In previous strategy: 🚀 Strategy #13: QQQ Trend Following Strategy, it shows how to use the AI Agent trend following skill into a trading strategy that out-perform S&P500 in risk-based return and max drawdown.

It is the AI Agent Skill i am using to build this trading system

Although the Strategy#13 didn’t out-perform S&P500 absolute return, it doesn’t mean the trend following indicator can’t build strategy with better result.

In this strategy, i show you how to use the same trend following indicator to build a new trading system with 3 different assets that outperform both absolute return and risk based return.

One simple indicator. 3 different assets.

No predict for tomorrow price. Just follow the current trend.

• Return Potential: ★★★☆☆

• Ease Of Adoption: ★★★★☆

• Drawdown Protection: ★★★★☆

🆘 Common Problems

Many people think that buying NVDA, SMCI, MRVL, etc when they are uptrend are not putting all eggs into a basket.

However, those symbols are highly correlated. Even worse, the uptrend strategy make it worser.

In order to build a diversify portfolio, you should check their correlation figure before including them in the portfolio.

Otherwise, the portfolio is fully exposed to one asset’s drawdown.

This strategy solves that by treating each asset’s trend as its own independent decision.

3 different assets are each measured against their own trend following indicator, and capital only flows into whichever combination is actually confirmed as trending.

🚨 If One Symbol Is Not Good Enough, What’s About Three Symbols?

A good indicator should be functional not only one symbol, but different symbols.

When one symbol in our previous strategy post proved that it can improve risk-based return, we should extend this.

In this strategy, we simply apply the same indicators to 3 assets that yield a result outperform S&P500 in both risk-based return and absolute return.

Results (2020–2026):

• Annual Returns: +15.9% ✅

• Sharpe Ratio: 1.08 ✅

• Profit Factor: 1.75 ✅

• Max Drawdown: -20.3% ✅ (🎉 40% better than buy and hold S&P500)

Paid content to unlock the following:

• 3 assets used in this trading system

• Full backtest setup details (instrument, fees, execution logic)

• Source code (Python + VectorBT) ready to run in Google Colab

BONUS For Paid Member ONLY - “Champion’s Fast Lane Playbook: Master Algo Trading in 12 Weeks” - My 8 modules self-study guides and notes for building a winning strategy.

(Paid Content Below)

⬇️⬇️⬇️

Read more

In 2025, I developed several trading systems and picked 5 to carry on in 2026. The details of the systems are as below:

The 5th Trading Systems terminated in paper trading phase, now i have 4 systems lived.

📈 Market Recap - PANIC MODE IS ON

The market was now stable in August and a good rebound for all my algos. As it is not a financial news or info blog, i am not going to summarize what’s happened in Aug.

🚨PANIC MODE IS ON🚨

Although the market and my algos are rebounding, panic mode is on due to my brokerage needs me to apply passkey due to regulatory requirement.

The impact is that Quantconnect, i.e. the Algo trading SaaS, that i am using is not able to connect to my brokerage account and my algos are all OFF due to passkey is not yet support. (as-per today update, they have a solution for FIX integration with Quantconnect, currently in beta. Hopefully, this can official launch that can resolve the issue soon.)

Here is my plan:

1. Short term plan

   1. Stock trading algo, i.e. Quant Trading System#3 - Hit & Run Stock Momentum Strategy, will switch off. Related fund will park in index fund ETFs

   2. Other strategies will switch to manual order submit based on the signal generated by those algo trading system.

   3. Raising ticket to IBKR to see if not to use passkey exemption is a solution

2. Mid-term plan

   1. I figured out a solution on running algo in my own Google Cloud Platform to connect to my IBKR directly. Currently running tests in limited capital live account.

   2. I am going to migrate current algos that are using daily or hour timeframe with limited number of symbols traded

   3. For other algos like a huge stock universe with minute timeframe, i will want for Quantconnect solution

3. Long-term plan

   1. Explore other brokerages to see if there is any better solution for integration

   2. Build my own algo trading platform

📣 Quick Update On My Four Systems

Quant Trading System#1 - Risk Parity Broad Market ETFs Strategy

🟢 This is the strategy that i will move to Google Cloud Platform. Currently in beta testing.

Quant Trading System#2 - Volatility Based & Alpha Factors ETFs Strategy

🔴 This strategy require a clean uptrend and not ideal for this kind of choppy market. It will be paused and i am looking for a brand new algo to replace this in Google Cloud Platform.

Quant Trading System#3 - Hit & Run Stock Momentum Strategy

🔴 This strategy will be paused and I am waiting for QuantConnect solution to redeploy

Quant Trading System#4 - Sector Rotation Driven ETF Strategy

🟡 I will carry on trading this strategy manually. Afterall, it is a monthly rebalance strategy, i just need to get its signal today and manual buy-in them.

But i think i will take this chance to enhance this strategy as monthly pick and rebalance is quite lagging i.e. You can see the huge drawdown in Jun as at that time, the semi-conductor related ETFs were moving nice in May, so, the strategy picked a couple of related ETFs and sledgehammered drop in June.

⚠️ Disclaimer: The content provided is intended solely for educational purposes and reflects the author’s personal experiences with data insights, backtesting and algorithmic trading. It does not constitute financial advice, investment recommendations, or financial services. All references to securities or financial instruments are for illustrative purposes only and should not be construed as endorsements or recommendations for trading. Performance figures presented are based on hypothetical backtested results, which are not indicative of future performance. Trading systems may contain errors, and no warranties are made regarding their accuracy or functionality. Past performance does not guarantee future results. Individuals are strongly advised to conduct thorough due diligence, including but not limit to forward testing in paper or simulated accounts for an appropriate duration, before deploying any trading strategy with real capital. Trading carries significant risks, including high volatility and the potential for complete loss of invested capital. Live trade the strategy at your own risk, as any losses incurred are not the responsibility of the author.

The foundation of a successful trading strategy lies in a clear and logical understanding of its underlying principles. Blindly adopting strategies without grasping their mechanics can lead to poor decisions, unexpected losses, and missed opportunities.

