I pulled every trading day from 2025 and analyzed which days of the week performed best. The results were clear: Wednesday dominated, Thursday underperformed, and the pattern was consistent across multiple stocks.
More interesting: I tested a simple strategy—buy Thursday, sell Wednesday. On Palantir, it returned 180% versus 124% for buy and hold. On the Nasdaq, it returned 41% versus 21%.
This is pure data. No predictions. Just what actually happened in 2025.
I analyzed six tickers: SPY, QQQ, PLTR, GOOGL, TSLA, and NVDA. Every trading day from January 1 to December 2, 2025. That's 264-288 observations per day of the week, depending on holidays.
For each day, I calculated:
Then I aggregated across all six tickers to find the overall pattern.
| Day | Avg Return | Win Rate | Up Days | Down Days | Total |
|---|---|---|---|---|---|
| Wednesday | +0.78% | 62.8% | 177 | 105 | 282 |
| Monday | +0.37% | 59.1% | 156 | 108 | 264 |
| Friday | +0.26% | 55.1% | 152 | 124 | 276 |
| Tuesday | +0.02% | 49.0% | 141 | 147 | 288 |
| Thursday | -0.49% | 50.0% | 132 | 132 | 264 |
| Day | Avg Return | Win Rate |
|---|---|---|
| Wednesday | +0.39% | 70.2% |
| Monday | +0.23% | 65.9% |
| Tuesday | +0.02% | 50.0% |
| Friday | -0.05% | 47.8% |
| Thursday | -0.23% | 54.5% |
| Day | Avg Return | Win Rate |
|---|---|---|
| Wednesday | +0.52% | 72.3% |
| Monday | +0.31% | 63.6% |
| Tuesday | +0.00% | 47.9% |
| Friday | -0.05% | 56.5% |
| Thursday | -0.31% | 50.0% |
| Day | Avg Return | Win Rate |
|---|---|---|
| Wednesday | +0.94% | 57.4% |
| Monday | +0.62% | 56.8% |
| Tuesday | +0.60% | 60.4% |
| Friday | +0.56% | 56.5% |
| Thursday | -0.51% | 50.0% |
| Day | Avg Return | Win Rate |
|---|---|---|
| Wednesday | +1.15% | 61.7% |
| Friday | +0.90% | 56.5% |
| Monday | +0.53% | 59.1% |
| Tuesday | -0.51% | 45.8% |
| Thursday | -1.44% | 29.5% |
This isn't cherry-picking. The pattern holds across different types of stocks:
Out of six tickers, Wednesday was the best or second-best day for five of them. Thursday was the worst or second-worst for five of them.
Knowing the pattern exists is interesting. But can you trade it?
I tested three strategies:
Capture the mid-week strength.
| Ticker | Return | Trades | vs Buy & Hold |
|---|---|---|---|
| SPY | +16.7% | 87 | +1.1% |
| QQQ | +21.7% | 87 | +0.8% |
| PLTR | +86.0% | 87 | -38.0% |
Works for indexes. Underperforms on strong uptrending stocks like PLTR because you're out of the market too often.
Buy after Thursday weakness, sell at Wednesday peak.
| Ticker | Return | Trades | vs Buy & Hold |
|---|---|---|---|
| SPY | +30.9% | 91 | +15.3% |
| QQQ | +40.9% | 91 | +20.0% |
| PLTR | +124.0% | 91 | +0.0% |
Buy at the weekly low, sell at the weekly high.
| Ticker | Return | Trades | vs Buy & Hold |
|---|---|---|---|
| SPY | +27.3% | 90 | +11.7% |
| QQQ | +35.4% | 90 | +14.5% |
| PLTR | +180.2% | 90 | +56.2% |
I don't know for certain. But here are some possibilities:
Wednesday Strength:
Thursday Weakness:
These are guesses. The data shows the pattern exists. It doesn't tell us why.
Day-of-week effects have been studied for decades. The classic findings:
My 2025 data shows different patterns. Monday was positive (+0.37%). Friday was barely positive (+0.26%). Wednesday dominated.
This could mean:
After publishing the initial analysis, I checked something important: mean versus median.
The mean (average) can be skewed by outliers. The median (middle value) is more robust. If they differ significantly, it means a few extreme days are driving the pattern.
