# On the power law of Y Combinator startups

As my long-time readers of my blog posts know, Rebel Fund maintains the largest and most comprehensive database of Y Combinator startups and founders that exists outside of YC itself. We use the millions of data points we’ve collected primarily to train our Rebel Theorem machine learning algorithm to predict which YC startups are most likely to drive investment returns for our fund, but they also allow us to understand and share unique insights on YC startups overall with the wider world.

Our database includes latest valuation estimates¹ for every YC startup in history, which we analyze to identify trends and patterns related to YC startup performance. In this post, I’ll share some surprising statistics on the steep power law that drives YC startup returns for venture funds like us, and argue for why a large portfolio strategy is optimal when it comes to investing in YC startups. This will be a very data-driven post so buckle up!

To start, it’s no secret that technology startup outcomes follow a power law distribution, but I don’t think even most venture investors fully appreciate how steep that curve is. The chart below plots the anonymized valuations of every YC startup from 2017 or earlier vintages (we chose these vintages since it takes 7 years on average for a YC startup to reach maturity) and I added a few waypoints for context:

As you can see, all the action is at the far left end of a *very* steep curve. The vast majority of startups enjoy no significant valuation growth at all, with a few big winners like Airbnb (~$100B valuation) dwarfing even other decacorns ($10B+ valuations), which in turn dwarf even other unicorns ($1B+ valuations), which in turn dwarf the minicorns ($100M+ valuations).

In fact, we’ve found that the unicorns, which represent about 6% of YC startups for these vintages, represent a whopping 90% of the total valuation growth. And of those, the decacorns, which represent just 0.6% of these YC companies, account for more than half of the unicorns’ valuation growth. And of those, just 2 companies — Airbnb & Stripe — represent more than half of the decacorns’ valuation growth!

This power law is so strong that the curve stays pretty steep no matter how much you zoom into it. Here’s it is for just the unicorns:

And here it is for just the decacorns:

An appreciation of the steepness of these curves has some big implications for venture investors — you really need to catch some outliers to achieve meaningful returns, and not all outliers are created equally.

The good news is that while unicorns² and especially decacorns are rare, they drive incredible financial returns for early investors. At Rebel, we’ve also been able to estimate initial valuations for YC startups based on their batch year and our internal valuation trends data, which coupled with our financing round data, allows us to estimate return multiples for early investors net of dilution, and even realized and unrealized investment IRR. Here are some financial return statistics for these same YC vintages:

You may notice is there’s a *big* difference between the median return numbers (which are terrible) and the averages (which are fantastic), which makes sense given how incredibly outlier-driven startup investing is. This is why you need a sound portfolio strategy when it comes to early-stage venture investing — a few random bets here and there is way too risky.

You also may notice that the IRR for ‘top companies’ ($150M+ valuations) and unicorns ($1B+ valuations) are *really* good, and even the overall average IRR is much better than other investment classes like public equities. So, YC startups are a great “asset class” if you know what you’re doing… and have some patience.

I mentioned above it takes 7 years on average for a YC startup to acheive an exit, and the valuation ramp to get there also follows an exponential curve. While the charts above focus on mature YC startups from 2005–2017, the chart below shows the average seed investor net return multiple for YC startups by years since their YC batch from the 2016 vintage to present:

As you’ll see, valuation growth is relatively slow until around the 3 to 5 year mark, and then accelerates dramatically as the big winners start taking off. That’s why most early-stage venture funds have a 10–year life and return the vast majority of capital to LPs in the last few years.

This brings me to the next stage of this post — let’s talk about the optimal portfolio strategy when it comes to YC startup investing.

There’s a common misconception, even amongst sophisticated institutional LPs, that a “concentrated” portfolio of 10 to 20 investments is optimal when it comes to early-stage venture capital funds. While I understand and will explain the logic behind this below, our data shows that it’s flat wrong after accounting for risk, at least when it comes to investing in early-stage Y Combinator startups.