The Role of Logical Clarity in Trading

A trading strategy’s logic defines the cause-and-effect relationship that drives its performance. For example, your hypothetical scenario of temperature changes influencing market rises sounds intriguing but lacks a credible causal link. Without a solid rationale, such a strategy isn’t worth pursuing, as it’s unlikely to hold up under scrutiny or in live markets. Understanding the logic helps you filter out flawed ideas early, saving time and resources.

In contrast, a strategy like mean-reversion, which assumes prices will return to their historical average after deviating, has a clear and testable logic. Adding a percentage-based stop-loss worsened performance because it contradicted the strategy’s core assumption. The further prices deviate from the mean, the higher the likelihood of reversion, so cutting losses prematurely undermines the strategy’s edge. This insight highlights how logic informs not just strategy design but also risk management.

Navigating Drawdowns with Logical Insight

When a strategy faces a significant drawdown—such as a 30% loss exceeding backtest results—logical clarity becomes a lifeline. Without understanding the strategy’s mechanics, you’re left guessing whether to continue or abandon it. A deep grasp of the logic allows you to evaluate whether the strategy’s assumptions still hold.

For instance, if a trend-following strategy relies on sustained price momentum but the market shifts to a range-bound state, the logic no longer applies, justifying a pause or switch. Conversely, if the assumptions remain valid, a drawdown might simply reflect normal variability, encouraging you to stay the course. This ability to make informed judgments during tough times stems directly from understanding the strategy’s foundation.

Adapting to Changing Markets

Markets are dynamic, shifting between trending, ranging, or volatile states. A strategy built on clear logic can be adjusted to these changes. For example, a momentum strategy might thrive in trending markets but struggle in choppy conditions. By understanding the logic—price momentum signals continuation—you can develop rules to detect market regime shifts (e.g., using volatility indicators) and adjust or pause the strategy accordingly. This adaptability ensures your approach remains relevant, reducing the risk of applying a strategy in unsuitable conditions.

Building Confidence and Discipline

Knowing why a strategy works fosters confidence, which is crucial for disciplined execution. Without this understanding, you might hesitate during drawdowns or exit trades prematurely, undermining performance.

For instance, a mean-reversion trader who understands that larger deviations increase reversion odds is less likely to panic during losses. This confidence promotes consistency, helping you stick to the plan even when emotions run high.

Setting Realistic Expectations

Logical insight helps align expectations with a strategy’s design. For example, a high-frequency scalping strategy might target small, frequent gains with occasional losses, while a swing-trading strategy aims for larger but less frequent profits. Understanding the logic clarifies these trade-offs, preventing unrealistic goals like expecting consistent daily profits from a long-term trend-following system. Realistic expectations reduce frustration and keep you grounded in the strategy’s realities.

Avoiding Black Box Pitfalls

Copy-pasting strategies or relying on “black box” systems—where the mechanics are opaque—can lead to disaster. Without understanding the logic, you’re vulnerable to hidden flaws, such as overfitting to historical data or biases that fail in live markets. By dissecting the strategy’s principles, you can test its robustness and avoid blindly trusting unproven systems. This vigilance protects your capital from strategies that look promising on paper but crumble in practice.

Conclusion

Understanding the underlying logic of a trading strategy is the cornerstone of success. It empowers you to make informed decisions during drawdowns, adapt to market shifts, execute with confidence, set realistic goals, avoid black box risks, apply strategies in suitable conditions, and refine them effectively. By prioritizing logical clarity, you build strategies that are not only robust but also adaptable, giving you an edge in the ever-changing world of trading.

(Disclaimer: The contents are for educational purposes only, sharing personal experiences and insights in developing and backtesting quantitative trading strategies using principles of quantitative finance. Developing or using quant trading strategies is a method or tool, not financial advice. The information provided, including coding techniques, backtesting methods, or statistical models, is general knowledge and not specific recommendations to buy or sell securities. It is not related to investment opportunities, passive income, debt relief, financial services, or financial consulting. I do not provide financial advice or services)

🔐 What Is a Passkey?

Passkey is a digital token that stores inside your device like phone, computer or physical digital key. It’s built on technology standard called FIDO2. You can’t type or copy this key.

Instead of remembering a password, you just use something already on your phone: your face (Face ID), your fingerprint, or your phone's PIN. When you log in, IBKR sends a little digital puzzle to your device, and only your device can solve that puzzle. Nothing that could be stolen or guessed ever travels across the web.

Indeed, it is a far more secure way, compare to current IB key, to protect you from hackers.

🔑 IB Key System

You may wonder that the existing IB key system is using your phone’s biometrics too, what’s the different?

The key different is that, those IB keys can be intercepted or faked by scamming lookalike fake website. This’s called “phishing” and it is exactly what passkey can addressed.

Using passkey solves the “phishing” attack entirely as the digital key is locked to IBKR’s website address. i.e. even if a fake site looks identical, your device won’t hand over the key as the website address is not IBKR official website address.

🆘 It is a more secure measure, what’s the problem then?

There isn’t any problem as normal user who is doing trading normally using IBKR apps. But as a algo trader, it is a huge issue.

I am hosting my algo in cloud infrastructure, and none of them are suitable for my case.

1. In order to login the passkey using my phone, i need to put my device nearby with Bluetooth ON

Issue: my algo is sitting in cloud that Bluetooth can’t reach it

2. Alternatively, i can insert the security key to machine’s USB port.

Issue: I don’t know how to insert the security key physically in the cloud

3. Using Software key like Proton Pass, 1Password, etc

Issue: It is not an option. Those software key can’t intercept IBKR gateway (IBKR gateway is a software to authenticate for IBKR API for algo trading) for authentication as they are browser based solution.

🤔 What’s Option Do It Have?

There are several options come up from my mind.

