Indicator Deep-Dive: Not All Signals Are Created Equal Across Asset Classes

The Question Behind the Numbers

Most retail traders treat technical indicators as universal tools — plug in RSI, MACD, or a moving-average crossover, and expect it to behave the same way whether you're trading EUR/USD, crude oil, or a mid-cap industrial stock. That assumption is expensive. The more interesting question — and the one this report attempts to address — is whether the same signal logic produces materially different outcomes depending on the asset class it's applied to. Our proprietary backtest and live signal system, which has now closed 19 signals across forex, equities, and commodities, offers an early but instructive answer: the asset class a signal fires in matters as much as the signal itself, possibly more.

An overall win rate of 42.1% and an average trade return of +0.12% across those 19 signals sounds modest, perhaps even discouraging at first read. But headline aggregates obscure the real story. Strip out one underperforming category and the picture changes substantially. This report breaks that apart, quantifies where our indicators are earning their keep, and — critically — where they are not.

Where the System Is and Isn't Working: A Category-by-Category Breakdown

The table below presents the full disaggregation of our 19 closed signals by asset class.

Forex is the largest single category in our dataset, accounting for 10 of the 19 closed signals. It posts a 50.0% win rate and an average return of +0.20% per trade. Those numbers are unspectacular but stable. For a signal-based system operating in a market that runs 24 hours, is dominated by institutional flow, and frequently mean-reverts around data events, breaking even on direction half the time while extracting a small positive edge on the average trade is a reasonable baseline. Forex also benefits from tight spreads and deep liquidity, which means the gross return figures are closer to net returns than they would be in less liquid markets.

Stocks, by contrast, tell a more compelling story — though with the important caveat that only four signals have closed in this category. Those four trades carry a 50.0% win rate and an average return of +2.33%. That average return figure is more than eleven times the forex average and dwarfs the overall system average of +0.12%. The intuition is plausible: equities tend to trend with more persistence than forex pairs once a catalyst breaks price from consolidation. Momentum and breakout indicators, in particular, tend to have wider payoff distributions in stocks, where a single earnings-driven move can extend a winner significantly. But with an n of four, we cannot draw confident conclusions. We flag this for what it is: a promising early read that requires substantially more data before we assign structural weight to it.

Commodities is where the system is visibly struggling. Five closed signals, a 20.0% win rate, and an average return of −1.79% per trade. That is not a rounding error or a single bad trade distorting a small sample — a 20% win rate means one winning signal out of every five. At that ratio, the average loser would need to be dramatically smaller than the average winner to keep aggregate returns positive, and the −1.79% average return indicates that is not happening here. Commodities markets are structurally different from forex and equities in ways that are likely punishing indicator-based systems: they are driven heavily by supply-side shocks, geopolitical events, and seasonal roll dynamics that do not encode cleanly into price-action indicators. A momentum signal that works beautifully on a currency pair or a growth stock can get caught badly in a commodity that is whipsawing around a supply report.

What the Overall 42.1% Win Rate Is Actually Telling Us

The system's headline win rate of 42.1% is below 50%, which means on a purely directional basis we are currently calling more trades wrong than right. That is worth sitting with rather than explaining away. However, the average return of +0.12% being positive while the win rate is below 50% is itself informative: it implies the system's winners are, in aggregate, larger than its losers. A system can be profitable at a sub-50% win rate if its average winner is sufficiently larger than its average loser — that is the fundamental math behind trend-following and breakout strategies.

The stocks category provides the clearest evidence of this dynamic. With a 50.0% win rate and a +2.33% average return, the winners in that category are pulling the average up meaningfully. The commodities category, however, appears to violate the necessary condition: a 20.0% win rate requires very large average winners to overcome a high loss frequency, and the −1.79% average return indicates the losses are not being offset.

What the aggregate numbers do not tell us is the distribution of returns within each category — whether those averages are being skewed by one or two outliers, or whether they reflect a consistent pattern across all signals in the group. With totals of 10, 4, and 5 signals respectively, outlier sensitivity is real and should not be minimized. We return to this in the methodology note.

Where We Could Be Wrong — and What We're Not Ready to Claim

Several interpretations of this data are plausible, and intellectual honesty requires naming them.

