Market Weather Forecast: MTF RSI+SMA — The Ultimate Trend Detector?

A beginner-friendly summary of the verification: “Market Weather Forecast: MTF RSI+SMA — The Ultimate Trend Detector?”.

What’s the idea?

Today, we’re diving into a common question among EA developers: Can we combine popular indicators across different timeframes to build a truly robust and profitable trading strategy? Specifically, we explored a strategy using Multiple Time Frame (MTF) analysis with the Relative Strength Index (RSI) and Simple Moving Average (SMA). Here’s the core concept: Imagine you’re trying to figure out if it’s a good time to buy. Instead of just looking at one chart, you “zoom out” to a higher timeframe for the big picture, then “zoom in” to a lower timeframe for the perfect entry.

1. Higher Timeframe (HTF) SMA for the “Vibe”: We used a longer SMA on a higher timeframe (like a Daily chart) to understand the overall market “mood” or bias. Is the market generally heading up or down? This gives us our directional filter.
2. Lower Timeframe (LTF) SMA for Trend Confirmation: Once we have the general direction from the HTF, we use a shorter SMA on a lower timeframe (like an H4 or H1 chart) to confirm that a clear trend is actually in play right now. No point buying into a dip if the short-term trend is flat or against us!
3. Lower Timeframe (LTF) RSI for Dip Buying: With the trend confirmed, we then looked for “dips” within that trend using the RSI on the lower timeframe. The idea is to buy when the price has pulled back a bit but is still expected to continue in the direction of the main trend. This is often called “trend-following dip buying.” A quick technical note: When working with MTF, it’s easy to accidentally “cheat” by looking at future data. To prevent this “look-ahead bias,” we used a careful resampling method with a one-period shift. This ensures our tests are fair and realistic.

How I tested it

To find the “optimal” settings for this strategy, we put it through a rigorous testing process. This isn’t just about picking a few numbers; it’s about systematically exploring many combinations to see what works best. We tested this strategy across four different combinations of base and higher timeframes:

• H1 (1-hour chart) as base, D1 (Daily chart) as HTF
• H1 as base, H4 (4-hour chart) as HTF
• H4 as base, D1 as HTF
• H4 as base, W1 (Weekly chart) as HTF For each of these timeframe pairs, we ran a “grid optimization.” Think of this like adjusting many dials on a stereo system to find the perfect sound. We tried 36 different combinations of parameters (the specific settings for the SMA periods and RSI levels) to find the ones that yielded the best historical results. We used two main methods to test for robustness:

1. Fixed-Period (In-Sample) Optimization: This is like giving a student all the test questions and answers beforehand, then asking them to achieve the best possible score on those specific questions. It finds the absolute best settings for a specific historical period. While useful for initial exploration, it’s prone to “overfitting” – meaning the strategy might be perfectly tuned to past noise, not future market behavior.
2. Walk-Forward Optimization (WFO): This is the gold standard for testing EAs (Expert Advisors – automated trading programs). It’s a much tougher, more realistic test. Imagine training a student on three years of material, then testing them on one year of new, unseen material. Then, you move the window forward, train them on the next three years, and test on the next unseen year, repeating this process. This simulates how an EA would perform in a constantly evolving market, where you might periodically re-optimize your settings. We used a 3-year training window and a 1-year validation window.
3. Forward Test with Fixed Parameters: As an additional check, we took the single best parameter set identified by the initial fixed-period optimization and simply ran it “forward” on new data across different currencies, without re-optimizing. This tests if a “one-size-fits-all” best setting truly exists.

What happened?

Let’s get to the exciting (and sometimes humbling) part – the results!

The “Too Good to Be True” In-Sample Results

When we first ran the fixed-period optimization, some of the “optimal solutions” looked incredibly attractive!

• One combination (H4 base + D1 HTF) showed a +15.5% profit over the entire test period, with a Profit Factor (PF) of 1.11. The Profit Factor is a simple way to measure how much profit you make for every dollar you lose (Gross Profit / Gross Loss). A PF greater than 1 means you’re profitable. So, 1.11 means for every $1 lost, you made $1.11. Not bad!
• Another combination (H1 base + H4 HTF) was even more impressive, showing a whopping +33.8% profit with a PF of 1.15! These numbers certainly grab your attention. It seemed like we were on to something potentially very lucrative. But, as experienced traders know, the real test comes next…

The Reality Check: Walk-Forward Optimization

This is where the rubber meets the road. Our Walk-Forward Optimization, the true test of robustness, delivered a starkly different picture:

• The strategy resulted in an overall loss of -3.0% across all the walk-forward validation periods.
• Even more telling, only 1 out of 5 of the individual 1-year validation windows actually generated a profit. Why such a massive difference? We observed that the “optimal” parameters selected by the optimizer were wildly unstable from one training window to the next. For example, the ideal SMA period would jump between 100 and 200, and the RSI buy level would swing from 30 to 50. This constant shifting is a classic sign of overfitting. It means the strategy was too finely tuned to the specific historical data of each training window, like a student who memorized answers rather than understanding the concepts. When faced with new, unseen data, it simply fell apart.

Fixed Parameters on New Data: Another Letdown

Even when we tried taking the single “best” parameter set from our initial fixed-period optimization and applying it to new data across different currencies (a “forward test”), the results were similarly disappointing:

• An overall loss of -11.0%.
• Only 2 out of 6 currency pairs showed any profit. This confirms the overfitting issue: what looks “optimal” for one specific historical period often crumbles when exposed to the unpredictable future.

What I learned

The main conclusion from this research is clear: the “optimal solutions” we found for this MTF RSI+SMA combination were a classic case of overfitting. Even though they showed impressive profits of +15% to +34% in fixed-period tests, these gains completely vanished when subjected to proper walk-forward validation and forward testing. This is a pattern we see repeatedly with many price-based strategies. They tend to latch onto historical noise rather than true, repeatable market behavior. However, there was one interesting observation:

• Low Drawdown (DD): The strategy exhibited relatively small drawdowns during the walk-forward tests, ranging from -0% to -7% annually. Drawdown (DD) refers to the maximum peak-to-trough decline in an account balance. Small drawdowns suggest that even though the strategy wasn’t profitable, it also wasn’t excessively risky. This means the strategy falls into the category of “no edge, but low risk.” It won’t be a standalone profit generator. But, in theory, a low-drawdown strategy could potentially serve as a component for diversification within a larger, more complex portfolio, especially if the goal is to reduce overall portfolio risk rather than generating direct profit from this specific method. Ultimately, this experiment reminds us that while indicators like RSI and SMA are valuable tools, simply combining them across multiple timeframes and optimizing for past performance isn’t a shortcut to consistent profits. True trading “edge” – that consistent, repeatable advantage – still stems from understanding the underlying market dynamics and trends.