The question behind the Doji
The Doji candlestick forms when a security closes near its opening price. In traditional technical analysis, this narrow real body is interpreted as a sign of indecision between buyers and sellers.
Traders frequently associate Doji patterns with potential trend reversals. But an attractive chart pattern does not necessarily imply a statistically reliable trading signal. To investigate, we examined SPY daily prices from January 2000 through September 2026.
For the traditional pattern explanation and annotated trading examples, see our original article: Magic Doji or Myth: How to Use Doji for Stock Trading.
How we defined and tested a Doji
A qualifying Doji was defined as a trading session where the absolute difference between the opening and closing price was no more than 10% of the session's high-low range. Zero-range sessions were excluded.
This identified 676 Doji sessions among 6,726 daily SPY observations, or approximately 10.05% of the sample. We evaluated subsequent close-to-close price returns over 1, 3, 5, and 10 trading sessions.
We also separated observations according to SPY's preceding five-session return: declines exceeding 2%, neutral moves within ±2%, and advances exceeding 2%. This allowed us to investigate whether a Doji's apparent predictive value changes with the preceding market trend.
First, does the Doji alone provide an advantage?
Across the full sample, Doji sessions showed only small differences from the broader market at short horizons. The largest observed directional difference appeared over 10 trading sessions.
Does a Doji alone predict positive SPY returns?
Frequency of positive forward returns after Doji sessions versus all SPY sessions.
These comparisons include Dojis in the all-session baseline. They describe historical outcomes, not independent trading signals.
Finding 1: Market context matters
A Doji does not have a uniform historical interpretation. The direction of the preceding market movement changes the distribution of subsequent outcomes.
Market context changes the Doji outcome
10-day positive-return frequency by preceding five-day SPY movement.
80 qualifying Dojis
494 qualifying Dojis
100 qualifying Dojis
The strongest directional result in our exploratory analysis appeared after Doji sessions in relatively neutral markets. However, the chart above compares Dojis with all sessions in the same market category, including other Dojis. It is descriptive rather than an independent non-Doji control test.
Finding 2: The neutral-market result persisted across decades
To examine historical stability, we compared qualifying neutral-market Dojis with neutral-market sessions that did not meet our Doji definition.
Across the full eligible sample, 494 neutral-market Doji observations with complete 10-day outcomes had a positive-return frequency approximately 7.35 percentage points higher than the corresponding non-Doji control.
Does the Doji advantage persist across market periods?
10-day positive-return frequency following neutral-market sessions.
158 Doji observations · 1459 non-Doji observations
215 Doji observations · 1675 non-Doji observations
121 Doji observations · 999 non-Doji observations
The directional advantage appeared in all three historical periods, but decreased from approximately 9.56 percentage points in 2000–2009 to 3.48 percentage points in 2020–September 2026.
Importantly, a higher frequency of positive outcomes does not necessarily mean a higher average return. In the most recent period, neutral-market Dojis had a higher positive-return frequency but a lower average 10-day return than non-Doji sessions.
Finding 3: The result survived several bootstrap specifications
Forward 10-day returns overlap, meaning adjacent observations cannot reasonably be treated as fully independent. We therefore used moving-block bootstrap resampling to examine the uncertainty around the neutral-market directional advantage.
How sensitive is the finding to bootstrap block length?
95% confidence intervals for the 10-day positive-return advantage.
The tested 10-, 20-, and 40-session bootstrap block lengths produced confidence intervals above zero. This supports the historical association under the tested specifications, but does not establish a profitable or independently validated trading strategy.
Does waiting for confirmation help?
We also tested whether a breakout beyond the previous Doji's high or low improved subsequent directional outcomes. Returns were measured from the confirmation session's closing price.
Does confirmation improve directional accuracy?
Five-day directional success following a breakout from the preceding Doji range.
239 Doji confirmations. Bullish success means a positive return; bearish success means a negative return.
185 Doji confirmations. Bullish success means a positive return; bearish success means a negative return.
Bullish confirmations showed mixed results across holding periods. Bearish confirmations were particularly unconvincing at the five-day horizon, where fewer observations subsequently produced negative returns than the broader downside-breakout comparison.
This challenges the assumption that a Doji followed by a downside breakout necessarily produces a reliable bearish signal.
Which findings remain statistically distinguishable?
We compared selected Doji outcomes with eligible non-Doji controls using moving-block bootstrap resampling. The results varied substantially depending on the signal definition.
Which effects are distinguishable from zero?
Observed differences and exploratory 95% moving-block bootstrap intervals.
The neutral-market 10-day positive-return advantage had a confidence interval above zero. The bullish one-day confirmation interval crossed zero, while the bearish five-day downside-frequency comparison was negative.
These exploratory tests were selected after examining multiple specifications. Their confidence intervals should not be interpreted as confirmatory evidence without adjustment and independent validation.
Does a Doji predict reversal or continuation?
A Doji is frequently interpreted as a potential turning point. But a reversal requires more than a candle's appearance. We tested whether Doji sessions were followed by a directional change or continuation more often than comparable non-Doji sessions.
We classified the preceding trend using the prior five-session SPY return. An increase greater than 2% defined an uptrend, while a decline greater than 2% defined a downtrend. Each Doji was compared with non-Doji sessions from the same preceding-trend category.
Reversal or continuation?
Doji advantage over non-Doji sessions in the same prior-trend regime. Positive values favor Doji.
How reliable are the differences?
Ten-day directional success-rate advantage and exploratory 95% moving-block bootstrap confidence intervals.
| Signal classification | Advantage | 95% CI (pp) |
|---|---|---|
| Bullish reversal | −2.35 pp | −13.30 to +8.80 |
| Bearish reversal | +1.03 pp | −8.64 to +10.41 |
| Bullish continuation | −0.93 pp | −10.27 to +8.39 |
| Bearish continuation | +2.35 pp | −8.67 to +14.06 |
All four 95% confidence intervals include zero. These results do not establish a reliable reversal or continuation advantage at the ten-day horizon.
Source: TNI analysis of SPY daily prices, 2000–2026. Twenty-session moving-block bootstrap, 2,000 replications. Intervals are exploratory and unadjusted for multiple testing.
At the ten-day horizon, bullish reversals showed a 2.35-percentage-point disadvantage, while bearish reversals showed a 1.03-point advantage. Neither result was statistically distinguishable from zero under the exploratory bootstrap.
Continuation outcomes were similarly inconclusive. These classifications measure the direction of the forward return, not whether a sustained new market trend developed. Reversal and continuation outcomes are also closely related, rather than four independent tests.
What this study does not establish
These results are exploratory. Multiple Doji thresholds, forward horizons, market regimes, and confirmation definitions were examined. Reported bootstrap probabilities and confidence intervals have not been adjusted for this broader search.
The analysis uses unadjusted SPY daily prices and close-to-close price returns, excluding dividends. It does not include commissions, bid-ask spreads, slippage, short-sale constraints, or portfolio-level position sizing. No untouched out-of-sample trading period was reserved for final validation.
Consequently, the observed association should be treated as a research hypothesis rather than evidence that Doji trading alone produces excess returns.
The bottom line
The evidence does not support treating every Doji as an automatic reversal signal. Instead, the historical results suggest that preceding market conditions influence how the pattern relates to subsequent SPY returns.
Our most interesting finding is the higher frequency of positive 10-day returns following Dojis in neutral-market conditions. Its declining advantage across decades, however, reinforces the need for caution and independent prospective testing.