1. Change to other brokerage that doesn’t require passkey

   🚫 It is not a perm solution, as using Passkey is regulatory requirement. Eventually, every brokerage in Hong Kong need to use Passkey, i think

2. Hosting the entire algo in my home such that i can reach out to the physical machine to authenticate.

   🚫 It is the least thing i want to do, as there are scalability and stability concept

3. Hosting the IBKR gateway in my home and the real algo logic in the cloud.

   🚫 It seems better than above but hosting at my home is the least thing i want to do even it is IBKR gateway only.

🚨 At Last, This Tiny Little Thing Saved My Algo Trading Journey

This is security key that support Passkey.

With this, i can host the IBKR gateway in cloud GCP VM as-of today. But I don’t need to insert this into the cloud as i can’t reach it.

Paid content to unlock the following:

• The exact way to authenticate IBKR gateway using passkey in cloud

• The sample step-by-step commands

  

BONUS For Paid Member ONLY - “Champion’s Fast Lane Playbook: Master Algo Trading in 12 Weeks” - My 8 modules self-study guides and notes for building a winning strategy.

(Paid Content Below)

⬇️⬇️⬇️

Read more

Algorithmic trading demands precision and constant optimization to stay profitable in dynamic markets. Monitoring key performance indicators (KPIs) provides critical insights into a strategy’s profitability, risk, and reliability, enabling traders to fine-tune their systems. Below, we explore six essential KPIs—Sharpe ratio, maximum drawdown, profit ratio, Calmar ratio, maximum loss—and explain their importance in algo trading.

1. Sharpe Ratio

The Sharpe ratio measures risk-adjusted returns by dividing the strategy’s average return (minus the risk-free rate) by the standard deviation of returns.

A higher ratio indicates better returns for the risk taken. For algo traders, this metric is vital for assessing whether profits justify volatility. A low Sharpe ratio may prompt adjustments to reduce risk exposure or enhance returns, ensuring the strategy remains efficient.

You should target a Sharpe Ratio at least 1.5 in at least 5 years timeframe.

2. Maximum Drawdown

Maximum drawdown is the largest peak-to-trough decline in portfolio value, expressed as a percentage.

It quantifies downside risk, showing the worst-case capital loss during a losing streak. Algo traders use this metric to evaluate strategy resilience, adjusting leverage or stop-loss levels if drawdowns are too severe, protecting capital in adverse market conditions.

Ensure maximum drawdown stays within 15-20%. Set up alerts to trigger if live trading reaches the backtested maximum drawdown level. If the drawdown exceeds this by an additional 10-20%, consider halting the algo trading, as this signals potential issues with the strategy requiring immediate review.

3. Profit Ratio

The profit ratio, or profit factor, is the ratio of gross profits to gross losses.

A ratio above 1 indicates profitability, with higher values showing stronger gains relative to losses. Algo traders rely on this metric to gauge trade efficiency. Target profit ratio at least greater than 1.5 is a good starting point.

4. Calmar Ratio

The Calmar ratio divides annualized returns by maximum drawdown, highlighting returns relative to downside risk.

A higher ratio reflects a strategy with strong returns and manageable losses. For algo traders, this metric guides risk management, especially in high-frequency or leveraged strategies, encouraging adjustments to minimize drawdowns while preserving profitability. Calmar Ratio 1.5 or above is advisable.

Calmar Ratio vs Sharpe Ratio

Calmar Ratio primary focus on tail risk or worst case as max drawdown is the denominator while Sharpe Ratio focus on total risk / volatility via standard deviation.

5. Maximum Loss

Maximum loss is the largest single-trade loss, expressed as a monetary value or percentage of capital. It reveals the worst-case risk per trade. Algo traders use this to set stop-loss limits or position sizes, preventing catastrophic losses and ensuring the strategy aligns with risk tolerance.

(Disclaimer: The contents are for educational purposes only, sharing personal experiences and insights in developing and backtesting quantitative trading strategies using principles of quantitative finance. Developing or using quant trading strategies is a method or tool, not financial advice. The information provided, including coding techniques, backtesting methods, or statistical models, is general knowledge and not specific recommendations to buy or sell securities. It is not related to investment opportunities, passive income, debt relief, financial services, or financial consulting. I do not provide financial advice or services)

I published the previous 🚀 Strategy #14: 3 Assets Trend Following Strategy a few days ago.

That made me ask a simple question: is a plain buy-and-hold strategy using the same 3 assets actually better than the trend-following indicator?

After all, if a simple buy-and-hold approach delivers similar — or even better — performance metrics, I would rather drop the trend-following system altogether.

This strategy setup is very simple, but the results are good.

• Return Potential: ★★★☆☆

• Ease Of Adoption: ★★★★★

• Drawdown Protection: ★★☆☆☆

🆘 Common Problems

Most systematic traders drill down into sophisticated indicators to build a trading system.

They start a simple indicator, and adding more and more indicators.

They think that it can help to build a “holy grail” of trading system.

They may be able to build a good backtest metrics, but it won’t work in live.

That is curve fitting.

From time to time, I suggest to my readers to keep the trading system as simple as possible.

If giving up a parameter and yield a similar result, give it up.

If giving up an indicator and yield a similar result, give it up.

That’s why I bring this “Simple 3 Assets Buy & Hold Strategy” here to check if buy-and-hold 3 assets strategy works well.

🚨 If Simple Strategy Works, Why Not Using It?

Simple strategies have one big advantage: they remove unnecessary complexity.

So I wanted to test this question, if a plain buy-and-hold strategy using the same 3 assets can deliver similar or even better results, then maybe the trend-following indicator is not needed at all.

That’s the real reason I built this comparison.

Paid content to unlock the following:

• The 3 symbols used in this strategy

• Full backtest setup details (instrument, fees, execution logic)

• Source code (Python + VectorBT) ready to run in Google Colab

BONUS For Paid Member ONLY - “Champion’s Fast Lane Playbook: Master Algo Trading in 12 Weeks” - My 8 modules self-study guides and notes for building a winning strategy.

