AI Research NIO

NIO 14-day RSI<30: forward 5-day returns vs non-event baseline (last ~3 years)

RSI(14) closing below 30 on NIO tilts toward a short-term bounce, but it’s a fragile tilt rather than a robust signal. Over the trailing ~3 years I flagged days with RSI<30 and measured strict forward 5-day returns: the 18 event days averaged +2.88% while the non-event baseline sat near -0.27%. That raw contrast looks interesting at first glance.

Digging deeper the evidence is mixed and noisy: statistical tests are borderline, the event sample is tiny, volatility on both sides is high, and RSI level has essentially no linear correlation with forward 5-day outcomes. The full report below shows the distributions, p-values, and a scatter of RSI versus forward returns so you can judge whether this is a tradable edge or mostly folklore.

The research question

For NIO over the past ~3 years, does the textbook RSI-below-30 'oversold' signal actually mark tradeable bounces, or does oversold just keep bleeding in a downtrend? Thesis: forward 5-day returns after the 14-day RSI closes under 30 come out statistically indistinguishable from — if not worse than — the everyday baseline, so 'oversold equals buy' is folklore in a weak name where weakness begets weakness rather than reversion.

How this was measured

Resampled NIO minute bars to daily closes over the trailing ~3 years (anchored to the latest available date). Computed 14-day RSI via Wilder-style EWMA smoothing (alpha=1/14). Flagged event days where RSI closed below 30 and measured the strict forward 5-trading-day return: close[t+5]/close[t]-1. Compared the event-day forward-return distribution to a non-event baseline (days with RSI≥30) using Welch's two-sample t-test, and reported a rank-based Mann–Whitney U p-value as a robustness check. A scatter shows the continuous relationship between RSI level and forward 5-day returns.

The key numbers

Analysis window (trading days)
752
2023-07-21 to 2026-06-23
Valid days for test (RSI+fwd5d)
733
18 events; 715 non-events
Event mean fwd 5d return
2.8819%
N=18 RSI<30 days
Event median fwd 5d return
2.8946%
Event fraction positive
66.67%
Baseline mean fwd 5d return
-0.2744%
N=715 non-event days
Baseline median fwd 5d return
-1.1189%
Baseline fraction positive
43.64%
Mean difference (event - baseline)
3.1563%
Welch t-statistic
1.716
Two-sided; positive favors RSI<30
Welch p-value
0.1030
p=0.1030 ≥ 0.05 → no statistically-clear difference
Mann–Whitney U p-value
0.0560
Rank-based, two-sided
Pearson r (RSI vs fwd 5d)
0.004
|r|=0.004 ≤ 0.3 → weak association

Reading the numbers

When RSI closed below 30 (18 days) the average forward 5‑day return was +2.88% vs a baseline mean of −0.27% (715 days). The raw gap is +3.16% but the two-sided Welch p=0.103 means that difference is not statistically clear.

The charts

Forward 5-day return: RSI<30 vs non-event days
What this chart says

This box plot lines up the 5‑day returns after RSI<30 (n=18) against the 715 non‑event days. The center of the RSI<30 group sits around +2.88% with outcomes spanning −10.71% to +18.73%, while non‑events center near −0.27% and span −20.88% to +41.24%. The eye should go to that higher central value for RSI<30, but the ranges overlap and the small event sample means the apparent advantage may not be reliable in practice.

Forward 5-day returns after RSI<30 events
What this chart says

The histogram shows the 18 individual forward‑5d outcomes after RSI<30: mean +2.88%, min −10.71%, max +18.73%, and 12 of 18 events were positive (66.7%). What stands out is dispersion — a few sizeable positive outcomes push the average up while several negatives persist — so a majority of events win but the sample is small and noisy, limiting confidence that each RSI<30 is a dependable bounce signal.

RSI level vs forward 5-day return (all valid days)
What this chart says

This scatter plots RSI‑14 versus forward 5‑day return across all 733 valid days: RSI runs from 24.70 to 87.80 (mean 48.14) while the average 5‑day return overall is about −0.20%. Look at the low‑RSI zone near 24.7 — returns there are mixed, with both notable losses and gains, so low RSI does not map cleanly to guaranteed reversion. Combined with the statistical tests (mean gap +3.16% but Welch p=0.103 and Mann–Whitney p=0.056), the plot supports the thesis that oversold readings in this name are noisy signals rather than a consistent trigger for bounce trades.

Forward 5-day return summary (events vs baseline)

groupNmeanmedianstdfraction_positive
RSI<30 (events)180.02880.02890.07660.6667
Non-event baseline715-0.0027-0.01120.09460.4364

Most recent RSI<30 events (last 20)

datersi14fwd_5d_return
2025-12-0325.210.0482
2025-12-0228.29-0.0069
2025-12-0129.6-0.0097
2025-11-2024.70.0261
2025-04-0827.410.1564
2024-04-1929.060.1873
2024-04-1729.150.0384
2024-04-1625.950.0707
2024-04-1527.220.0179
2024-04-1229.79-0.0779
2024-04-0829.46-0.1071
2024-04-0529.62-0.0659
2024-03-2828.88-0.0265
2024-02-0528.370.1257
2024-02-0229.660.065
2024-01-3129.90.0388
2024-01-2228.060.0318
2024-01-1929.10.0065

The takeaway

Short answer: RSI(14) closing under 30 on NIO tilts toward a small 5-day bounce, but the evidence is suggestive rather than conclusive. The 18 RSI<30 days averaged +2.88% forward 5-day return (median +2.89% and 66.7% positive) versus the 715 non-event days which averaged -0.27% (median -1.12% and 43.6% positive) — a raw mean gap of about 3.16 percentage points. Statistical tests are mixed and weak: Welch p = 0.103 (roughly a 10-in-100 chance this gap is luck) and Mann–Whitney p = 0.056 (roughly a 5.6-in-100 chance) — borderline at best, not a clean reject of no effect. The signal is noisy: event returns have a standard deviation of ~7.66% (baseline ~9.46%) and the Pearson correlation between RSI and forward 5-day return is essentially zero (r ≈ 0.0039), so lower RSI doesn’t consistently predict larger short-term gains. Practical takeaway: RSI<30 produced more positive 5-day outcomes in this window, but with only 18 events, high variance, and marginal p-values, it’s a weak, risky edge — not a reliable “oversold = buy” rule on its own.

The fine print