First, the commodities underperformance could be regime-specific rather than structural. If those five signals fired during a period of unusually high volatility — driven by, say, energy supply disruptions or agricultural weather events — the results might look very different under normal conditions. We do not have sufficient signal history to separate regime effects from indicator effects in commodities. This is preliminary data, and we are not ready to conclude that our indicators simply do not work in commodity markets. What we can say is that they have not worked so far across five closed signals.

Second, the strong average return in stocks (+2.33%) may be driven by one or two large winners in a four-signal sample. If a single trade returned +8% and the other three were flat to slightly negative, the average would look impressive while telling us almost nothing about expected future performance. We do not have within-category distribution data to rule this out, which means the stocks result, as encouraging as it looks, should be treated as a hypothesis to test rather than a conclusion to act on.

Third, and most importantly: 19 total signals is a small number. In academic finance, conventional thresholds for statistical significance in strategy testing typically require dozens to hundreds of observations, depending on the variance of returns. Our full dataset does not meet that threshold for any single category. We are not claiming statistical significance here. We are presenting early directional evidence and treating it accordingly.

The most dangerous reading of this report would be to aggressively increase position sizing in stocks signals and abandon commodity signals entirely based on these 19 data points. That would be acting with more conviction than the data supports.

What Traders Should Actually Do With This

Despite the caveats, there are concrete, low-regret adjustments that follow logically from what the data shows — adjustments calibrated to the uncertainty of the sample rather than its surface-level conclusions.

• Apply asymmetric position sizing by category, not uniform sizing across all signals. Given that commodities signals have produced a −1.79% average return across five trades and a 20.0% win rate, sizing them equally to forex or equities signals is not justified. Until the commodities sample expands and the win rate improves, reducing position size in that category relative to the others is a prudent risk management response — not an abandonment of the signals, but a reflection of current uncertainty.
• Treat the forex category as the system's anchor. Ten closed signals at 50.0% win rate and +0.20% average return is not glamorous, but it is the most reliable read we have. The forex results benefit from the largest sample size in our dataset and represent a positive-expectancy baseline. Traders building exposure around our signals should treat forex as the core allocation until the other categories accumulate more history.
• Watch the next five to ten stocks signals closely. The +2.33% average return on four closed stock signals is the most interesting number in the dataset and the most fragile. The next handful of stock signals will either reinforce or revise that figure substantially. Pay attention to whether winners in that category are genuinely larger than losers, or whether the average is being sustained by a single outlier.
• Do not dismiss commodity signals entirely. A 20.0% win rate over five trades could reflect a short-term rough patch as readily as a structural indicator mismatch. Continue tracking these signals, but do so at reduced size and with tighter risk parameters. If the win rate remains below 35% after 15 or more commodity signals, that would be a more meaningful signal to reconsider the indicator logic applied in that category.
• Look at the conditions under which signals fired. The data presented here does not break down by market regime — trending vs. range-bound, high-volatility vs. low-volatility. As our signal database grows, that cross-tabulation will become the next important analysis. For now, traders who are keeping individual trade logs should note the VIX level, trend strength, and volatility context for each signal they act on. That data will matter later.

Methodology: What This Data Can and Cannot Support

The numbers in this report are drawn from Stocks365's proprietary backtest and live signal system as of April 17, 2026. The dataset covers 19 closed signals across three asset classes: forex (n=10), stocks (n=4), and commodities (n=5). Return figures represent average return per closed trade within each category. We are not disclosing the specific indicators, timeframes, or instruments underlying each signal in this publication, as those details are reserved for subscribers to our signal service.

The sample sizes here are small by any rigorous quantitative standard. Nineteen total signals — and category samples as small as four — are not sufficient to support claims of statistical significance, and we make no such claims. The win rates and average returns presented should be understood as directional indicators of early system behavior, not stable long-run estimates. Averages over small samples are sensitive to individual outliers, and the true expected return of each category could differ materially from the figures shown here once more signals accumulate.

We publish these numbers because transparency is more useful than silence, and because early data — clearly labeled as such — gives readers the ability to form their own calibrated views. As signal count grows, we will update this analysis and flag where initial readings held or where they did not. That is the exercise: not to be right on nineteen trades, but to build a system whose accuracy can be honestly measured over hundreds.