Is v1.4.0 REALLY Optimal? We Tested 2 New Ideas to Find Out!

A beginner-friendly summary of the verification: “Is v1.4.0 REALLY Optimal? We Tested 2 New Ideas to Find Out!”.

What’s the idea?

We’re always on the hunt for ways to make our algorithmic trading systems (EAs) even better. This time, we put our Core System v1.4.0 under the microscope to test two specific ideas:

1. Stock Market Filters: Could using different stock market indices as a filter improve performance for our JPY (Japanese Yen) cross pairs?
2. Pyramiding: Is adding to winning positions (a technique called pyramiding) actually beneficial for this system, especially with v1.4.0’s new “overlay” features? Our goal, as always, is to find real, verifiable edges, not just introduce complexity for complexity’s sake.

How I tested it

Let’s break down how we tackled each question.

Filtering with Stock Indices

Our Core System v1.4.0 primarily trades JPY cross pairs (like EUR/JPY or GBP/JPY). The Japanese Yen is often considered a “safe-haven” currency, meaning it tends to strengthen during times of global uncertainty or “risk-off” sentiment, and weaken when confidence returns. This connection between global risk and JPY crosses made us wonder if a stock market index could act as a good leading indicator or filter. We tested four scenarios for our JPY cross trading:

• US500 Filter: Using the S&P 500 index (a major US stock market index) as a filter. The hypothesis here is that the US market often leads global risk sentiment.
• JP225 Filter: Using the Nikkei 225 index (Japan’s main stock market index) as a filter. Perhaps a local index would be more relevant for JPY crosses.
• Both US500 AND JP225 Filter: This would only allow trades if both conditions from the US and Japanese markets were met.
• No Filter: This served as our baseline, representing the current v1.4.0 system without any stock market filtering. We measured performance in terms of monthly return, while keeping the maximum drawdown (DD) at a 10% equivalent. Drawdown (DD) is the peak-to-trough decline in your capital, and keeping it controlled is crucial for long-term survival.

Re-evaluating Pyramiding

Pyramiding is a strategy where you add more units to an already profitable position, essentially “pyramiding” into a winner. The idea is to amplify gains when a strong trend develops. We’ve explored this before in “Research 53,” which suggested it wasn’t ideal for our proprietary trading (prop trading) approach, where efficiency and consistent returns are key. However, v1.4.0 includes some new “overlay” features that might manage positions more intelligently, so it was worth a fresh look. Here’s how we tested it:

• We compared the standard v1.4.0 system (no pyramiding) against versions that allowed for a maximum of 2 or 3 additional units to be added to a winning trade.
• Crucially, to make a fair comparison, we kept the total maximum risk for a full position roughly the same across all scenarios. This meant that when we allowed pyramiding, the risk per unit had to be smaller. Think of it like having a fixed budget for a trade; if you buy more units, each unit costs less. We again looked at monthly returns, normalized to a 10% drawdown equivalent.

What happened?

Time for the results! And frankly, they were quite illuminating.

Stock Market Filters: US500 Wins!

Here’s how the different stock market filters performed (monthly return at 10% DD equivalent):

• US500 Filter: A respectable +0.22% per month.
• JP225 Filter: +0.18% per month. Interestingly, this was pretty much the same as having no filter at all.
• Both US500 AND JP225 Filter: +0.20% per month. What this means: The US500 filter clearly outperformed the others. This strongly suggests that the global stock market (led by the US) is a better “crystal ball” for predicting overall risk sentiment than the Japanese market alone. The US market’s movements often precede shifts in global investor confidence, which then impacts demand for safe-haven currencies like the JPY. So, the existing v1.4.0 system, which already incorporates the US500 as a filter, is indeed using the optimal approach here. Great confirmation!

Pyramiding: Not a Winner

The results for pyramiding were clear and, perhaps, a bit disappointing for those hoping for an easy boost:

• Standard (no pyramiding): +0.23% per month (at 10% DD equivalent).
• Max 2 additional units (pyramiding): +0.21% per month.
• Max 3 additional units (pyramiding): +0.19% per month. What this means: Performance actually declined as we allowed more pyramiding! In other words, adding to winning positions reduced our overall monthly returns. This was a monotonic deterioration – the more we pyramided, the worse the results. Why this happened: Remember how we kept the total maximum risk constant? When you add more units, but keep the total risk the same, each individual unit carries less risk. This results in smaller average position sizes for each trade. While this might also reduce drawdown (which it did, implicitly, as we normalized to 10% DD), it ultimately leads to what we call “under-capitalization” for the potential of each trade. You’re effectively spreading your bets too thin across more units, rather than concentrating your capital on the most promising initial entry. This re-confirms our findings from “Research 53”: pyramiding is not an effective strategy for improving performance with this system in a prop trading context.

What I learned

This research, while not yielding a new breakthrough, was incredibly valuable. It solidified our understanding of the v1.4.0 system:

• The US500 filter is superior for JPY crosses because the US market is a more effective leading indicator for global risk sentiment.
• Pyramiding does not improve returns for Core System v1.4.0; in fact, it actively harms performance by leading to under-capitalization.
• Core System v1.4.0 is already highly optimized for the price data it uses. We call this being “on the frontier of price data” – meaning, with the current inputs and logic, it’s performing as well as it can. The big takeaway? No changes are needed for our current v1.4.0 system! It’s already running efficiently. For future improvements, we’ll need to look beyond current price data optimization. This might involve exploring entirely new data sources (like interest rate differentials or “carry trade” strategies, though these often come with specific prop trading constraints), or even venturing into different asset classes. Of course, continued real-world “forward testing” remains crucial to validate everything we do. Sometimes, the best discovery is realizing that what you’ve already built is robust and optimal. That’s a win in itself!

How this connects

This verification builds on earlier ones (what failed before and what I tried this time, comparisons between approaches).

• Pyramiding Strategy: Did Chasing Wins Bring Max Profit or Risk?