Stefan Tzvetanov

Educational only

[Stefan Tzvetanov @learn_phoenix]

Netflix's earnings which sent the stock up 15% after the print serve as an important lesson about market expectations.

It seems like the stock has come a long way since Ackman sold his entire NFLX stake in April 2022, taking a $435M loss.

I still remember listening to the @theallinpod a year ago when @chamath critiqued NFLX’s business model, claiming it will struggle to maintain growth due to saturation in key markets and increasing competition. He also questioned the sustainability of producing a vast array of differentiated high-quality content, arguing that it may dilute quality and brand identity. All valid points.

What I liked about the recent Q4 report and NFLX's 4x growth over the past 2.5 years is that it highlights the importance of expectations in fundamental investing. Market expectations are complex, shaped by narratives, numbers and the opinions of various participants. As long as the company keeps surpassing these ever-changing expectations, the stock tends to rise. This is exactly what has happened with Netflix. Each time a new challenge or bar has been set, NFLX has managed to exceed it so far.

The bar for subscriber growth was already pretty high prior to the print. The buyside was expecting a big beat coming out of Tyson vs Paul, Squid Game 2, Christmas NFL, but the actual print blew even those expectations away. Subscriber numbers are easy to get right directionally (relative to consensus) using alt data, but determining the exact magnitude of the beat is tricky. Clearly, the size of the beat mattered this time.

Like other tech companies, NFLX has faced foreign exchange (FX) challenges - as the dollar has strengthened against other currencies, revenues earned earned abroad translate into fewer dollars. This all seemed to have been priced in ahead of earnings. Retention was also a concern, but NFLX's CEO indicated positive retention behaviour on the call and clarified that the upside in the sub adds wasn't necessarily driven by the Paul vs Tyson fight or Christmas NFL alone.

NFLX will no longer report subs. Attention now turns to monetizing the existing subscriber base. The announced price hikes along with their existing ads plan are good first steps. The company is well-positioned to execute the strategy. As @RadnorCapital summarized it well: “The cash machine is working - scale has enabled them to invest back into the business, while expanding margins. These investments drive engagement, which drives retention and pricing power. Members spend an average of 2 hours per day on the platform, which is meaningfully higher than any other video platform with the exception of YouTube (which I view as the best media asset around).”

That said, valuation at ~35x is quite high and the stock is coming off a nearly perfect set of tailwinds. Expectations are as high as ever - can they keep exceeding them?

[Stefan Tzvetanov @learn_phoenix]

https://t.co/KbIgmMeTLZ

[Stefan Tzvetanov @learn_phoenix]

HOW TO CRUSH THE BALYASNY 4-HOUR CASE STUDY

"Here’s a stock you have never seen before. You have 4 hours to build a model and a long/short thesis with limited access."

This case study format has recently regained traction mainly because Balyasny has been aggressively interviewing junior analysts. Coming off arguably their best year relative to the top multi-managers in 2025, they are likely looking to build on their momentum and are hiring across all levels. The firm is recruiting seasoned senior analysts from other MMs, offering them a fast track to a PM position, and has finally launched their version of the Point72 Academy and the Citadel Associates Program, filling summer intern classes. This is on top of recruiting proven senior leaders from the top MMs across risk, data and operations functions over the past 18-24 months.

The multi-hour in-person case study has also become a common format used by many other hiring PMs and long/short equity funds, including Citadel. The reality is it can be challenging to develop a thesis on a stock you know nothing about in just a few hours. On the bright side, you don’t have to spend seven sleepless nights on a take-home case study while working a full-time job. You show up, spend a few hours working on it and you’re done. It’s also good practice for ramping up on any stock quickly and building a thesis with reasonable forecasts.

We won’t spend time philosophizing about whether this is a good measure of someone’s ability to do the job. Instead, we are going to cover all aspects of the case study:
-- How to understand the business and figure out the core debates around the stock
-- How to build a variant view - finding the “anchor” of your thesis
-- How to forecast and build revenue and margins
-- How to get to the 1-year price target
-- How to calculate risk/reward with a probability-weighted price target
-- Best practices: write-up and other questions.

Setting the stage
For the purposes of this write-up, we will assume that you only get access to the bare minimum: the most recent handful of 10-Qs, a 10-K and earnings transcripts. A bonus would be a spreadsheet with some embedded historicals, a comp table and sell-side estimates or reports.

