My investment method (part 4)

Here are parts 1, 2, and 3.

When people talk about valuation, they often drift toward extremes.

I’ve seen investors treat valuation with meticulous attention to decimal places, with assumptions so elaborate that the model itself starts to eat up all rational arguments. I’ve seen others treat valuation as something to be honored in theory but completely ignored in practice. But probably most often, I’ve seen people treat valuation as a plug variable, something to be massaged until the output aligns with whatever enthusiasm they already had for the asset.

You’ll hear people say that it’s all right to pay a fair price for a great business. My advice is: don’t listen to them. A good investor never pays a fair price for anything. There are two sides to every trade, and there’s no prize for being fair. There’s only the willingness to act when others won’t. A fair price is what you pay when you can’t decide whether you’re the one exploiting an opportunity or the one being exploited. And a fair price for a great business means you give what you get, and that isn’t a particularly compelling way to compound wealth.

Investors aren’t in the same boat together. This is the whole point of Ben Graham’s metaphor about Mr. Market.

The practical problem about valuation isn’t deciding whether valuation matters in an investment, because in the end, it’s the only variable that counts. The problem is learning how to approach it in the real world, where numbers slip and shift every few months — sometimes every few weeks — and where compound interest and its runway can make a huge difference in the intrinsic value you estimate. In other words, a valuation can never be a destination for an asset’s true intrinsic worth, because all valuations are biased, and the future will never play out as imagined.

As new information surfaces, you’ll have to adjust and adapt your valuation estimate. This is called Bayesian updating. And as information changes, remember that it’s not just your value that’s changing, but so will the price, and price will change more violently than value. The question isn’t whether you’re wrong, but whether your wrongness is survivable and whether your rightness is rewarded enough to matter.

So here’s what valuation is: it’s a way to make your assumptions explicit, to see what must be true for your investment to work, and to understand how much room you have for error.

That’s where simplicity helps. A valuation with thirty inputs may feel more thoughtful, but your understanding of it tends to be inversely proportional to the number of inputs you allow yourself to juggle. Valuation is conceptually simple, but it’s human nature to insist on making it convoluted, either to intimidate outsiders or to confer an illusion of insight to yourself. A simple approach to valuation with transparent assumptions will always serve you best.

Here are three other important principles when it comes to valuation:

• Quality growth is always rarer to come across than you think. Quality growth requires that a firm find places to continuously reinvest capital at returns meaningfully above its cost of capital. A firm can grow while destroying value at the same time, and the larger it gets, the harder finding new avenues becomes. If you project growth rates higher than or in line with the recent past, you better have a damn good reason.
• Don’t mistake accounting for valuation. Accounting is a descriptive tool built on conventions, but the real economics of a business hide beneath those conventions. D&A schedules rarely map reality, revenue recognition can distort timing, and capitalization policies vary. In other words, GAAP profits can diverge sharply from owner earnings and FCF. Owner earnings and FCF are what matter in a valuation.
• Make your valuation explicit, but don’t rely on precision. You don’t need to be correct — because you won’t be — but you should be clear about your stance and your process to the degree that someone else could, in theory, replicate what builds up to your value estimate and identify where they disagree. (I’m not talking about publicizing your valuations, only that your own framework should be laid out clearly enough that you can return to it months later without rewriting history in your mind.)
• All you really need to know is whether the asset you’re buying is good enough and mispriced enough for you to make an informed investment decision. Ben Graham once said, “You don’t have to know a man’s exact weight to know that he’s fat.” Two informed analysts will always produce different valuations. If a company trades at less than 2/3rds of a conservative valuation, the valuation doesn’t need to be precise but reasonable enough to show an obvious bargain. What matters, and what I’ve emphasized in previous parts of this series, is that the real challenge in investing isn’t analytical difficulty, but overcoming psychological discomfort. Because the stock market is a pari-mutuel system, great investment opportunities should, by definition, look unappealing to someone else and will require patience and tolerance for neglect.

