Here’s a fact that breaks most people’s brains: In venture capital, one startup can be worth more than the other 99 in your portfolio combined.

Not “a little more.” Not “twice as much.” We’re talking 50x, 1,000x, occasionally 10,000x returns. The entire game runs on a handful of wildly improbable winners that shouldn’t exist but do.

This creates an obvious problem. If you’re a venture fund, and your whole business model depends on finding needles that are rarer than needles, how many haystacks should you search?

Most early-stage funds bet on 20 to 40 companies per fund. At Team Ignite, we think that number is way too low. We’re building portfolios of 300 to 500 companies. And we have math to explain why.

The Lottery Ticket Problem

Imagine you’re playing a lottery where one ticket in 10,000 wins $100 million, but most tickets are worthless. How many tickets would you buy?

The obvious answer: as many as you can afford, as long as you’re good at picking tickets.

Now replace “lottery tickets” with “startups” and you’ve got the venture capital model. Except there’s a twist. Unlike actual lotteries, some investors are genuinely better at picking. They see patterns, spot talent, understand markets. That skill matters.

But here’s what also matters, and what most people underestimate: volume.

Even if you’re great at picking, the math of power law distributions means you need a lot of shots to reliably capture the extreme winners. A 20-company portfolio might produce solid returns if you hit one 50x outcome. But the companies that generate 1,000x or 10,000x? Those are so rare that unless you’re taking hundreds of swings, you’re mostly hoping to get lucky.

What We Actually Did

We didn’t just theorize about this. We built a model.

Here’s what we know about how startups actually perform:

About half go to zero
A quarter return 1x to 3x (you get your money back, maybe a little more)
A smaller chunk hits 3x to 10x
A few percent land between 10x and 50x
A tiny sliver goes 50x to 200x
Then there’s the freaky stuff: 200x, 1,000x, 10,000x, even higher

This isn’t a neat bell curve. It’s a power law, meaning the right tail (the rare, massive wins) completely dominates the math.

So we ran thousands of simulations. We created imaginary funds with 50 companies, 100 companies, 300, 500, all the way up to 1,000. Each simulation pulled randomly from the same distribution of outcomes. Then we tracked two things:

What was the typical fund result? (The median)
What were the odds of catching at least one monster winner?

The results were striking.

Finding One: The Average Lies

First surprise: the average (or “expected”) return barely changed as portfolio size increased. Whether you had 50 companies or 500, the mean outcome stayed roughly the same.

Why? Because in power law math, the extreme winners carry so much weight that they define the average no matter what. Adding more companies doesn’t change the theoretical expected value if you’re drawing from the same distribution.

So if you only looked at averages, you’d think portfolio size doesn’t matter.

That would be extremely wrong.

Finding Two: The Median Tells the Truth

The median fund outcome (the typical result, the one in the middle) told a completely different story.

As we increased portfolio size, the median return climbed steadily and significantly:

A 50-company fund produced decent returns in a typical scenario
A 100-company fund did meaningfully better
Around 300 to 500 companies, the median outcome was several times higher than at 50

Translation: more shots don’t just give you a chance at the mega-outcome. They make strong results normal instead of lucky.

This is what limited partners (LPs, the people who invest in venture funds) actually care about. They’re less interested in your one-in-a-hundred best case and much more interested in what’s likely to happen fund after fund, year after year.

Finding Three: You Can Engineer Luck

The most intuitive insight came from tracking tail capture. We measured the odds that a fund contained at least one truly gigantic winner.

Under our assumptions:

A 50-company fund had a small chance of landing a 1,000x outcome and a very small chance of a 10,000x
A 300-company fund had better than 50/50 odds of a 1,000x and meaningful odds of a 10,000x
A 500-company fund pushed those probabilities higher still

The takeaway: portfolio size isn’t just about diversification. It’s a lever that controls how much you’re relying on luck versus design when hunting for the companies that define the entire asset class.

At 300 to 500 positions, extreme winners stop being rare miracles and start being probable outcomes.

Why We Landed on 300 to 500

So why not 1,000? Why not 50?

The answer is that 300 to 500 hits a sweet spot across four dimensions.

