A technical indicator does not create new information. It transforms historical price, volume, volatility, or trading-range data into another representation.
A moving average smooths the price. RSI summarizes the balance between recent gains and losses. Bollinger Bands compare the current price with a recent average and volatility range. These indicators may make the data easier to interpret, but they cannot manufacture predictive information that was not present in the underlying data.
The largest problem begins when traders search for the settings that produced the strongest historical result.
There Is No Single RSI or Moving-Average Strategy
Consider an apparently simple RSI strategy. The investor must still choose the RSI period, entry and exit thresholds, whether to trade long only or in both directions, the holding period, the stop-loss, the profit target, and any trend or volatility filters.
Changing any of those choices creates a different strategy. Even a simplified setup with dozens of choices per parameter can create millions of possible RSI strategies, and that is before combining multiple indicators or testing several markets and timeframes.
Combining Indicators Makes the Search Space Explode
Suppose a trader has 100 possible indicators or filters available. A strategy using one indicator has 100 broad choices. A strategy chaining together five indicators can already produce up to 10 billion broad combinations, even before considering each indicator's internal settings.
A moving-average filter, RSI trigger, volatility screen, stop-loss rule, and trailing exit may each offer dozens or hundreds of variants. The actual strategy universe becomes far larger than the number of tools visible on a trading platform.
With Enough Tests, Strong Backtests Appear by Chance
Imagine testing 1,000 strategies that have no real predictive ability. Some will lose, many will look average, and a few will appear surprisingly strong simply because the historical sample happened to favor them.
Expand that process to 100,000, 10 million, or 1 billion strategies and the strongest result will look increasingly impressive even if none of the underlying strategies has genuine edge. The winner can be luck, not skill.
A Sharpe Ratio of 2 May Be the Winner of a Search
Suppose a researcher tests an enormous number of technical strategies over ten years of data. Even if all of them have zero true alpha, the best one may still show high returns, limited drawdowns, and an apparently strong Sharpe ratio.
The final winner is usually selected partly because it received unusually favorable historical luck. When that luck disappears on new data, performance often falls sharply. What looks like decay may simply be the disappearance of luck that was mistaken for alpha.
The Final Strategy Hides the Failed Alternatives
A published strategy might say: buy when the 37-day moving average crosses above the 91-day moving average and volatility is below its 120-day average. Presented that way, it can look as though those exact values were chosen for strong theoretical reasons.
But the actual process may have involved hundreds of short and long moving averages, multiple volatility windows, delays, exits, and backtest periods. The final strategy is only the surviving winner, while the failed alternatives remain hidden even though they are statistically relevant.
Indicator Agreement Is Often Not Independent Confirmation
Traders often combine RSI, MACD, moving averages, rate of change, and stochastic oscillators and feel reassured when all of them point in the same direction. But many of these indicators are different descriptions of the same underlying price momentum.
Agreement between related indicators can create the illusion of several independent confirmations when, in reality, they are all derived from the same historical series. That appearance of confirmation adds less information than it seems.
Smooth Indicators Can Make Randomness Look Structured
Raw prices are noisy. Indicators such as moving averages smooth that noise, making the resulting line look cleaner, more stable, and more meaningful. But smoothness is not evidence of predictability.
A 200-day moving average changes slowly because almost all of its data is reused from one day to the next. Its persistence is built into the calculation. Technical charts can therefore transform unstable price movements into visually convincing trends without improving the ability to forecast what happens next.
Precise Settings Are Often a Warning Sign
In the TSLA example below, the 55-day SMA finishes at +10.7%, the 54-day version ends at +13.5%, and the 56-day version drops to -4.4%. That kind of sharp change from a one-day parameter shift is difficult to reconcile with a broad economic mechanism.
A more credible strategy should show a broad region of reasonable performance across nearby settings. The goal is not to find one perfect parameter combination, but to determine whether a stable relationship exists across many reasonable variants.
The path matters as much as the endpoint. In the default TSLA 55-day SMA example above, a backtest stopped near one of the stronger peaks could have made the same setup look close to a 100% gain and worth selecting. A test run judged near one of the weaker stretches around a -15% drawdown could just as easily have pushed you to discard that exact setup and keep searching across other SMA settings or entirely different indicators.
What Stronger Evidence Would Look Like
A credible price-based strategy should not depend on one perfect historical configuration. It should show broad parameter stability, genuine out-of-sample testing, evidence across different market environments, and comparison with random or shuffled strategies to test whether the research process can manufacture apparent alpha from noise.
Most importantly, it should be linked to a persistent economic, behavioral, institutional, or market-structure mechanism. The indicator itself is not the explanation. Without a reason the effect should continue, the backtest alone is difficult to interpret.
The Central Problem
Technical-strategy research combines two unfavorable conditions: the predictive information in historical prices is usually weak, and the number of possible strategies is enormous. That means weak signal is hard to separate from ordinary market noise at the same time that massive search makes it easy to fit that noise exceptionally well.
The most important question is not whether a strategy performed well historically. It is how many indicators, settings, combinations, markets, periods, and rule changes were considered before this specific result was selected. Without knowing the size of that search, strong historical performance can be much less informative than it appears.
Build research frameworks that test ideas harder than a chart does.
Explore CatCapital ResearchCatCapital research outputs are for informational purposes only and are not financial advice. Investors should make their own decisions and consider their own objectives, constraints, and risk tolerance.