1. Introduction

1.1 Background

Technical analysis remains the primary decision-making framework for retail traders, with surveys indicating that over 80% of active retail participants rely on at least one technical indicator for entry and exit timing (Menkhoff, 2010). The most commonly cited indicators in retail education — RSI, MACD, Stochastic Oscillator, and Bollinger Bands — form the backbone of introductory trading courses, YouTube tutorials, and fintech platform default configurations.

Despite this widespread adoption, rigorous empirical testing of these indicators at the individual-signal level remains sparse in accessible literature. Academic studies have largely focused on aggregate portfolio returns (Brock et al., 1992; Lo et al., 2000) rather than the signal-level win rates and expected returns that matter most to active traders making discrete entry/exit decisions.

1.2 Research Questions

This study addresses three primary questions:

1. Which individual technical indicators produce statistically significant directional signals in current market conditions?
2. Does combining multiple indicators (confluence trading) improve signal quality, and if so, which combinations are optimal?
3. What role does volume play in signal confirmation, and can volume filtering transform marginal signals into tradeable edges?

1.3 Contribution

Our study contributes to the literature in three ways:

• We test 25 indicators simultaneously across 250 tickers, providing the largest cross-sectional comparison available in accessible format
• We evaluate 45 combination strategies, systematically testing the "confluence" hypothesis popular in retail trading education
• We identify volume confirmation as a dominant signal modifier, quantifying its impact with statistical rigor

2. Methodology

2.1 Data

Universe: 250 U.S. equities selected by options flow activity from the Unusual Whales flow database, representing the most actively-traded names by institutional options participants. The universe includes mega-cap technology (AAPL, MSFT, NVDA, GOOGL, META, AMZN, TSLA), major ETFs (SPY, QQQ, IWM, DIA, XLE, XLF), mid-cap growth (PLTR, COIN, SOFI, HOOD, MSTR, IONQ), and traditional sectors (BA, GS, JPM, UNH, LLY, HD).

Period: April 1, 2025 – April 1, 2026 (252 trading days)

Data Source: Yahoo Finance daily OHLCV, validated against Alpaca Markets data for the top 20 tickers

Forward Return Horizons: 1, 3, 5, 10, and 20 trading days from signal generation

2.2 Indicators Tested

We implemented 25 indicators spanning five categories:

2.3 Signal Definition

Each indicator generates discrete entry signals:

• Buy signal (+1): Indicator transitions from non-bullish to bullish state
• Sell signal (-1): Indicator transitions from non-bearish to bearish state
• Deduplication: Signals within 3 bars of a previous identical signal are suppressed

2.4 Combination Strategy Design

For combination (confluence) strategies, we define 14 continuous indicator states rather than discrete signals:

2.5 Performance Metrics

For each signal, we measure:

• Win Rate (WR): Percentage of signals producing positive forward returns
• Average Return: Mean percentage return across all signals at each horizon
• Median Return: Median percentage return (robust to outliers)
• Sharpe-like ratio: Mean return / standard deviation of returns

2.6 Statistical Significance

With sample sizes ranging from 85 to 16,226 signals per indicator, we apply the following significance threshold: a signal is considered to have genuine predictive power if its 5-day win rate deviates from 50% by more than 2 standard errors.

For n=1,000 and p=0.50, SE ≈ 1.58%, meaning win rates outside [46.8%, 53.2%] are significant at the 95% level.

3. Results

3.1 Individual Indicator Performance