Portfolio Analysis in Python Using QuantStats

There's a particular kind of software that used to require a Bloomberg terminal and a quant team to operate. Now it requires a pip install. QuantStats is one of those packages — and whether that's democratization or dangerous simplification depends entirely on what you do with it.

A recent tutorial from the NeuralNine channel on YouTube walks through the library's capabilities in about 23 minutes, building from basic data retrieval up through multi-asset portfolio construction and a rudimentary MACD trading strategy. The presenter is upfront about the framing from the start: "I am primarily here a technical expert — I'm here to teach you guys the technical part of it, how to implement stuff, how to automate stuff. I'm not the kind of guy who explains to you all the ratios and all the interpretations."

That's an honest disclaimer, and it shapes everything that follows. The video is a technical walkthrough, not a finance course. Keeping that distinction in mind is the key to getting value from it.

What QuantStats Actually Is

QuantStats describes itself as "portfolio analytics for quants" and bundles four distinct capabilities into a single package: data retrieval (historical returns by ticker symbol), statistical metrics (Sharpe ratio, Sortino ratio, CAGR, max drawdown, alpha, beta), visualizations (earnings curves, drawdown plots, monthly heat maps, Monte Carlo simulations), and report generation.

That last one is the party trick. A single line — qs.reports.html(returns, benchmark, output='report.html') — produces a full HTML document containing every major performance metric and a complete suite of charts, formatted and labeled, ready to share. For an analyst who would otherwise spend an afternoon in Excel stitching this together, that's a genuine time saver.

The library pulls historical return data directly via its own utilities, so you don't need a separate data pipeline just to get started. Plug in a ticker symbol, get back a Pandas Series of daily percentage returns, and you're working with the right abstraction immediately — returns, not raw prices — because that's what portfolio performance analysis actually requires.

The Numbers It Produces

Walking through a single-stock example with Apple (AAPL), the tutorial surfaces a few figures worth pausing on. The Sharpe ratio comes in at 0.62, which the presenter notes is probably below the 1-to-2 range most practitioners consider acceptable. The compound annual growth rate lands around 19%. The maximum drawdown — the largest peak-to-trough decline across the full history — sits at roughly 81%.

That drawdown number deserves attention. An 81% drawdown means that at some point in Apple's history, the stock fell from its peak by more than four-fifths of its value before recovering. Most retail investors have never stress-tested their holdings against that kind of scenario. The fact that QuantStats surfaces it in a single function call (qs.stats.max_drawdown(returns)) is legitimately useful — it forces the question into view.

The Monte Carlo simulation feature adds another dimension. The library can reshuffle the last 252 trading days — roughly a year — across hundreds of simulated paths, plotting a confidence band around the actual outcome. As the presenter describes it: "This is more like an analysis for a worst-case scenario for tail risks. So you can see if you would have survived this movement." The framing is intuitive. Whether the simulation's assumptions — that past daily return distributions adequately model future tail risk — hold up is a question for someone with more financial modeling depth than a Python tutorial can provide.

Building a Multi-Asset Portfolio

The tutorial's more interesting practical section involves constructing a weighted portfolio from scratch. The example uses five stocks — Apple, Nvidia, Tesla, Microsoft, and Meta — with manually assigned weights (30% Nvidia, 20% each for Apple and Microsoft, 15% each for Tesla and Meta). The mechanics are pure Pandas: download returns for each ticker, build a DataFrame, multiply by a weights Series, sum across the column axis, and you have a single return stream representing the blended portfolio.

Running that through qs.reports.html() against the S&P 500 as benchmark generates a full comparative report. The tutorial notes the portfolio's Sharpe ratio comes out favorably — likely driven by Nvidia's steep run — which illustrates something QuantStats cannot account for: the difference between historical performance and forward-looking expectation. A backtest that includes Nvidia over the last several years is going to look good. That's not necessarily a trading strategy; it's a history lesson.

The MACD Experiment

The final section ventures into vectorized backtesting territory, which the presenter acknowledges is outside QuantStats' core purpose. Using TA-Lib to calculate the Moving Average Convergence Divergence (MACD) indicator on Apple's price history since 2010, the tutorial implements a simple rule: if MACD is above its signal line at the close, go long the next day. Otherwise, sit out.

The result, when compared against a pure buy-and-hold strategy in a QuantStats report, is instructive: "buy and hold would be the better choice here — we actually don't make very good returns with the MACD." The MACD strategy does produce a slightly better Sharpe ratio, because it's in the market less often and therefore takes on less volatility. But the absolute returns trail the passive strategy.

This is a well-documented pattern in quantitative finance research. Simple technical indicators, applied mechanically, tend to underperform passive strategies on broad U.S. equities over long horizons — particularly after transaction costs, which this backtest doesn't model at all. The tutorial doesn't oversell the strategy; it presents the output and lets the numbers speak. That's the right approach for a technical walkthrough.

What it can't do is fully account for the methodological limitations of vectorized backtesting: no slippage, no transaction costs, no position sizing beyond binary long-or-flat, no consideration of the data-mining bias that comes from testing a well-known indicator on a well-known stock. These aren't criticisms of the tutorial — they're the natural boundary of what a 23-minute introduction can cover.

The Honest Question This Raises

QuantStats is a well-engineered library. The API is clean, the report output is polished, and the barrier to entry is genuinely low. A data scientist with no finance background can be producing Sharpe ratios and drawdown plots within an afternoon.

That accessibility is mostly a good thing. Tools that were previously locked behind expensive platforms or specialized teams reaching a wider audience generally produces better outcomes — more informed investors, more rigorous analysis, more people asking the right questions about their portfolios. The monthly heat map alone, showing which calendar months have historically been brutal for a given asset, is the kind of visualization that retail investors almost never see presented clearly.

The tension worth naming is this: QuantStats makes the presentation of financial analysis easy. It does not make financial analysis easy. The two things are not the same. A well-formatted HTML report with a drawdown chart and a Monte Carlo confidence band carries an implicit authority that the underlying assumptions may not deserve. Knowing that Apple's historical Sharpe ratio was 0.62 is useful context. Knowing what to do with that number — how to compare it across asset classes, how to adjust for the current interest rate environment, how to think about its stability going forward — requires a different kind of expertise entirely.

The presenter is refreshingly direct about this gap: "None of this is financial advice, by the way, because I'm just a programmer, data scientist, machine learning engineer — but I'm not a financial expert." Most tools don't come with that caveat built into the tutorial. Perhaps they should.

For Python practitioners looking to build reporting infrastructure, analyze historical return data, or get a structured introduction to quantitative finance concepts, QuantStats earns its install. For anyone tempted to treat the output as a substitute for financial judgment, the MACD backtest result is worth keeping close — a technically correct report, pointing toward a strategy that underperformed doing nothing.

---

Bob Reynolds is Senior Technology Correspondent at BuzzRAG.