The best way to learn valuation is to practice it every day. Make it your intent to value assets, companies, and investments, not just talking about valuing them. Valuation is a universal discipline across all public stocks, private businesses, and speculative assets. Through practice and experience, you’ll learn that a dollar earned in one jurisdiction carries different risks than a dollar earned in another. You’ll learn that two businesses can produce the same accounting earnings yet have wildly different intrinsic values. You’ll learn to account for truncation risk, because some businesses won’t survive long enough to justify the projections embedded in a going concern valuation, and that this risk can’t be buried in a discount rate. (In a DCF, this going concern assumption mostly finds its place in the terminal value, and in relative valuation, it shows up implicitly when valuing a company based on how other companies — most of which are healthy — are priced by the market at a given time.)

Speaking of which, discount rates — or the cost of capital — are a touchy and potentially dangerous subject among investors in different camps. Some treat the discount rate as the precise output of a theoretical model like CAPM. Others grasp the concept but decide to abandon it altogether and use a uniform rate for everything. Charlie Munger once called the cost of capital a “perfectly amazing mental malfunction.”

I’ll tell you what the cost of capital isn’t. Contrary to popular opinion, it’s not just the opportunity cost of putting money into one asset instead of something else. Of course, all capital has an opportunity cost. But it depends on what that “something else” is, because two different assets can have very different levels and types of risk. And people encounter (or stumble into) different opportunity sets, so the cost of capital will always mean different things to different people.

So I’m not one of those value investors who debunk modern portfolio theory. “Risk-adjusted cost of capital” can become an abstract concept in theory, but it’s obviously a directionally right one. You don’t have to juggle with betas to believe in risk-adjusted returns. It’s just a matter of common sense. Of course a risk-free asset like a government bond has lower cost of capital than a corporate bond. Of course the lower you are in the capital stack, the higher a return you should demand. And of course a company operating in a currency issued by a country with meaningful default risk has a higher cost of capital than one operating in a stable currency with little national debt. (I can count on two hands the times I’ve seen someone pitch a stock while conveniently leaving out the fact, or not even realizing, that the company earns all of its cash flows in a country where the 10-year sovereign yield sits in the double digits, yet every peer they compare it to operates in an economy with a low-single-digit 10-year.)

When people have issues with the concept of the cost of capital, what they really mean is that they have issues with the cost of equity, not the cost of debt or other debt-like financing. They’re talking about how to estimate the required rate of return on the equity. The problem with the cost of equity is that, because it’s essentially guesswork, it can too easily become an escape hatch to justify almost any investment case. You see it with “quality compounders” being bid up to 50x earnings because investors treat them as perpetual bonds. But what investors are really doing is finding a stock everyone else likes, and then smuggling that popularity into the discount rate. They’ll say that a Costco paying out <1% of earnings while growing in the mid-single digits could be just as cheap as another company growing >10% per year while paying out 5% in dividends and buybacks. This might in theory be true, but it’s potentially very dangerous.

Charlie Munger also said this (which will ring clear if you’ve looked at enough acquisitive businesses): “A corporation’s cost of capital is 1/4 of 1% below the return on capital of any deal the CEO wants to do.”

The rule to follow is simple: if your investment case depends too much on your discount rate, you’re spending too much time on the wrong investment case. The solution is to move on and find something more obvious. There are >55k listed stocks out there. You’ll want a wide margin between your required rate of return and the prospective return before you even consider investing.

(Btw, there’ll be some who read this and point out that Buffett has said for years that the US Treasury rate is the rate he used to discount future cash flows. But what he has really been saying is that he used certainty-equivalent cash flows. Instead of adjusting the discount rate upward to account for risk, he adjusted the cash flows downward to the degree he could count on with his highest degree of confidence. Once you reduce the cash flows to a level of near-certainty, the treasury rate becomes a natural yardstick. And that’s what it is: a yardstick. Buffett didn’t sit around actually discounting cash flows. He evaluated opportunities in his head and compared them by the quality and durability of the cash they could deliver after he’d mentally stripped out anything uncertain. In other words, Buffett wasn’t pretending that every business is safe enough as an alternative to a government bond. He was choosing businesses where, after adjusting the cash flows for risk, the remaining stream behaved enough like a government bond that the comparison became meaningful. And from there, it’s a matter of judging which certainty-equivalent stream offers the better long-term return.)

