The hypothesis for this project
Patterns on a random walk
Traders see setups in candles, RSI, moving-average crosses, MACD, and Bollinger Bands. Those shapes also appear in pure noise. You can pick the windows that worked and the chart looks real.
The fake market
The symbol is NOIZ. It is not a currency, a coin, or an index. Each one-minute close is the last close times a random shock. The expected log move is zero. There is no news, no order book, and no edge.
We first build 1,051,200 one-minute bars. That is 730 days, 24 hours a day, from 2024-08-20 to 2026-08-20 UTC. We then group those minutes into 15-minute, 1-hour, 4-hour, and daily bars. Open is the first print. High is the max. Low is the min. Close is the last print.
The generator uses seed 20260820. Annual volatility is 18 percent. The start price is 100.00. Anyone can rebuild the same path.
1 minute
1,051,200
bars
15 minute
70,080
bars
1 hour
17,520
bars
4 hour
4,380
bars
Daily bars
730
full sample
Start → end
100.00 → 94.07
-5.9%
Range
83.80 – 118.50
high / low of the walk
Seed
20260820
fixed, reproducible
Generated series
NOIZ looks like a market
Daily shows the full two years. Lower timeframes show a recent slice so the candles stay readable. The hidden history is still the same random walk.
730 daily bars · 2024-08-20 to 2026-08-19
Twelve popular rules
We scanned the same NOIZ path with rules that show up in retail charts and trading books. Each rule has a fixed entry and a fixed hold. We do not trail, scale, or skip losers.
- RSI 14 oversold bounce and overbought fade
- SMA 20 / 50 golden cross and death cross
- MACD line cross of its signal line, both ways
- Bollinger 20, 2-sigma band bounce and fade
- Bullish engulfing, bearish engulfing, hammer, and shooting star
The scan uses 15-minute, 1-hour, 4-hour, and daily bars. That is 51,328 signals. 25,591 of them made money after the fixed hold. The win rate is 49.9 percent. That is about half.
The full sample first
Before the pretty charts, here is every signal. Average result sits near zero. The best window is large. The worst window is large too. That gap is what cherry picking uses.
| Rule | Signals | Win rate | Average | Best | Worst |
|---|---|---|---|---|---|
| RSI oversold bounce | 1,209 | 50.4% | -0.00% | +6.07% | -5.64% |
| RSI overbought fade | 1,245 | 49.6% | -0.02% | +4.50% | -7.09% |
| SMA 20 / 50 golden cross | 1,036 | 48.9% | -0.03% | +5.44% | -4.34% |
| SMA 20 / 50 death cross | 1,036 | 48.9% | -0.00% | +5.47% | -6.27% |
| MACD bullish cross | 3,817 | 49.6% | -0.00% | +8.16% | -6.36% |
| MACD bearish cross | 3,815 | 50.5% | +0.00% | +5.08% | -3.81% |
| Bollinger lower-band bounce | 2,866 | 50.7% | +0.00% | +8.87% | -5.74% |
| Bollinger upper-band fade | 2,813 | 50.2% | -0.00% | +3.84% | -9.26% |
| Bullish engulfing | 11,581 | 49.2% | -0.00% | +7.35% | -6.96% |
| Bearish engulfing | 11,532 | 49.9% | +0.00% | +5.51% | -6.93% |
| Hammer reversal | 5,162 | 49.5% | -0.00% | +7.05% | -4.59% |
| Shooting star | 5,216 | 51.1% | +0.01% | +5.86% | -7.96% |
Now the cherry picks
For each rule we kept one winning 4-hour window when we could. We skipped nearby times so the gallery is not one swing shown twelve ways. The hold is the same rule used in the table. We chose the sample after we saw the result.
If you post only these charts, NOIZ looks tradable. The table above is the rest of the data.
Cherry-picked RSI
RSI oversold bounce
Buy when 14-period RSI crosses back up through 30. Exit 12 bars later.
NOIZ · 4 hour · Entry Dec 23, 2024 04:00 UTC · Hold 12 bars
Cherry-picked RSI
RSI overbought fade
Sell when 14-period RSI crosses back down through 70. Exit 12 bars later.
NOIZ · 4 hour · Entry Jan 3, 2025 00:00 UTC · Hold 12 bars
Cherry-picked Moving average
SMA 20 / 50 golden cross
Buy when SMA 20 crosses above SMA 50. Exit 16 bars later.
NOIZ · 4 hour · Entry Nov 4, 2024 00:00 UTC · Hold 16 bars
Cherry-picked Moving average
SMA 20 / 50 death cross
Sell when SMA 20 crosses below SMA 50. Exit 16 bars later.
NOIZ · 4 hour · Entry Dec 19, 2024 04:00 UTC · Hold 16 bars
Cherry-picked MACD
MACD bullish cross
Buy when the MACD line crosses above its signal line. Exit 12 bars later.
NOIZ · 4 hour · Entry May 23, 2026 08:00 UTC · Hold 12 bars
Cherry-picked MACD
MACD bearish cross
Sell when the MACD line crosses below its signal line. Exit 12 bars later.
NOIZ · 4 hour · Entry Feb 12, 2026 16:00 UTC · Hold 12 bars
Cherry-picked Bollinger
Bollinger lower-band bounce
Buy when price closes back above the lower 20, 2-sigma band. Exit 10 bars later.
NOIZ · 4 hour · Entry Aug 2, 2026 12:00 UTC · Hold 10 bars
Cherry-picked Bollinger
Bollinger upper-band fade
Sell when price closes back below the upper 20, 2-sigma band. Exit 10 bars later.
NOIZ · 4 hour · Entry Sep 1, 2025 00:00 UTC · Hold 10 bars
Cherry-picked Candlestick
Bullish engulfing
Buy after a down candle whose body is wrapped by the next up candle. Exit 8 bars later.
NOIZ · 4 hour · Entry Jul 31, 2025 00:00 UTC · Hold 8 bars
Cherry-picked Candlestick
Bearish engulfing
Sell after an up candle whose body is wrapped by the next down candle. Exit 8 bars later.
NOIZ · 4 hour · Entry Feb 6, 2025 12:00 UTC · Hold 8 bars
Cherry-picked Candlestick
Hammer reversal
Buy a hammer: small body, long lower wick, little upper wick. Exit 8 bars later.
NOIZ · 4 hour · Entry Oct 24, 2025 20:00 UTC · Hold 8 bars
Cherry-picked Candlestick
Shooting star
Sell a shooting star: small body, long upper wick, little lower wick. Exit 8 bars later.
NOIZ · 4 hour · Entry Aug 5, 2026 00:00 UTC · Hold 8 bars
What this does not prove
This does not prove that every rule fails on real markets. It proves a weaker and more useful point. Success on a chart is cheap when you choose the sample after you see the result.
A random walk still trends for a while. It still prints hammers, crosses, and RSI bounces. After the fact, those shapes are easy to name. The naming does not give the next bar an edge.
Hypothesis paths show turn-timing shapes. The product does not predict direction. This is not investment advice. SabioLabs, LLC.