Tail capture. In this range, you’re sampling the right tail of the distribution densely enough that backing a 100x or 1,000x company becomes likely, not hopeful. You’re not waiting for an invitation to the one mega-round per vintage. You’re creating your own opportunities.

Repeatability. The median fund performance in this zone is strong and stable. That means you can deliver high-quality outcomes consistently, which is what LPs actually want.

Downside protection. More shots don’t eliminate risk (early-stage is still early-stage), but they significantly reduce the chance that a fund gets dragged down by a few unlucky failures.

Operational reality. You can’t run a 1,000-company fund with a small team and actually support founders well. There’s a limit to what’s executable. Our infrastructure (more on that in a second) lets us operate at scale, but we’re not trying to index the entire market.

Why This Works Now (and Didn’t Before)

Two big shifts make this strategy viable today in a way it wasn’t a decade ago.

First: AI changed the sourcing game.

We see roughly 10,000 early-stage startups per year globally. A traditional fund would drown in that volume. We use AI to filter and rank the entire flow before a human ever opens a deck.

The machine ingests signals from product traction, hiring patterns, open source contributions, founder backgrounds, and dozens of other data sources. Then it surfaces the top 1 to 2 percent for our partners to evaluate deeply.

This means we can take hundreds of shots per year without sacrificing selection quality. Volume and rigor aren’t trade-offs anymore.

Second: founder expectations evolved.

Founders increasingly value speed, clarity, and tactical expertise over long courtship rituals. They don’t want a six-month process. They want smart money that moves fast and helps them solve real problems.

A high-volume, high-conviction strategy fits that world. We can make decisions quickly, be transparent about our thesis, and then get to work.

The Risks (and How We Manage Them)

A strategy like this obviously has risks. If you just spray capital everywhere, you end up with shallow relationships, weak support, and eventually adverse selection. You can also accidentally become an index fund and lose the edge that comes from strong conviction.

We think about three guardrails:

Selection quality. Volume only helps if you maintain high standards. Our filter isn’t “invest in everything interesting.” It’s “use AI to surface the potentially great, then be ruthlessly selective within the top 1 percent of what we see.”

Follow-on discipline. Not every company gets the same attention or capital. A power law portfolio requires doubling and tripling down on the rare companies showing true breakout potential. We’re comfortable with that concentration.

Domain focus. We stick to B2B SaaS, AI, fintech, and marketplaces generally and avoid things we don’t understand like Biotech. These are sectors where we have deep pattern recognition and can add real value. High volume doesn’t mean chasing every shiny thing.

What This Means for You

If you’re an LP, the pitch is simple. Venture is power-law driven. Capturing extreme winners is a function of both skill and sample size. Modern infrastructure makes it possible to run a much larger, higher-quality portfolio than was feasible before. We’re intentionally designing fund construction so that 100x and 1,000x outcomes are statistically probable, not just aspirational.

If you’re a founder, the message is equally clear. We’re set up to see a very broad slice of the market. When we partner with you, it’s because we believe your company has the potential to sit in that right tail. We’ve built a platform and network (Team Ignite) that can support a large community while still leaning in hardest when you start to break out.

One Last Thing

Early-stage venture isn’t about gentle diversification around an average. It’s about finding and backing the rare companies that define decades.

If that’s the game, portfolio construction should reflect it. More shots on goal, with real selection discipline, isn’t a gimmick. It’s an attempt to match strategy to the actual mathematics of the asset class.

The power law is brutal and beautiful. We’re just trying to play it honestly.

Appendix A: The Math Behind the Model

Here is the simple quantitative engine powering our portfolio construction thesis. Early stage venture does not behave like a bell curve. It behaves like a power law. The right tail is so heavy that it determines almost the entire distribution of fund outcomes.

1. A realistic outcome distribution for individual startups

We modeled a fund as a set of draws from a stylized distribution that roughly matches what decades of venture data shows. For each startup:

50 percent go to zero
25 percent return 1 to 3 times capital
15 percent return 3 to 10 times
7 percent return 10 to 50 times
2 percent return 50 to 200 times
0.7 percent return 200 to 1,000 times
0.2 percent return 1,000 to 10,000 times
0.1 percent return 10,000 to 100,000 times

The exact percentages matter less than the shape. This is not a normal distribution. It is a very fat tail. A tiny number of outcomes contribute a giant share of value.