(Paid Content Below)

⬇️⬇️⬇️

Read more

For those new to quant trading, assessing progress can be daunting. Unlike traditional trading, quant trading relies on automated systems driven by data and code, making it critical to have objective measures of success. The foundation of this evaluation lies in backtesting—simulating your trading strategy on historical data to gauge its performance. Backtesting over at least five-year timeframe is ideal, as it captures a full market cycle, including ups and downs, providing a robust picture of your strategy’s resilience. By systematically measuring your progress through defined levels, you can ensure your algo trading system is on the right track.

Level 1: Don’t Blow Up Your Account

The first milestone in algo trading is simple but critical: avoid wiping out your account. It may sound obvious, but many beginners underestimate how quickly losses can spiral. A strategy that survives a five-year backtest without catastrophic losses demonstrates basic stability. This level is about ensuring your risk management—such as stop-losses, position sizing, and diversification—is robust enough to weather market volatility. If your backtest shows your account staying intact over the long run, you’re laying a solid foundation for further growth.

Level 2: Achieve Positive Returns

The next step is to generate positive returns in your backtests. This milestone confirms that your strategy is not only surviving but also profiting. Positive returns over a five-year period indicate that your algorithm is capturing market opportunities effectively. While this may seem like a basic goal, reaching consistent profitability in backtesting is a clear sign you’re moving in the right direction. It validates that your trading logic and parameters have potential, setting the stage for more advanced benchmarks.

Level 3: Outperform Buy-and-Hold S&P 500

Once you’ve achieved positive returns, the next question is: why bother with quant trading if it can’t beat a simple benchmark like buy-and-hold S&P 500? This level is a significant achievement, as outperforming a major index like the S&P 500 shows your strategy adds value beyond passive investing. In backtesting, compare your strategy’s annualized returns to the S&P 500’s performance over the same period. If your returns are higher, celebrate this milestone—it’s proof your system is on the right track, assuming your backtesting methodology is solid.

Level 4: Monitor KPIs Against the Benchmark

At this stage, it’s time to dive deeper into your strategy’s performance by analyzing key performance indicators (KPIs). These metrics provide a detailed view of your system’s strengths and weaknesses compared to the S&P 500. For a comprehensive list of KPIs, refer to the post Mastering Algo Trading: 5 Key KPIs to Boost Performance. Focus on metrics like:

• Sharpe Ratio: Measures risk-adjusted returns—higher is better.

• CAGR (Compound Annual Growth Rate): Reflects consistent growth—aim to beat the benchmark.

• Max Drawdown: Tracks the largest peak-to-trough loss—lower is preferable.

• Calmar Ratio: Combines return and drawdown—higher indicates better risk-adjusted performance.

If your strategy’s KPIs surpass those of the S&P 500, you’re building a robust system that balances profitability and risk effectively.

Level 5: Be a Better Version of Yourself

Once you’ve cleared the benchmark, your new goal is self-improvement. At this level, you’re no longer just competing with the market—you’re striving to optimize your own system. Fine-tune your algorithm to enhance KPIs, reduce drawdowns, and boost returns. Experiment with new parameters, incorporate additional data, or refine risk management techniques. The aim is to achieve “market wizard” status, where your strategy consistently delivers exceptional performance. This level is about relentless improvement and pushing your system to its full potential.

(Disclaimer: The contents are for educational purposes only, sharing personal experiences and insights in developing and backtesting quantitative trading strategies using principles of quantitative finance. Developing or using quant trading strategies is a method or tool, not financial advice. The information provided, including coding techniques, backtesting methods, or statistical models, is general knowledge and not specific recommendations to buy or sell securities. It is not related to investment opportunities, passive income, debt relief, financial services, or financial consulting. I do not provide financial advice or services)

In my previous post, i shared my first news-driven strategy.

It is just a start.

Warren Buffett told us not to put all our eggs in one basket.

Okay.

I decided to build a few different LLM-driven news tools — not just with different prompts, but from entirely different perspectives.

This time, I am not asking the LLM to analyze the news and suggest BUY or SELL.

Instead, I ask the LLM to find the market’s downside catalysts.

🧠 Why There Is Another LLM Trading Strategy?

At the very beginning, I plan to have a couple of LLMs to generate a number of trading signals.

I don’t plan to depend on a single prompt or a single model. At the end of the day, I am going to put real money into trading.

Diversification of the strategies, i.e., different prompts, different models, and different methodologies, is always better than depending on a single one.

After several weeks of dry run for my very first LLM news trading strategy, I find that it is a little bit optimistic about the market news.

It is not a bad thing, as there are more bull days than bear days in the history of the stock market.

I am not going to refine the prompts. Instead, I am going to develop a brand new LLM trading strategy using a totally different perspective.

The way of analyzing the news is not simply asking the AI model if it is a BUY/SELL recommendation for a given piece of news.

Instead, when there is a market downturn, I ask the AI model to find out the reason for the drawdown and track it.

Part 1/4 Prompts -

System Prompt:

You are an expert financial analysis assistant. Your task is to analyze news and find the exact catalysts/reasons for the price drop in {SYMBOL}. You must follow instructions precisely and return output ONLY in the specified JSON format.

User Prompt:

Today’s date (US Eastern Time) is {YYYY-MM-DD}. Search for and extract the main catalysts and reasons for the price drop of {SYMBOL} on the {session} session of {YYYY-MM-DD}. The price change during this session was {pct_change%} (reference price: {reference_price}, current price: {current_price}).

<🔒 - Paid Member For Full Prompt>

Expected output format: {”reasons”: [{”headline”: “...”, “explanation”: “...”, “source_urls”: [”...”]}]}

🚨 UNLOCK: The Market Crash LLM News Analysts Workflow To Help You Survive In The Market

Everyone is “guru” in the bull market. The real gurus are those who can survive in major drawdown and keep their profit!

This market crash LLM news analysts workflow is exactly designing to helping people to survive in major drawdown.

You now have my system prompt to analysis the news, but it is only part of it. The key of this news analysis trading strategy is not a single prompt, but the overall workflow how different prompts work together.

