Can We Build a Macro-Based Currency Strength Meter With AI?

I was looking to build my own macro fundamental based currency strength meter in order to identify possible shifts in sentiment towards a market, that wasn't just based off technicals.

Don't get me wrong, I love my current currency strength meter, but the one I use daily is based off technical movements within the market. I then use my own knowledge and experience of the market to identify the "why" the market could move, and use the currency strength meter to add confluence.

The tricky part of building the new macro meter is the fundamental drivers we use in order to identify opportunities. In my head, there are a few that are non-negotiable, they have to be in there. For example indicators such as interest rate differentials, GDP differentials and inflation measures. However, each economy can have subtle differences. For instance, New Zealand's economy relies on dairy product exports as it makes up a large portion of trade. This is a metric that would only impact the NZD.

So as you can see it's not as easy as just writing down the same metrics. At least that's my thoughts right now, if I can prove myself wrong then great, I am always open-minded when it comes to trading and building any kind of trading tool.

As I was finding it a little difficult I thought I would turn to AI tools to help me. I have always been curious to know how traders are incorporating AI into trading. Whether it be trading bots or algorithms that learn to adapt to the market or offer trading signals, and I do have an interview with a trader coming very soon, that uses AI tools to trade the markets, so be on the lookout for that one.

But I wondered if a macro fundamental strength meter could be built using AI.

Now traders I am not very skilled when it comes to coding, it is something I have done in the past but I am a complete amateur. It is something I am interested in getting into, but I am not ready for that challenge just yet.

So to begin I asked chatGPT a pretty basic question:

"If you were to become an advanced forex trader that used macro data to determine opportunities in the market what would you use and why?"

Here's what it spat out:

"If I were to become an advanced forex trader using macro data to drive decisions, I'd focus on building a global macro framework—essentially a lens for identifying currency mispricings and macro divergences between countries. Here's what I’d use and why, broken down into key categories:"

So far so good, that's exactly what we are trying to build the tool for. It's to find macro data divergences between countries to exploit trading opportunities.

It then goes on to break down the markets into categories.

1. Interest Rate Differentials and Central Bank Policy: The AI wants us to use the central bank interest rates from around the world and compare them. It also wants us to look into interest rate expectations, and the reason why is that in its words, "interest rates are the single most influential driver of current movements". Traders are constantly trying to price in where rates are going. To do this we need to look towards "forward rate curves/OIS spreads, Fed Funds Futures, STIRs and so on".

2. Inflation Metrics: The AI pointed to CPI, Core CPI, PCE and PMI's as important indicators to watch as well as inflation expectations from surveys and wage inflation data. The reason behind this comes back to interest rate expectations. For example, higher inflation can see a hawkish central bank, which often means higher yields and a stronger currency.

3. Growth Data and Leading Indicators: For this the AI prompt suggests we follow GDP growth, PMI/ISM, retail sales, industrial production, durable goods, consumer confidence and labour market data such as NFP. Why? Well, capital flows to countries with better returns on investment which is driven by strong productivity and growth.

4. Current Account / Trade Balance / Capital Flows: Sustained current account deficits can lead to FX depreciation over time, especially if not funded by stable capital inflows.

5. Sentiment and Positioning: This was interesting to me because it pointed to using the CFTC or CoT reports to understand sentiment which is something I incorporate into my trading analysis anyway. It wants to use this information the understand where the crowd is positioned and can help with contrarian trades or timing exits.

After some information I asked the model to give me a recent example that uses all the points provided to identify a trading idea. It then went on to point out a USD/JPY trading idea because of the following reasons:

Macro FX Trade Setup: Long USD/JPY

🌍 Macro Context

🏦 Central Bank Divergence: Fed vs. BoJ

• Fed is still relatively hawkish: U.S. inflation is sticky at 3.5%, unemployment is low (~4%), and wage growth is steady. Fed officials are guiding for higher-for-longer rates.

• BoJ has just barely exited YCC (Yield Curve Control) but remains ultra-dovish. Real rates in Japan are still deeply negative.

✅ Interest rate differential strongly favors the USD.

📈 Inflation & Growth Divergence

• U.S. growth is outperforming: Q3 GDP beat expectations at 2.3% annualized.

• Japan is seeing slowing export demand (especially from China), and inflation is cooling below 2%, giving BoJ more room to stay easy.

✅ Growth and inflation divergence adds fuel to the trade.

💸 Capital Flows / Current Account

• Japan still has a current account surplus but faces capital outflows due to unattractive bond yields.

• U.S. is attracting foreign investment due to high yields and stronger equity markets.

✅ Capital flows support USD strength.

📊 Risk Sentiment

• Risk sentiment is neutral to mildly risk-on. Not a huge tailwind or headwind.

• USD/JPY tends to rise in risk-on conditions, while JPY strengthens in panic.

➖ Not a major driver, but not hurting the trade either.

💡 Valuation

• USD/JPY is not extremely overvalued yet based on REER or PPP.

• Some short-term technical consolidation gives room for breakout continuation.

✅ No strong valuation headwinds.

📉 Positioning

• CoT data shows some long USD/JPY positioning, but not extreme.

• Implied vol is low, which is good for tight stop-loss setups.

✅ No positioning red flags; option market is calm.

It was said to be long around November 10th 2024, this would have led to a good profit.

Ok, so I think this is a good starting point for the macro indicator, as the points the AI tool has made are actually points that can be consistent across most assets. This makes the tool more efficient to use.

Looks like I will be going into the lab for a while to try and get something together.

This is what I love about markets, it's building and thinking of ways to try and get something out of them.

Fun stuff! Speak soon!