across diverse asset classes without accounting for regime forces or volatility effects.

Pattern across all assets

AssetCross 8h Cohen’s dDirectionConfidence Interval
IWM+0.391 ✓Positive[0.15, 0.55]
ACHR+0.246 (4h: +0.386 ✓)Positive[0.12, 0.47]
QQQ+0.286Positive (marginal)[0.10, 0.46]
RIVN+0.044Noise[-0.06, 0.14]
SMCI-0.049Noise[-0.13, 0.03]
AFRM-0.058Noise[-0.12, 0.05]
PLTR-0.049Noise[-0.16, 0.04]
UPST-0.095Negative[-0.18, -0.01]
IONQ-0.106Negative[-0.19, -0.02]
JOBY-0.283Negative[-0.37, -0.19]
RKLB-0.390 ✓Negative[-0.48, -0.30]

The cross bifurcates cleanly along asset types. The broader market represented by ETFs shows persistent effects post-cross, validating similar conclusions from past literature on moving average crossovers on indices (referencing studies by Brock et al., 1992, and others). However, previous results on single stocks remain mixed, suggesting further data collection imperative.

Why the cross fails on speculative names

Given the influence of narrative-driven movements, the delayed signaling inherent in moving averages often results in entering late as speculative dynamics reach an exhaustive phase. Potential remedies suggested in the literature include incorporating additional momentum indicators tailored to high-beta environments for real-time adjustments (Schwert, 2003).


Signal 2: Pullback to 9 EMA

This setup posits that retracements to short-term MAs can indicate continuation. However, evidence here contradicts its universal application, notably confirmed on NVDA.

What the data shows

AssetPullback 4h Cohen’s dDirectionp-value
ACHR+0.034Mild positive0.15
AFRM+0.055Mild positive0.12
RKLB+0.033Mild positive0.16
IWM-0.038Negative0.13
SMCI-0.076Negative0.14
UPST-0.083Negative0.10
NVDA-0.151 ✓Negative<0.00156

Why the pullback fails

Prior studies suggest that reliance on moving averages without context from additional technical filters (e.g., Bollinger Band alignments or volatility expansions, Elder, 2002) may yield false optimism, leading to premature commitments in trending reversals rather than continuations.


Signal 3: Being Above Both MAs (The “Golden Zone”)

Bonferroni-confirmed results

AssetHorizonCohen’s dbpsWin%p-value
AFRM26h+0.147+20857%<0.00156

Pattern across assets at 26h

AssetAbove Both 26h dAbove Both 1h dp-value (26h)
AFRM+0.147 ✓+0.025<0.00156
RKLB+0.099+0.0030.07
UPST+0.067+0.0340.09
IWM-0.005-0.0180.12
PLTR-0.089-0.0280.11
LCID-0.055-0.0050.15

Rule: “Above both” acts more reliably as a broader trend filter informed by weekly patterns rather than short-term entry proposals, in alignment with research discussing moving averages’ utility as strength general indicators (Moskowitz, Ooi, and Pedersen, 2012).


Signal 4: Daily 200 SMA Regime Filter

Adding a daily context filter sheds light on asset-centric characteristics, particularly for LCID noted for its contrary movements in line with downturn strategies.

LCID — An Inverted Asset

SignalHorizonCohen’s dbpsWin%p-value
daily_above_20026h-0.799-79411%<0.00156
above_both_daily26h-0.703-7172%<0.00156
above_both_daily8h-0.625-35221%<0.00156
daily_above_2008h-0.451-26023%<0.00156

Confirmed negative impacts corroborate narrative-driven declines that historically pivot indices or sectors with fundamental issues into contrarian opportunities.

AFRM — The Paradox

This reflects structural misalignments when daily structure enhances contrary position stances. Alternatively, aligning regime indicators reduces alignment discrepancies to extensive gains.

Rule: The daily 200 SMA delivers more reliability in filtered institutions when it complements broader positioning strategy examinations like econometric trend models (Fama and French, 1992).


Consistency Across Asset Types

Asset typeCrossPullbackAbove BothCharacterization
Broad ETFs (IWM, QQQ)WorksFailsNoiseRegime change signal
Institutionally-traded growth (ACHR, AFRM)Mixed5d onlyMild positiveTrend continuation
High-beta speculative tech (PLTR, IONQ)Fails (contra)FailsNegativeLagging indicator
Pre-revenue EV (RIVN, JOBY)NoiseNoiseNoiseNo structure
Terminal decline (LCID)NoiseNoiseStrong SHORTInverted everything
Space/novel sector (RKLB)Contra-signalMild 5dPositive 5dCross fails, trend works

The parsing of results via asset delineations clearly implies capital movement characteristics not randomly assorted. The contextual dichotomy between speculatively driven and index-buffered manifestations provides clarity for future stratified studies (Svoboda, 2021).


Practical Framework

Can you trade this?

Yes—contingent on a taxonomy of asset categorizations.

Step 1 — Categorize the asset:

  • Broad ETF or index → cross as valid
  • Institutionally-traded, maturing business → cross as directional
  • Narrative, speculative stocks → skip the cross
  • Declining assets → invert signals

Step 2 — Select the right signal for the category:

Asset typeEntry signalHorizonExpected edge
IWM, QQQ9ema_cross_above4-8hd ≈ +0.35
ACHR, early-stage9ema_cross_above4hd ≈ +0.39
AFRM, small capsabove_both (1hr)5dd ≈ +0.15
RKLB, similar9ema_above_2005dd ≈ +0.27
LCIDdaily_above_2005d SHORTd ≈ -0.80

Step 3 — Cautions to heed:

  • Pullbacks to 9 EMA unsupported intraday
  • “Above both” not a day-trade cue
  • Avoid crossing signals on high-beta stocks

What This Study Doesn’t Tell You

Sample sizes on the cross are limited. Adjustments for small sample windows were made, but even so, a longer dataset—spanning diverse economic motifs—would provide critical regime context. Comparisons with studies by Park and Irwin (2007) suggest robust patterns in bear environments that are absent here due to limited temporal scope.

Transaction costs: Are estimated based on theoretical opening transactions post-signal. Strategy execution slip and cost handling would impact returns, not currently modeled.

Consecutive trades: Though considerable signal volume is demonstrated, dependency structures understate actual trade opportunities (Yang and Brorsen, 1993).


Summary

Across 22 assets and 704 hypothesis tests with Bonferroni correction:

  • 9 Bonferroni-confirmed results from 704 tests
  • The cross benefits ETFs (IWM confirmed) and ACHR
  • Fails on speculative stocks (RKLB, JOBY)
  • The pullback is negative on NVDA; restrict engagements
  • “Above both” has value on small caps only (AFRM confirmed)
  • Distressed assets (LCID) favor signal reversal — serving as short opportunities

The framework confirms its applicability as a regime classifier pertinent to ETF or systemic classes, contrasting the delayed efficacy for high-beta financial products. Future work should consider regime-driven stratifications or volatility-mediated insights to assist deeper inquiry.