In the paid section, I reveal the exact workflow, LLM prompts and full explanation on how I track the market drawdown catalysts and convert to tradeable decision:

• The Complete System Design: You have all the prompts, but it is not enough. The critical path is to how they glue together for a complete function.

• All Four Prompts: There are 4 prompts (Part of the 1st prompt shared above) included in this news analysts system. They all have different functionality yet critical to the final outcome.

• How To Aggregate With Previous News Analysis LLM: There are 2 tools now. But to use them? I will show you

• Next Step: Don’t jump in too soon, learn how to launch it phase by phase without risking your money!

BONUS For Paid Member ONLY - “Champion’s Fast Lane Playbook: Master Algo Trading in 12 Weeks” - My 8 modules self-study guides and notes for building a winning strategy.

🔥 Unlock the full blueprint now and survive in the market can’t be easier

Join paid to read the full breakdown 👇👇👇

Read more

The main problem for most investors is that, they don’t have an edge.

An edge is some trading method that it can yield a positive outcome in a long run, not just by luck.

One of the “edge” trading method out there is probably trend following system.

Imaging if you can escape from down trend market like Covid crash in 2020 or rate-hike bear market in 2022.

That is why I built an AI agent skill to help to identify if a stock is in uptrend, inspired by legendary trader, Mark Minervini.

“My goal is not to buy at the lowest or cheapest price but at the ‘right’ price, just as a stock is ready to move significantly higher. Trying to pick a bottom is unnecessary and a waste of time; it misses the whole point.”

— Mark Minervini

The AI agent just to answer one useful question:

Is this stock acting right enough to be invested right now?

So I tested it on QQQ.

Just one mechanical decision, made using only information that was available on that trading day.

• Return Potential: ★★☆☆☆

• Ease Of Adoption: ★★★★☆

• Drawdown Protection: ★★★☆☆

🆘 Common Problems

Most of the algo traders who are backtesting their strategy begin in the right way.

But then, they want to have a better backtest results.

They add more parameters, more indicators, more rules etc.

The backtest performance metrics improve, a higher Sharpe ratio, a lower drawdown.

But once the strategy goes live, the performance is a disaster.

It is because the strategy learning the past too well.

That is curve fitting.

🚨 Simplicity Is the Edge. Buy The Stock At The Right Time!

The goal here was not to build the smartest-looking strategy.

It was to build one that could be explained in a sentence.

When QQQ is in an uptrend, own QQQ. When it is not, hold cash.

That is it.

The strategy uses one trend indicator with two parameters.

Results (2020–2026):

• Sharpe Ratio: 1.03 ✅

• Profit Factor: 2.21✅

• Max Drawdown: -20.2% ✅ (🎉 40% better than buy and hold S&P500)

Paid content to unlock the following:

• The code for the up trend detection to build the indicator

• Full backtest setup details (instrument, fees, execution logic)

• Source code (Python + VectorBT) ready to run in Google Colab

BONUS For Paid Member ONLY - “Champion’s Fast Lane Playbook: Master Algo Trading in 12 Weeks” - My 8 modules self-study guides and notes for building a winning strategy.

(Paid Content Below)

⬇️⬇️⬇️

Read more

In 2025, I developed several trading systems and picked 5 to carry on in 2026. The details of the systems are as below:

The 5th Trading Systems terminated in paper trading phase, now i have 4 systems lived.

📈 Market Recap - Another Struggling Month

July is another high volatility month for the momentum stocks. A sledgehammer-like drop for high beta tech stocks like computer chip, memory, etc.

However, the S&P500 dropped only 0.1%.

It is not a good month for my portfolios especially most of my strategies are long related strategies.

📣 Quick Update On My Four Systems

Quant Trading System#1 - Risk Parity Broad Market ETFs Strategy

🟢 The design of this system is based on low-correlated assets. From Mar to Jun, those assets have been moving in the same direction, making the system’s returns highly volatile. In Jul, the correlation among those assets started to drop down. Making this strategy is the “best” among other strategies.

Quant Trading System#2 - Volatility Based & Alpha Factors ETFs Strategy

🟡 July market regime is not favor for this high beta strategy.

Quant Trading System#3 - Hit & Run Stock Momentum Strategy

🟢 In July, there is not much trading opportunity for hit and run for momentum stocks. After all, this choppy market environment made a lot of trades end up with stop loss.

Quant Trading System#4 - Sector Rotation Driven ETF Strategy

🟢 This is another disappointed month after Jun. The stock pick for July is similar to June, it looks like a good one at beginning of the month when we do the rotation. But after that, those ETFs went down for the whole month.

We’ll have another pick this month — let’s see.

⚠️ Disclaimer: The content provided is intended solely for educational purposes and reflects the author’s personal experiences with data insights, backtesting and algorithmic trading. It does not constitute financial advice, investment recommendations, or financial services. All references to securities or financial instruments are for illustrative purposes only and should not be construed as endorsements or recommendations for trading. Performance figures presented are based on hypothetical backtested results, which are not indicative of future performance. Trading systems may contain errors, and no warranties are made regarding their accuracy or functionality. Past performance does not guarantee future results. Individuals are strongly advised to conduct thorough due diligence, including but not limit to forward testing in paper or simulated accounts for an appropriate duration, before deploying any trading strategy with real capital. Trading carries significant risks, including high volatility and the potential for complete loss of invested capital. Live trade the strategy at your own risk, as any losses incurred are not the responsibility of the author.

Leopold Aschenbrenner, a former OpenAI Research, caught up a once in a life opportunity earlier than everyone else.

He made huge amount of money by his correct view on AI-boom, the biggest market themes of the decade. This AI-boom benefits a wide range of industrial plate like compute, chips, memory, data centers, and power, etc.

He is clever enough to b…

Read more

This post is for anyone who wants to understand how top-tier institutional traders build and protect edge.

Most retail traders stare at stock charts. Meanwhile, the rates desks at Goldman Sachs are quietly harvesting premium from the options market using a repeatable, data-driven workflow that most people never see.