This brings us to the margin of safety, a pillar of my investment method. “Margin of safety” is a phrase that has been repeated so often that it tends to get lost in the weeds, acknowledged with a nod, but then abandoned in exchange for the shiny new toy that everyone else cares about.

“Yes, I know about Ben Graham and the margin of safety. Have you seen this B2B SaaS business growing 25% a year with only 3x turns of debt, trading 20% below peers? That discount should be enough of a margin of safety when you’ve got a rocket ship like this.”

Where investors tend to miss the meaning behind the margin of safety is when they use it as a poetic way to tell themselves they bought something cheap. But the margin of safety isn’t an abstract comfort cushion. And it isn’t just synonymous with high expected returns. A margin of safety is about survivability. If I have to choose between the investment that protects me when I’m wrong and the one that rewards me when I’m right, I’ll choose the former every time. Paradoxically, it’s usually the former that ends up generating the most reward over time.

So investing with a margin of safety is about focusing intensely on the downside. You don’t want your investment mistakes to originate from paying too high a price relative to some measure of earnings. You don’t want to look back at any investment that went south and say, “If only I’d have bought this at x times earnings instead, then it would have worked out fine.”

Your errors — which are inevitable — should originate from something unusual happening. Something that blindsided you that you probably knew was a risk but a low-probability risk. Obviously, if you understood the future of a business perfectly, you’d need very little margin of safety. But perfect information doesn’t exist when it comes to the future. There are only confidence intervals.

The answer lies in having layers of redundancy in the price you pay for a stock. Redundancy is ambiguous because it seems like waste if nothing unusual happens. But in the stock market, something unusual happens all the time. And in the moment an investment stops cooperating with your thesis, redundancy becomes the only thing standing between a temporary setback and permanent impairment.

Here are some examples of the redundancies I’m talking about:

• Paying a price that assumes muted growth, conservative margins, and unexciting capital allocation, so the business doesn’t need to surprise you on the upside for the investment to work.
• Buying companies with balance sheets that can absorb a few bad years without forcing dilutive equity issuance or distressed refinancing.
• Insisting on economics resilient enough to handle periods of industry stagnation or macro shocks, not just one good year extrapolated forward.
• Owning businesses with management teams who’ve historically behaved rationally with excess cash rather than rolling it into acquisitions that destroy value.
• Choosing industries where demand may wobble but rarely collapses, allowing you to survive the full cycle rather than hope you happened to buy at the right point in it.
• Preferably paying a price that’s supported, or is at least partially protected, by the company’s liquidation value or tangible asset base, so the downside is anchored in something other than a forward earnings projection.

With this concept firmly in mind, you understand that share price volatility isn’t what you should be afraid of or even care about. Share price volatility, and the volatility in relation to the market, is a detached phenomenon that can occur for a variety of reasons outside a company’s fundamentals, such as liquidity droughts, index rebalancing, margin calls, risk-parity unwinds, ETF flows, or just someone on the other side needing cash more urgently than you do. This makes volatility as much an opportunity as a threat, because volatility measures upward variation just as much as downward variation. Stable results should mean little to you if you can optimize the endpoint. So rather than fixate on share price volatility, your time and effort are better spent understanding how the company’s behavior can make the future more predictable and reduce the true risks of your investment. Real risk is the risk that the balance sheet buckles under stress, the risk that management does something irreversible with your money, or the risk that you misanalyze the business. Some of the safest investments I’ve ever made — and you’ll ever make — have been when my purchase was done at the peak of downside volatility.

Now onto how to actually value a business.

There are two avenues to go down here, and both are — or should be — interrelated: valuation and appraisal. They sound similar, and people use them interchangeably (mostly by calling it “valuation” when what they’re really doing is an appraisal), and while the two approaches should, in theory, converge on the same value, they’re not the same thing.

A valuation — which, yes, is usually done by discounting all future cash flows back to present value — tells you whether the business, on its own terms, makes sense to own. Appraisals, on the other hand, are what investors typically rely on to justify an investment (and convince others of the same) because investors are social beings. We like to anchor our reasoning in things that are both based on other people’s opinions and immediately visible. We like to point to how other companies are priced, what multiples the industry trades at, and what the last five deals in the space looked like. We like to show that the market has already blessed our idea by giving similar businesses similar, or higher, “valuations.”