2. The simulation

For each portfolio size, from 50 to 1,000 investments, we ran thousands of Monte Carlo simulations. Each simulated fund was created by randomly drawing from the outcome distribution above.

For every simulated fund we calculated:

Fund MOIC (average multiple of invested capital)
Median and percentile outcomes across simulations
The probability the fund contained at least one 50x, 100x, 1,000x, or 10,000x outcome

This gives a clean way to see how portfolio size affects both the typical result and the probability of capturing the extreme winners.

3. What the math showed

Three findings mattered most.

First. The expected return does not change much as portfolio size grows. In a power law world the mean is dominated by the rarest winners, so both small and large portfolios have similar expected values.

Second. The median outcome rises sharply with portfolio size. With 50 positions the typical fund does fine. With 300 to 500 positions the typical fund does significantly better. More shots make strong performance normal rather than lucky.

Third. The probability of capturing extreme winners increases non linearly. Under these assumptions:

A 50 position fund has a small chance of a 1,000x outcome and an even smaller chance of a 10,000x
A 300 position fund has better than fifty percent odds of a 1,000x and meaningful odds of a 10,000x
A 500 position fund makes those outcomes even more likely

Portfolio size becomes a lever that determines how much of your fund’s success depends on luck. Once you reach the 300 to 500 range you are sampling the right tail densely enough that extreme winners stop being rare miracles and start being statistically expected.

4. Assumptions about deal quality

None of this assumes that deal quality stays constant if a fund scales recklessly. The model isolates the effect of portfolio size while holding selection ability constant. The only reason this portfolio size is feasible for us is that we review more than ten thousand companies per year and use AI screening to ensure that partners focus on the highest potential opportunities.

The conclusion from the math is simple. In a power law world, the optimal number of shots on goal is far larger than most early stage funds take. If your sourcing engine and filtering discipline are strong enough, a 300 to 500 company portfolio is not a spray and pray strategy. It is the statistically rational way to capture the winners that define the asset class.

Appendix B: The Actual Model and Equations

This section provides the precise mathematical structure behind the simulations described in Appendix A. It shows exactly how portfolio outcomes were generated, what we assumed, and what we did not assume.

1. Defining the underlying return distribution

We represent each startup investment as a random variable XXX drawn from a discrete power-law-shaped distribution. The distribution is defined by outcome buckets and their probabilities:

P(X=mi)=piP(X = m_i) = p_iP(X=mi)=pi

Where the mim_imi values are multiples on invested capital and the pip_ipi values are the empirical probabilities:

Multiple mim_imiProbability pip_ipi0x0.501–3x0.253–10x0.1510–50x0.0750–200x0.02200–1,000x0.0071,000–10,000x0.00210,000–100,000x0.001

To simulate, each non-point bucket is sampled from a log-uniform distribution inside that band (reflecting that outcomes inside a bucket themselves follow a heavy tail):

X∣X∈[a,b] ∼ exp⁡(U(ln⁡a,ln⁡b))X \mid X \in [a,b] \; \sim \; \exp\left( U(\ln a, \ln b) \right)X∣X∈[a,b]∼exp(U(lna,lnb))

This avoids the unrealistic assumption that a “10–50x” outcome always returns 30x.

2. Constructing a fund

A simulated portfolio of size NNN is:

Portfolio Returns=1N∑i=1NXi\text{Portfolio Returns} = \frac{1}{N} \sum_{i=1}^{N} X_iPortfolio Returns=N1i=1∑NXi

Where each XiX_iXi is an independent draw from the distribution above.

We ran KKK Monte Carlo trials (typically K=10,000K = 10{,}000K=10,000) for each portfolio size.

Across the KKK trials we record:

Mean portfolio MOIC
Median portfolio MOIC
5th, 25th, 75th, 95th percentiles
Tail capture probability

P(max⁡(X1,…,XN)≥T)P\left(\max(X_1,\dots,X_N) \geq T\right)P(max(X1,…,XN)≥T)

for thresholds T=50x,100x,1,000x,10,000xT = 50x, 100x, 1{,}000x, 10{,}000xT=50x,100x,1,000x,10,000x

This last expression is the mathematical core of “shots on goal.”