On 2010, USD swaption implied volatility exploded. Since then, it’s come back down but not evenly. Some parts of the surface are back at historic lows. Others are still running hot.

That gap is where the money is.

Goldman Sachs published this workflow as an open-source notebook in their gs-quant library on GitHub (ref: here). I studied and found out how the industrial giant quant trader trades in six steps workflow.

You don’t need a Goldman badge or a Marquee API token to understand the framework. The process itself is what matters.

Step 1 — Scan the Volatility Landscape

This is where every trade begins: a systematic scan of the volatility surface.

The setup: The notebook creates eight USD swaptions — receivers and payers, each with a 3-month expiry, on both 1-year and 10-year underlying swap tails. Strikes are set at-the-money (ATM) and 50 basis points away from ATM.

currency = Currency.USD
expiry = '3m'
notional = 1e8

tenors = ('1y', '10y')
r_strikes = ('ATMF', 'ATMF-50')
p_strikes = ('ATMF', 'ATMF+50')

receivers = [IRSwaption(PayReceive.Receive, tenor, currency, expiration_date=expiry, strike=strike, notional_amount=notional, name='{},{},{}'.format('R',tenor,strike)) for strike in r_strikes for tenor in tenors]
payers = [IRSwaption(PayReceive.Pay, tenor, currency, expiration_date=expiry, strike=strike, notional_amount=notional, name='{},{},{}'.format('P',tenor,strike)) for strike in p_strikes for tenor in tenors]

swaptions = payers + receivers
port = Portfolio(swaptions)

dates = pd.date_range('2018-01-01', '2020-01-01', freq='BM').date.tolist()
dates.append(dt.date(2020,1,24))

with HistoricalPricingContext(dates=dates): 
    vols = port.calc(IRAnnualImpliedVol)

Two years of historical implied volatility data is pulled for every swaption in the basket, sampled at month-ends.

Why this matters: This is essentially a heat map of how expensive or cheap the options market is across the curve: right now versus where it has been historically.

In the chart plot

• 1-year tail vols have fully retraced the summer spike — back at their lows

• 10-year tail vols have NOT fully retraced — still elevated relative to history

What a Goldman trader sees here: An asymmetry. If 10-year tail vol is still expensive but the fundamental driver has faded, there's a systematic premium sitting there, waiting to be harvested by someone willing to sell options on that part of the curve.

But knowing where vol is expensive is only Step 1. The real edge comes from what happens next.

Step 2 — Dig Into the Skew (Confirm the Trade Thesis)

Implied volatility alone doesn’t tell the whole story.

Skew, i.e. the difference in implied vol between out-of-the-money strikes and at-the-money strikes, reveals how the market is pricing tail risk on each side.

What the notebook measures:

• Receiver Skew - Vol of 50bp OTM receiver minus ATM receiver

• Payer Skew - Vol of 50bp OTM payer minus ATM payer

Panel 1 — 3m10y Skew: Both receiver and payer skew on the 10-year tail remain high. The vol smile (both wings elevated) is steep and roughly symmetrical.

Panel 2 — 3m1y Skew: A very different picture. Payer skew has flattened while receiver skew remains rich. The market is paying up for downside rate protection on the short end only.

Update to now, you can see that both vol level and vol smile (both wings) are unusually elevated. The market is overpricing rate volatility in both directions on the long end.

BUT. What To Trade?

Disclaimer: This commentary is for educational purposes only and is based on publicly available Goldman Sachs material and the gs-quant GitHub repository notebook. It is not affiliated with Goldman Sachs, endorsed by Goldman Sachs, or intended as investment advice. Any trade setup discussed here is illustrative only and should not be construed as a recommendation for any current or future position.

🔒 UNLOCK: Here’s Where It Gets Actionable...

Knowing where the opportunity is (Steps 1–2) is just the starting point.

The questions that separate a profitable desk from someone who simply blows up selling options are:

• ❓ Exactly what strikes and expiry maximise the risk/reward?

• ❓ How does P&L evolve month by month — and how bad can it get?

• ❓ When during the trade’s life is a vol spike most dangerous?

• ❓ How far do rates need to move before you actually lose money?

If you've ever wanted to think about rates volatility the way an institutional desk does. NOT guessing, NOT gambling, BUT running a repeatable process — this is the post.

Join 1,800+ quant traders already using these steps now 👇👇👇

Read more

In the world of quantitative trading, backtesting is a cornerstone for developing strategies, rooted in the fundamental assumption that historical market patterns can repeat in the future. This belief drives traders to analyze past data, hoping to uncover strategies that will perform well in live markets.

But does a successful backtest truly guarantee su…

Read more

This post is for anyone who wants to diversify the portfolio to decrease risk but not sacrificing return.

Warren Buffett's most famous rule: Don't put all your eggs in one basket.

Simple advice. Everyone knows it.

So you build a portfolio with 5 stocks:
• NVIDIA
• AMD
• TSMC
• Broadcom
• Micron

Five different companies. Five different baskets. You're diversified, right?

WRONG.

During the March 2026 Iran war selloff, all five dropped 15-20% in three weeks. Same direction. Same timing. Same pain.

Here's the uncomfortable truth most retail traders miss:

The mistake isn't the number of stocks. It's the hidden way they behave together under stress.

You can own 50 stocks and still watch your portfolio crater in unison. Or you can own 8 stocks that move independently and cut your drawdown in half.

Institutional quant desks don't guess at this. They measure it.

But retail traders? They count stocks and hope for the best.

But Won't Diversification Kill My Returns?

Most retail investors avoid proper diversification because they think it means giving up gains.

More stocks = lower returns = boring portfolio.

I thought the same thing.

Then I tested it.

I rebuilt my portfolio using the exact framework institutional desks use to measure true diversification.

Same return potential. 48% less drawdown.

I didn't add random stocks. I added the RIGHT stocks using a specific measurement that shows how they'll behave when markets break.