From the get-go, the most important part of the case study is to tell a coherent story with a variant view, price targets and R/R. That’s it. The point is not to build some overly granular model. Keep it efficient, understand the business, build a reasonable thesis and quantify it with the forecasts of your core drivers. Keep everything high level and only zoom in and be detailed about what matters.

Full write-up linked in the comments.

[Stefan Tzvetanov @learn_phoenix]

Duolingo (DUOL) Down 26% Post Earnings Explained

Q3 revenue and EBITDA beat, but 4Q bookings and EBITDA guidance were below consensus, and more importantly, management said they are prioritizing user growth and teaching quality over near-term monetization.

This is bad because it signals problems with their current conversion from free to paid subscribers based on their freemium model and is a negative for revenue growth, at least in the near term. Furthermore, DAU y/y growth is expected to decelerate from the reported ~36% in Q3 to ~30% in Q4 based on the Sep/Oct run rate. Management also noted that while the AI-powered Max tier is growing, it's performing below their initial high expectations.

In a world of so many free AI language tools and turns, this is concerning. At the same time, gross margin compression is expected to continue in FY25 due to higher AI compute costs (DUOL is one of the top OpenAI API users).

Now, the stock will likely recover at some point. The business is still fine - revenue grew +41% y/y, bookings +33% y/y.

The key piece, regardless of whether you are bullish or bearish on the stock, is that investors see it as a secularly challenged business because of AI (competition from AI language tools and shrinking margins because of AI compute costs), so from now on you will be seeing such excessive moves up and down, +-20/30%.

It is and will be a very volatile stock.

A lot of ups and downs yet to come as bookings/user data deteriorates or improves.

[Stefan Tzvetanov @learn_phoenix]

Fair question

[Compound248 💰 @compound248]

Apparently, if you ask @sama how his $13 billion money-losing for profit not-for-profit can support $1.4 TRILLION of spend commitments, he gets very defensive and starts to attack non-existent short sellers.

I cannot wait for @OpenAI to be a public company.

[Stefan Tzvetanov @learn_phoenix]

Citadel Securities says institutional investors are bearish

[Stefan Tzvetanov @learn_phoenix]

https://t.co/cwRHorr2bk

[Stefan Tzvetanov @learn_phoenix]

MODELING MARGINS LIKE A HEDGE FUND PRO
A 4-step framework top long/short analysts use to model gross margins

When modeling margins, you can take previous period’s COGS and SG&A as percentages of revenue and roll them forward (take a look at image 1).

But by treating all costs as variable, you are assuming every dollar of revenue brings an equal dollar of cost. That’s just not how real businesses scale.

Instead, you can split costs into fixed vs variable. Fixed costs grow very slowly, perhaps with inflation, while variable costs scale with revenue (take a look at image 2).

That small structural change of splitting fixed vs variable costs compounds quickly. As revenue grows 8-9% per year, those mostly fixed dollars get spread over a larger base. Then you see gross margins expanding from 30% to 38% and EBITDA margins climbing from 6% to 17%.

That 6% → 17% jump happens precisely because the cost base is only partially variable. Since a portion of the cost base stays fixed, each incremental dollar of revenue flows disproportionately to profit. In other words, you are seeing leverage on a partially fixed cost base (an important feature of operating leverage). In table 1, EBITDA margin stays flat at 6% because you treat all costs as variable.

Modeling margins generally comes down to identifying fixed and variable costs and deeply understanding the components of COGS and SG&A. Let’s dive deeper and define a 4-step framework that top hedge fund analysts use to model gross margins. Before we get into it, we need to discuss a few important caveats.

If you want to read the full write-up, go to the link in the comments.

[Stefan Tzvetanov @learn_phoenix]

I think Citadel and Millennium experienced more severe Liberation Day drawdowns than peers and haven't yet fully recovered. But this is likely due to higher sector/style correlations across pods or over-concentration.

The problem has become that within sectors (e.g., internet/tech), many pods have been increasingly doing the same types of trades (i.e., not only do different pods cover the same sector, but they also start converging in style - using the same data, looking at inflections, etc.).

The trades of each individual pod tend to be highly correlated to other pods within that sector. That means you can’t get the benefit of sum of squares, meaning that the overall portfolio risk doesn’t shrink as you add pods, but it grows.

It eats up into returns - more aggressive hedging at the platform level or more volatility exposure during drawdowns.

[Stefan Tzvetanov @learn_phoenix]

Citadel and Millennium experienced more severe Liberation Day drawdowns than peers - part of the reason is likely higher sector/style correlations across pods or over-concentration.