3. Tail capture probability rises non-linearly

A key analytic quantity is the probability that at least one investment in the portfolio clears a high multiple TTT:

P(hit T)=1−(1−pT)NP(\text{hit } T) = 1 - (1 - p_T)^NP(hit T)=1−(1−pT)N

Where pTp_TpT is the probability that an individual startup exceeds threshold TTT.

For example, if

p1,000x=0.002,p_{1{,}000x} = 0.002,p1,000x=0.002,

then for a 50-deal portfolio:

P(≥1 company at 1,000x)=1−(1−0.002)50≈1−0.904≈9.6%.P(\text{≥1 company at } 1{,}000x) = 1 - (1 - 0.002)^{50} \approx 1 - 0.904 \approx 9.6\%.P(≥1 company at 1,000x)=1−(1−0.002)50≈1−0.904≈ 9.6%.

For a 300-deal portfolio:

1−(1−0.002)300≈1−0.548≈45.2%.1 - (1 - 0.002)^{300} \approx 1 - 0.548 \approx 45.2\%.1−(1−0.002)300≈1−0.548≈ 45.2%.

For a 500-deal portfolio:

1−(1−0.002)500≈1−0.367≈63.3%.1 - (1 - 0.002)^{500} \approx 1 - 0.367 \approx 63.3\%.1−(1−0.002)500≈1−0.367≈ 63.3%.

This is the mathematical basis for the core claim: in a power-law world, increasing N increases the odds of capturing ultra-rare outcomes far more than it changes the expected value.

It is also why the median return improves meaningfully with more positions even though the mean stays almost constant.

4. What we did not assume (and why)

The model intentionally isolates the impact of portfolio size. It does not assume:

Equal access to all deals
Equal ability to get into top performers
Uniform value add or founder selection behavior
Zero degradation in selectivity as volume increases

Those are real questions, but they are orthogonal to the specific mathematical question we were testing:

Holding deal quality constant, how does the distribution of fund outcomes change as you vary the number of draws from a power-law distribution?

This is the same approach used in academic power-law modeling, Rebel Fund’s YC Monte Carlo studies, and Steve Crossan’s “rational VC portfolio size” work.

5. How we map the math to reality

To connect the model to Team Ignite’s operating reality:

We see ~10,000 companies per year, and our AI ranking system restricts partner review to the top 1–2 percent of the funnel.
We do not invest in the “top 1 percent by outcome,” we invest in the “top 1 percent by observed signal.”
Access effects are real, but our acceptance rates in competitive rounds (>98 percent for YC) keep the “selection ability” assumption reasonable for modeling.
Support is not uniform across 300–500 companies; instead we concentrate follow-on capital and attention on breakout signals, which the model reflects indirectly through dilution-adjusted MOIC.

The model does not claim that the 300th company in our portfolio is as good as the 30th. It claims that if your sourcing and signal engine keeps average deal quality above a baseline, the mathematics of power-law returns reward more samples.

6. Full algorithm used in the simulation (readable form)

For each portfolio size N in [50, 100, 200, 300, 400, 500, 1000]: For trial = 1 to K: returns = [] For i = 1 to N: Draw bucket j from categorical distribution p_j Draw X from log-uniform(min_j, max_j) for that bucket Append X to returns portfolio_MOIC[trial] = mean(returns) portfolio_max[trial] = max(returns) Record: mean(portfolio_MOIC) median(portfolio_MOIC) percentiles(portfolio_MOIC) P(portfolio_max ≥ threshold)

This is the actual model.

Appendix C - FAQs

What Is the Difference Between Mean and Median, and Why Does It Matter in Venture?

Why do we look at both mean and median in this model?

Because in a power law world they tell completely different stories.

The mean is the average. It is heavily influenced by the rarest and largest outcomes. If one company in a fund returns ten thousand times capital, the mean jumps even if the rest of the portfolio is mediocre.

The median is the middle. Half of all simulated funds perform better than the median and half perform worse. The median tells you what a typical fund looks like, not the theoretical average.