Allocators have reached this conclusion because a lot of the top-performing venture funds historically had concentrated portfolios, and with a small portfolio, a single great investment can ‘return the fund’³

This is true, but there’s survivorship bias — a lot of the worst-performing funds *also* had concentrated portfolios, but they don’t come back to market so you never hear about them. And while yes, it’s wonderful if a single investment can return the fund, what happens if a manager misses that one great investment? We’ve already discussed how rare unicorns and decacorns are, even amongst prestigious YC startups, and thus they’re easy to miss in a more concentrated portfolio, even amongst skilled managers.

To drive this point home, we ran a series of Monte Carlo simulations to empirically test the performance of different YC startup portfolio sizes. For the uninitiated, these are computer simulations that run thousands of random trials to estimate probable outcome distributions based on certain variables (in this case, different YC startup portfolio sizes).

Before sharing our results, I’ll caveat that these simulations differ from the real world in two important ways: 1) no investor actually invests at random, and 2) they assume perfect “access” to YC startups, which isn’t the case for most investors (more on that later).

At first we ran the simulations with the startups ungrouped, such that each trial would include randomly selected 2005–2017 vintage YC startups to see what expected returns various portfolio sizes would achieve. The problem we encountered is the dispersion in YC startup outcomes is so great that the results became too “noisy” to make sense of. For example, if by pure chance a sample portfolio included an Airbnb or Stripe, that one company would impact the portfolio returns so much that no other variable matters.

So, we instead grouped the startups into two categories, unicorns and non-unicorns, such that every startup in each category was treated the same. While simplistic, this quieted the noise so we could see the impact of different portfolio sizes on expected unicorn rates (which is all that matters anyway since the unicorns drive 90% of YC startup returns).

In the interest of brevity I won’t share every portfolio size that we tested, but these examples are illustative:

Here you can see that with a 10-company portfolio, a YC startup investor has a ~55% chance of striking out completely (no unicorns), a ~35% chance of hitting a single unicorn, and a ~10% chance of hitting two unicorns. While a 10% or 20% unicorn rate is fantastic and well-above the 6% average for these vintages, this investor’s odds are better to strike out completely and break even at best… not good⁴.

For a 50-company portfolio, things start to look much better:

Now we’re seeing a normal distribution or “bell curve” start to form around the 6% average unicorn rate, with the vast majority of outcomes in the 3–9% unicorn range. While this investor has virtually no chance of getting a 15% unicorn rate or higher, they also have virtually no chance of striking out completely.

With a 150-company portfolio, the outcome distributions get much tighter around the 6% average unicorn rate. This investor will almost certainly not exceed a 10% unicorn rate (at least not by chance — more on that below) but they’ll also almost certainly not get below a 2.5% unicorn rate, which is sufficient to acheive a modest financial return (similar to public equities).

This analysis actually *underplays* the value of a larger portfolio size since every unicorn is treated equally. In reality, a single decacorn investment can return 1000x to early investors, and maximizing one’s portfolio size also maximizes their chances of catching one of these super-outliers that essentially guarantees exceptional portfolio performance.

Another way to look at the optimal risk/return profile of different portfolio sizes is analyzing their average unicorn rate, standard deviation, and Sharpe ratio. Without going into all the technical details, an optimal portfolio acheives the highest return (unicorn rate) with the lowest risk (standard deviation) and highest Sharpe ratio (risk/return balance).

Here we see that the average unicorn percentage is quite consistent at the different portfolio sizes, but the standard deviations and Sharpe ratios only start to stabilize around 50-75 companies, and continue to improve as portfolio sizes grow. In other words, to get the highest return with the lowest risk, you want as large a YC startup portfolio as possible (though you start to see diminishing marginal returns after 50-75 companies).