Here's what I found:
| Day | Mean | Median | Difference |
|---|---|---|---|
| Monday | +0.375% | +0.361% | +0.014% |
| Tuesday | +0.023% | -0.038% | +0.061% |
| Wednesday | +0.783% | +0.344% | +0.439% |
| Thursday | -0.488% | -0.004% | -0.483% |
| Friday | +0.261% | +0.317% | -0.056% |
Wednesday's advantage shrinks dramatically. The mean shows +0.78%, but the median shows +0.34%—barely better than Monday (+0.36%) or Friday (+0.32%). A few big Wednesday gains skew the average.
Thursday's weakness is also overstated. The mean is -0.49%, but the median is essentially flat at -0.00%. Most Thursdays are neutral. A few big drops pull the average down.
The days are actually much closer. Looking at medians, Monday/Wednesday/Friday are all around +0.3-0.4%. The dramatic differences in the mean are driven by extreme days.
| Rank | Day | Median Return |
|---|---|---|
| 1 | Monday | +0.361% |
| 2 | Wednesday | +0.344% |
| 3 | Friday | +0.317% |
| 4 | Thursday | -0.004% |
| 5 | Tuesday | -0.038% |
The ranking changes completely. Monday edges out Wednesday. The top three days are nearly identical.
One pattern holds up: Tesla's Thursday weakness. Mean: -1.44%. Median: -1.08%. Even removing outliers, Tesla consistently underperforms on Thursday. This appears to be a real, consistent pattern.
The buy-Thursday-sell-Wednesday strategy still worked in 2025 (180% on PLTR, 41% on QQQ). But the underlying pattern is weaker than it first appeared.
You're not capturing consistent daily differences. You're hoping to catch the occasional big Wednesday gain or avoid the occasional big Thursday drop.
That's less reliable. More like gambling on outliers than exploiting a consistent edge.
Before you start trading this pattern, understand the limitations:
1. One Year of Data
This is 2025 only. The pattern might not persist in 2026 or beyond.
2. Transaction Costs
These strategies require 87-91 trades per year. Even at zero commission, there's bid-ask spread and slippage. That adds up.
3. Taxes
Every trade triggers short-term capital gains. That's taxed at your income rate (up to 37%). Buy and hold gets long-term rates (15-20%).
4. Execution
I used closing prices. Real trades happen at market prices. You'll get worse fills.
5. Overfitting
I found the optimal strategy on the same data I tested it on. It will perform worse on new data.
6. Market Regime
2025 was mostly bullish. Bear markets behave differently.
7. Psychology
Can you actually execute 90 trades per year mechanically? Most people can't.
For most people: no.
The pattern is real. The data is clear. But turning it into consistent profits is hard:
For day traders or algorithmic traders with low costs and tax-advantaged accounts, it might be worth testing. For everyone else, it's more interesting than actionable.
After analyzing every trading day in 2025, here's what matters:
1. The pattern is real. Wednesday outperformed. Thursday underperformed. Consistently.
2. It's tradeable in theory. Buy Thursday, sell Wednesday beat buy and hold on indexes and PLTR.
3. It's hard in practice. Costs, taxes, and execution make it difficult for most investors.
4. It might not last. One year of data isn't enough to build a strategy on.
5. It's worth monitoring. If the pattern persists into 2026, it becomes more interesting.
In 2025, Wednesday had the highest average return (+0.78%) and Thursday the lowest (-0.49%). But median analysis shows the pattern is driven by outliers. Remove extreme days, and Wednesday/Monday/Friday are nearly identical (+0.3-0.4%).
A buy-Thursday-sell-Wednesday strategy still outperformed: SPY (+15%), QQQ (+20%), PLTR (+56%). But you're not exploiting consistent daily differences—you're betting on catching occasional big moves.
This is one year of data with significant limitations. Transaction costs, taxes, and execution challenges make it hard to profit from. The median analysis makes it even less compelling.
Interesting pattern. Weaker than it first appeared. Not investment advice.
Final Thought: The initial analysis showed Wednesday averaging +0.78% and Thursday -0.49%. But median analysis reveals the pattern is driven by outliers. Most days, Wednesday/Monday/Friday are nearly identical (+0.3-0.4%). The dramatic differences come from a few extreme days.
The trading strategies still worked in 2025. But you're not exploiting a consistent edge—you're betting on catching occasional big moves. That's less reliable than it first appeared.
This is why you dig deeper. The surface-level pattern looked strong. The median analysis shows it's weaker. Always question your first conclusion.