The problem has become that within sectors (e.g., internet/tech), many pods have been increasingly doing the same types of trades (i.e., not only do different pods cover the same sector, but they also start converging in style - using the same data, looking at inflections, etc.).

The trades of each individual pod tend to be highly correlated to other pods within that sector. That means you can’t get the benefit of sum of squares, meaning that the overall portfolio risk doesn’t shrink as you add pods, but it grows.

It eats up into returns - more aggressive hedging at the platform level or more volatility exposure during drawdowns.

[Blue Horseshoe @BlueHorse88]

Multi-strategy giants like Citadel and Millennium are only up mid-single digits this year, despite strong equity and credit markets. On the surface, that looks like underperformance but it may tell us more about the plumbing of the system than about trade ideas.

These funds depend on dealer financing to scale hundreds of small, market-neutral trades. If banks’ balance sheets are stressed (weighed down by deeply underwater HTM portfolios and squeezed further by an inverted front end) their willingness to extend balance sheet capacity shrinks. The result: higher funding costs, tighter leverage, and muted hedge fund performance.

In other words, the underperformance of pod-model multi-strats could be a shadow indicator of liquidity stress inside the banking system even as headline markets rally.

[Stefan Tzvetanov @learn_phoenix]

Even if you see ETFs as the top positions of multi-manager hedge funds (Citadel, Millennium, etc), it doesn’t mean they are “investing” in ETFs. Pods are usually market-neutral, so ETFs are a convenient and cheap way to hedge style/sector/factor exposures. Return correlations often pop up due to similar style and sector exposures across teams - this is exactly what it can mean to neutralize them at the firm wide level.

[Stefan Tzvetanov @learn_phoenix]

@blueprintsmb22 Stock is up 21% in the past month, so a lot is likely already priced in. Q4 trends through end of Oct will matter a lot too, heading into the Nov/Dec holiday period

[Stefan Tzvetanov @learn_phoenix]

4. NOPAT and ROIC are key for understanding the intuition behind multiples. You won't be using the NOPAT formula directly, but it shows you what's underneath the multiples that we use every day.

[Stefan Tzvetanov @learn_phoenix]

4 THINGS TOP LONG/SHORT EQUITY ANALYSTS KNOW ABOUT VALUATION MULTIPLES

Quick thread. 🧵

[Stefan Tzvetanov @learn_phoenix]

3. All valuation frameworks are steady-state - they tell you whether a stock price is fair/expensive/cheap in the current state of the world - interest rates, risk environment, sector narratives. If those changes, valuation changes with them (and the change can be quite dramatic!).

[Stefan Tzvetanov @learn_phoenix]

@MikeFritzell Thanks for putting this together.

Could you add https://t.co/0v2SUVx5Vm to Sector niches / Learning? It's an educational Substack on fundamental long/short equity.

[Stefan Tzvetanov @learn_phoenix]

HOW TO PITCH A STOCK LIKE A HEDGE FUND PRO: 10 STEPS

1. Lead with the conclusion and clearly state your variant view

Comment and follow if you'd like to get a template with a full example (CHWY).

2. Quantify your variant view in terms of KPIs and valuation vs consensus

3. Provide evidence for your variant view - what data supports your variant view?

4. Analyze sentiment - how do investors and sell-side feel about the stock?

5. Evaluate positioning - are hedge funds long, short or neutral?

6. Identify the catalyst - what event will prove or disprove your variant view?

7. Explain the setup - why is now the right time for your idea?

8. Outline the biggest risk in your variant view - what's the biggest counterpoint and why is it unlikely to take place?

9. Provide a risk/reward analysis for your variant view - include probabilities and expected value

10. What work do you need to do to gain incremental conviction?

Comment and follow if you'd like to get a template with a full example (CHWY).

[Stefan Tzvetanov @learn_phoenix]

https://t.co/WVzk52B69y

[Stefan Tzvetanov @learn_phoenix]

LEGENDARY CITADEL PORTFOLIO MANAGER: LESSONS FROM OUR CONVERSATION

I call him old school because he grew up in the early Citadel world - he was part of the great talent they developed back then and was shaped by the risk-oriented culture that laid the foundation for the success the firm has today. I also call him a modern great because he is still absolutely killing it in multi-manager land right now.

In our conversation, we covered aspects of idea generation, portfolio construction, career advice and more…