Now here’s the kicker — these charts only examine the component of *chance* at different portfolio sizes, not *skill*. If an investor is skilled at selecting the best startups for their portfolio, they can “break free” from these averages, effectively increasing their unicorn rate and thus returns, while still minimizing risk with a larger portfolio size.

For example, if an investor with a 150-company portfolio investor can simply avoid the bottom half of YC startups, then their portfolio would acheive around a ~12% unicorn rate with very little risk. This would translate into a ~42% expected portfolio IRR net of dilution, which is of course excellent⁵.

While selection skill adds value at every portfolio size, the other side of the coin is *access*. Unlike public equities, which anyone with a brokerage account can buy at will, YC startup investing requires access to the founders, who ultimately get to choose which investors they’d like to partner with — and since it’s not uncommon to see ‘hot’ YC startup rounds heavily oversubscribed, the founders can often afford to be choosy.

To underscore the importance of access, imagine that you’re targeting a 20-company portfolio, which statistically should expect about 1 unicorn, and you get into 19 of your target deals… but the *one company* that rejects your investment offer ends up being that unicorn. In this sad but entirely plausible scenario, your expected portfolio returns are close to zero.

This is why when I started building Rebel in 2019, my first priority was to optimize for YC startup access. Our approach to this is a streamlined and founder-friendly investing process, a team of top YC alumni as partners (co-founding startups now valued at over $100B in aggregate), helpful portfolio support at major company inflection points, pre-Demo Day investing, a check size that’s easy for founders to accept, and much more. It’s paid off — we ended our first fund with a >98% deal win rate and our second fund has a 100% deal win rate so far after dozens of investments.

Without meaning to, in this post I’ve divulged the key aspects of Rebel’s investment strategy. We target a large portfolio of 150+ startups per fund, while aiming to invest in the top 10% of YC startups in each batch (which represent ~96% of all valuation growth) via a data-driven multi-stage diligence process, and avoiding adverse selection at all costs. Even if we don’t select the top startups perfectly, all we *really* need to do outperform YC averages, which isn’t too difficult with our infrastructure. We also reserve 30% of our capital to “double down” on our best investments in later rounds, which is a tactic for concentrating capital in the outliers without the risk associated with more concentrated portfolio sizes, albeit at higher valuations.

I warned this post would get heavy, but I hope it gives you an appreciation of the steep power law curve in venture investing, why a large portfolio strategy is optimal, and the importance of strong selection and access when it comes to YC startup investing. Happy hunting!

¹We use a combination and public and private data sources to estimate latest YC startup valuations in near real-time. We get some valuation data directly, and also infer private company valuations from various clues like the size, series, and timing of announced financing rounds and even a startup’s position on YC’s Top Private Companies by Valuation list. Public YC companies are marked to their latest market capitalizations. While our valuation estimates aren’t perfect, they’re quite accurate overall.

²I should make clear that while ‘unicorn’ is a binary term (a startup is either a unicorn or it’s not) the unicorn mark is a rather arbitrary waypoint on our startup valuation curve. In other words, a $950M company is 95% as valuable as a $1B company, and nothing for early investors to balk at. That said, “unicorn” is still decent shorthand for an outlier startup.

³There are less quantifiable reasons that some LPs and GPs prefer concentrated portfolios as well, like having more influence and control over portfolio founders or more efficient portfolio management, but here I’m focused just on the mathematics of it.

⁴I really mean “not good” for an institutional fund like us entrusted with many millions of other people’s dollars. If you’re an angel investor betting with money you can afford to lose, then a smaller portfolio could make sense. There are also many good reasons to invest in startups beyond financial returns, like supporting the next generation of innovators, helping out a friend, learning about technology trends, or just having fun!

⁵All of the data in this post is based on YC startups. While I suspect non-YC startups follow a similar power law dynamic, YC only accepts <1% of startup applicants and we have lots of data showing that YC startups outperform their non-YC peers by a wide margin. So, non-YC focused investors probably need to assume more risk with more concentrated portfolios to acheive these kinds of returns.