The difference between a 31% drawdown and a 16% drawdown isn't luck.

It's knowing WHICH stocks to add and WHEN to stop adding them.

That measurement framework is what separates retail guessing from institutional precision.

And it's exactly what I'm sharing inside the paid section.

Here's What Most Investors Miss

Most people think diversification means owning more stocks.

It doesn’t.

You can own 5 stocks, 10 stocks, even 20 stocks — and still watch them fall together when the market turns. That’s because the real risk isn’t how many names you hold. It’s how they behave together under stress.

That’s the part most retail investors never measure.

I used to think adding more stocks automatically made a portfolio safer too. Then I tested it properly and found something surprising: after a certain point, adding more names does almost nothing unless they improve the right kind of diversification.

Inside the paid section, I’ll show you the exact framework I use to measure that, decide how many stocks are actually enough, and know when adding another position stops helping.

🔒UNLOCK: The Missing Part Of Putting Eggs Into The Basket

I spent 2,000+ hours to find out how to protect my capital (i.e. lower max drawdown) WITHOUT sacrificing my return (i.e. better Sharpe Ratio).

The missing part here is to measure the true diversification.

The full decision matrix:
✅ Every risk target — 10%, 15%, 20%, 25%, 30% drawdown scenarios
✅ Every correlation band — what happens at 0.3 vs 0.5 vs 0.7 vs 0.9
✅ Every stock count threshold — exactly how many stocks you need at each level
✅ Cost-effectiveness breakpoints — when adding another stock wastes effort

This isn't theory.

I used this exact framework to rebuild my portfolio from 5 stocks to 12 and cut my max drawdown by half without sacrificing returns.

The paid content shows you how to build a true diversified portfolio.

No guessing. No hoping. No Googling calculators and wondering what to do with the number.

Just the complete framework and your specific answers.

And this exact measurement methodology is the secret that helped me win Quantiacs Q22 Quant Contest Competition

The Exact Framework: How Many Assets, What Correlation, and When To Stop Adding

This is the paid section. Everything below is based on real simulation data — 400 scenarios, 16 years of data

🔥 Unlock The Exact Diversification Framework I Used To Cut Drawdown By Half — Proven In a Live Trading Contest 👇👇👇

Read more

Survivorship bias is a critical issue in algorithmic trading, where focusing only on successful assets leads to skewed perceptions and flawed strategies. Many people’s preference for small coins over BTC stems from survivorship bias, where only the “winners” (e.g., DOGE, SHIB) are remembered, while the 99.9% of small coins that crashed to zero are forgo…

Read more

This post is for anyone who wants to understand how top-tier institutional traders build and protect edge.

I was 10% down in a few weeks when a news shock hit my portfolio.

Most of my strategies are long. That is intentional. My defense against drawdown was not market timing, but holding uncorrelated assets next to my stock portfolio. Different sleeves, different stories, so one bad week would not hit everything at once.

Then the shock came.

The “uncorrelated” pieces moved the same way as equities at the same time. Correlation went to one under stress. Diversification did not fail in theory. It failed on the day I needed it.

That was the real lesson: my hedge stack was built for slow risk, not for a news-driven drawdown.

Why I started looking

I still want to run long. I am not flipping to cash every time headlines get loud. But after the Iran escalation ran for a month and took 10% with it, I needed something better than "hold uncorrelated assets and wait."

I went looking for how professional desks actually handle this. Not a textbook answer. Not “buy puts.” The real workflow a bank quant runs when a shock hits and the front end of the vol curve starts pricing fear differently from the months ahead — exactly what happened in March during the geopolitical flare-up.

Standard resources gave me nothing useful. Options courses teach the instruments. They do not show the decision process behind them.

What Goldman Sachs actually published

Goldman Sachs’ gs-quant repo (here) contains a public Jupyter notebook that most retail traders scroll past. It is not a tutorial. It is a working research pipeline, the kind a desk runs before risk goes on.

What caught my attention: the notebook does not just explain that vol curves can invert. It shows you how to find the inversion across all G10 pairs, confirm whether it is extreme enough to act on, and structure the expression so you are not just buying volatility and hoping.

That three-step logic, i.e. screen, confirm, structure, is what I had been missing. Not a new asset class. A repeatable process for a specific market condition that shows up every time a headline shock front-loads fear into the near term.

Below the paywall, I map that pipeline step by step, so you have a process ready before the next headline week arrives.

🔒UNLOCK: The Missing Implementation Layer

Understanding the concept is just the starting point.

The difference between reading about this and knowing when to act comes down to three decision points, and the ones Goldman's notebook is actually built around.

Below the paywall, you will immediately unlock the full Python workflow:

• The GS Screen Cutoffs: The exact numerical thresholds used to validate a trade before risking capital.

• The Go/No-Go Visualization: How to plot the term structure distortion to make an instant trading decision.

• The Trading Setup: Exact GS quant trading setup breaks down in response to this kind of event

Includes raw Python code snippets and a link to the complete, downloadable .ipynb research notebook.

Join 1,800+ quant traders already equip with this workflow 👇👇👇

Inside the Goldman Sachs Curve Inversions Notebook — Step by Step

This is a detailed walkthrough of Goldman Sachs' public gs-quant notebook. Every code block below is taken directly from the notebook. Every output is real. The workflow is reusable, i.e. not just for the 2020 EURUSD example Goldman used, but for any headline shock that front-loads fear into the near end of the vol curve.

Disclaimer: This commentary is for educational purposes only and is based on publicly available Goldman Sachs material and the gs-quant GitHub repository notebook. It is not affiliated with Goldman Sachs, endorsed by Goldman Sachs, or intended as investment advice. Any trade setup discussed here is illustrative only and should not be construed as a recommendation for any current or future position.

Read more

They think the problem is the model. Usually the problem is the workflow.

The first version of this post was about what mattered. This one is about what breaks.

These are 10 more rules I learned after enough bad prompts, fake backtests, broken scripts, and expensive trial-and-error to sto…

Read more

After winning, I moved on to building multiple trading systems for my own money.