Why does the mean not change much with portfolio size?

In a power law distribution the extreme winners dominate the average so completely that the expected value barely moves when you add more positions. Whether you draw fifty times or five hundred times from a distribution that contains ten thousand times outcomes, the average is anchored by those long tail possibilities.

This is why mean return is not a helpful guide for how many investments a fund should make.

Why does the median rise sharply as the portfolio gets larger?

Because larger portfolios sample the right tail more consistently. With a small portfolio you only occasionally catch a sixty times or one hundred times winner and you almost never catch a thousand times winner. With a large portfolio you are pulling enough times from the distribution that the probability of hitting the meaningful winners becomes much higher.

This makes the typical result of a three hundred or five hundred position fund much stronger than the typical result of a fifty position fund.

What does this mean in real life for a fund manager?

It means that fund size is a lever that controls how dependent you are on luck.

A small portfolio can hit a mega outcome but the odds are low. A large portfolio can still miss but the odds are materially lower. Once you reach the three hundred to five hundred range the portfolio is dense enough that one or more right tail outcomes are not unusual. They become expected features of the distribution.

In practical terms this means that a fund constructed around enough shots on goal can deliver strong outcomes more consistently across vintages. LPs care deeply about this because they prefer repeatability to one off heroics.

What does this mean for founders?

It means the fund is set up to support a wide funnel while still recognizing that a few companies will drive the majority of value creation. The math does not reduce the importance of conviction. It simply makes it more likely that the fund is positioned to find and support the companies that break out.

Why does this matter so much in venture but not in most other asset classes?

Because most asset classes follow distributions without extreme outliers. In public equities or real estate the mean and median are usually close. In early stage venture they are often miles apart because the biggest winners are so large relative to everything else.

Understanding the gap between mean and median is the key to understanding why traditional portfolio sizes are too small for the shape of this market.

Sources, Notes, and Further Reading

A few readers asked where the underlying probabilities in the model come from and how generalizable they are across sectors and geographies. The core power law dynamics are extremely well documented. Here are the strongest public sources.

AngelList Power Law Data
AngelList published one of the clearest analyses showing that early stage startup outcomes follow a steep power law, with a small number of companies generating the vast majority of total value.
https://www.angellist.com/blog/what-angellist-data-says-about-power-law-returns-in-venture-capital

Correlation Ventures and Horsley Bridge Findings
Across twenty-plus years of actual VC portfolio data, the same pattern appears:
• 50 to 60 percent of startups go to zero
• Roughly a quarter return low single-digit multiples
• A very small percentage drives nearly all gains
This underpins the left side and the center of the distribution we use.

Hustle Fund’s Power Law Explanation
A concise summary of why early stage investing structurally produces a tiny number of massive winners and a long tail of failures.
https://www.hustlefund.vc/post/power-law-in-startup-investing

Skalata VC’s Return Curve
A helpful public breakdown of how venture returns concentrate heavily in a few companies and why patience matters in the long tail.
https://www.skalata.vc/blog/how-venture-returns-really-work-power-law-patience-and-portfolio-construction

YC-Specific Data
YC is one of the steepest power law ecosystems in the world. Multiple analyses from PitchBook and Rebel Fund show:
• About 6 percent of YC companies become unicorns
• About 0.6 percent become decacorns
• A handful of companies generate most of the cumulative valuation
Airbnb and Stripe alone account for more than 25 percent of YC’s historic value creation.
These curves are unusually steep compared to most markets, which is why YC-style pipelines tend to produce fatter right tails than the average sector.

A Note on Timing and IRR

The model in this article focuses on the shape of the outcome distribution, not the timing of liquidity. One LP pointed out correctly that 2x to 4x outcomes often return capital in two to six years, while thousand-times outcomes may take a decade or more. That timing difference affects IRR and DPI but does not change the underlying portfolio construction logic.

A broader portfolio tends to produce:
• Early DPI from modest exits or tenders
• Long-term TVPI durability from the right tail that takes years to mature

This time separation is a feature of the strategy, not a flaw. High-volume portfolios return capital earlier while preserving exposure to the long-duration winners.

More Shots on Goal in a Power Law World