In January 2026, I launched 4 systems live and 1 system in pilot. Those systems backtested with CAGR ranging from +23% to +74%.

But I didn't stop there.

More ideas keep coming. New …

Read more

A few days ago, I wrote about using an LLM to label pre-open NASDAQ news as YES or NO.

This week I rebuilt the entire workflow.

Why? The first version was thinking like a commentator, not a trading system.

That difference matters.

A naive sentiment model reads a scary headline and wants to sell. A trading system asks a harder question: did this type of news actually move the market down in the current regime, or does it just sound dangerous?

That was the flaw in version one.

It relied on “news sounds negative, therefore risk-off.” But markets don’t trade on tone alone. They trade on how participants respond to similar catalysts recently.

That insight changed the architecture.

The new version doesn’t treat the LLM as a predictor. It treats the LLM as a market-response learner.

Read today’s pre-market news. Compare it to the last month of similar news. Only issue a SELL when the market has proven this type of catalyst actually matters.

✏️ What changed

This isn’t just a prompt wrapped around a news feed anymore.

I built a full pipeline: news collection → analysis → signal generation → backtest.

Swap the model? Just change the config. Every new model runs through the same backtest to prove it works before I trust it.

The biggest upgrade? A far more sophisticated LLM prompt loaded with my 20+ years of trading knowledge responding to news events.

The model can say SELL. But the system still has rules.

Threat level matters. Confidence matters. Whether similar news got ignored last week matters. Whether a familiar theme is escalating into something worse matters.

This version solves the real problem: false alarms.

A bad SELL hurts more than most traders realize. Step out of NASDAQ on a scary headline the market ignores? You don’t just miss a trade. You underperform buy-and-hold for no reason.

That’s why I built the system to raise the bar for SELL signals, not maximize how often the model sounds smart.

Most of the models are able to dodge away from the drawdown with similar total return i.e. better risk-based return. The top 2 models are generating better total return as well.

Hooray!

🚨 UNLOCK: The LLM News Workflow That Outperformed Buy-and-Hold in QQQ

The free section showed the problem. It did not show the part that makes this tradeable.

In the paid section, I reveal the exact workflow I now use to turn pre-market NASDAQ news into a cleaner, more disciplined signal, and the production-style design choices that made this version much better than the first:

• ✅ The Exact Prompt: The final prompt structure that stopped the model from acting like a commentator and forced it to think like a market-response learner.

• ✅ The Memory Engine: How the 30-day rolling memory is built so the LLM can compare today’s headlines with how the market actually reacted to similar news recently.

• ✅ The Decision Layer: The hard-rule framework that filters weak SELL calls, reduces false alarms, and stops the system from overreacting to scary-but-ignored headlines.

• ✅ The News Filter: The pre-processing step that removes technical-analysis noise, analyst predictions, and low-value commentary before the LLM ever sees the input.

• ✅ Potential Issue When Backtest LLM For Trading: You don’t want blow up your account once live using LLM. This is a MUST READ if you are trying to use LLM for trading

BONUS For Paid Member ONLY - “Champion’s Fast Lane Playbook: Master Algo Trading in 12 Weeks” - My 8 modules self-study guides and notes for building a winning strategy.

🔥 Unlock the full blueprint for my latest news-driven trading system with a process built for faster response to the markets.

Join paid to read the full breakdown 👇👇👇

Read more

In 2025, I developed several trading systems and picked 5 to carry on in 2026. The details of the systems are as below:

The 5th Trading Systems terminated in paper trading phase, now i have 4 systems lived.

📈 Market Recap - A Struggling Month

June was a rollercoaster for the stock market, especially for the high-growth technology and computer chip companies that had been leading the way up. While the month started with high hopes, those feelings quickly reversed. The turning point was a massive drop in early June, where the semiconductor sector lost over $1 trillion in value in just one day, showing how fast things can change when the market is overly focused on a few hot stocks.

📣 Quick Update On My Four Systems

Quant Trading System#1 - Risk Parity Broad Market ETFs Strategy

🟢 The design of this system is based on low-correlated assets. However, over the past few months, those assets have been moving in the same direction, making the system’s returns highly volatile.

To react faster to the market, I’m exploring whether AI-based news sentiment filtering can be integrated into my trading system. If you’re interested, check out my latest post here

Quant Trading System#2 - Volatility Based & Alpha Factors ETFs Strategy

🟢 It is a high-beta trading system; as long as there is no sudden market drop and the market is moving up, it will beat the market.

Quant Trading System#3 - Hit & Run Stock Momentum Strategy

🟢 In early June, our market-regime detection system issued a RED signal and sold all stocks, moving to 100% cash. Since then, the algo hasn’t picked any stocks, and the P&L has remained flat.

Check out my market regime detection system below

👇👇👇

Quant Trading System#4 - Sector Rotation Driven ETF Strategy

🟢 This system rotates ETFs every month. Momentum is one of the factors used to evaluate selections. In early Jun, semi-conductor and AI-related ETFs were strong, so the portfolio held them. It suffered during the market sell-off in early June.

We’ll have another pick this month — let’s see.

⚠️ Disclaimer: The content provided is intended solely for educational purposes and reflects the author’s personal experiences with data insights, backtesting and algorithmic trading. It does not constitute financial advice, investment recommendations, or financial services. All references to securities or financial instruments are for illustrative purposes only and should not be construed as endorsements or recommendations for trading. Performance figures presented are based on hypothetical backtested results, which are not indicative of future performance. Trading systems may contain errors, and no warranties are made regarding their accuracy or functionality. Past performance does not guarantee future results. Individuals are strongly advised to conduct thorough due diligence, including but not limit to forward testing in paper or simulated accounts for an appropriate duration, before deploying any trading strategy with real capital. Trading carries significant risks, including high volatility and the potential for complete loss of invested capital. Live trade the strategy at your own risk, as any losses incurred are not the responsibility of the author.