Most valuations you read out there are appraisals masquerading as discounted cash flow analyses, where the analyst uses the DCF to back into a number that was already implied by the peer multiple they had in mind.

And it makes sense why that happens. A DCF leaves the hardest part of the job to you. It asks you to project future cash flows into perpetuity. And anyone who’s done a DCF on a growing company will have noticed that most of the value produced sits far out in the terminal years, often well beyond the boundary where anyone can forecast with any degree of confidence. A one-point change in the terminal growth rate or the cost of capital can swing your valuation by 20-30% without you changing anything about the underlying business.

To understand why that is, you can convert the terminal value in a DCF into a cash flow multiple, and that multiple is simply one over the difference between the cost of capital and the terminal growth rate. If the terminal growth rate is 4% and the cost of capital 8%, the terminal multiple will be 25x the out-year cash flow (1 / [8% – 4%]). Just turning the cost of capital up to 9% and the growth rate down to 3% yields a 16x multiple instead. And as we talked about previously, the further you reduce the cost of capital, the more exponential the “justified” multiple becomes. A 6% cost of capital with a 4% growth rate would yield a 50x multiple on the cash flows. You’ll only experience a few opportunities in your lifetime where you can forecast cash flows with that overwhelming degree of confidence, and even then, you’ll struggle to justify the math.

And then you’ve got the other problem we already talked about, which is truncation risk, or the risk that the business won’t live long enough for most of these projected cash flows to materialize. That risk can’t be folded into a discount rate, because the discount rate is, according to Damodaran, a “blunt instrument that was never intended to include failure risk.” A DCF assumes a going concern into perpetuity, which many businesses will never achieve. This is why, when you’re dealing with a business in distress — or one with a meaningful probability of entering distress — the DCF becomes almost useless in its standard form. You can raise the discount rate, change the margins, and tweak the runway, but none of it will capture the binary nature of failure. Instead, you have to incorporate a probability of survival for each year of projected cash flows, haircut those cash flows accordingly, and then estimate whatever value might still be scraped from the ashes.

These are the reasons why the appraisal approach appeals to most investors. They want to play the pricing game without having to walk through the discipline of stating what the future cash flows actually are. When the future is murky, people fall back on what feels observable: “This trades at half of where peers trade,” or “This is below book value,” or “This is priced like it’s going bankrupt even though management says liquidity is fine.” It’s valuation by proximity. If a group of companies trades at 15x EBITDA, 10x feels cheaper. If the same group trades at 30x, then 22x feels “conservative.” Appraisal is a shortcut, a way of saying you think the business is worth more, without having to specify why or under what assumptions, providing the comfort of valuation without the commitment of valuation. But skipping the hard part of the job doesn’t remove the hard part. It merely hides it.

Appraisals are how people perform “relative valuations.” Instead of grappling with the economics of the business, investors look for reassurance in what others have paid either for the asset itself in the past or for its peers in the present. But what that type of appraisal is doing is borrowing conviction from the crowd, and that’s why it’s bound to be dangerous on its own. When you rely too heavily on appraisal, you implicitly assume that the market itself is roughly correct, just not for your precious snowflake. And even if that assumption held, peer comparisons are never as clean as they appear. Two companies might share an industry label and nothing else. Their capital allocation policies can diverge in ways that meaningfully change intrinsic value. One firm might have a hidden non-operating asset that could be spun off to reduce leverage or release cash without disturbing the core business, while another might be encumbered by contracts or minority interests that trap capital rather than free it. The keyword here is resource conversion option, and such differences never show up in the tidy rows of a comp table unless you deliberately adjust for them. Rarely do investors do that. They line the companies up, look at the mean or median, and call the comparison complete.

All of what we’ve discussed so far should make it clear why valuation is an art, not a science. The art lies in knowing which tool and which lens to apply at which moment. A DCF or contingent-claim valuation forces you to articulate the economics and think like an owner, while an appraisal forces you to acknowledge the environment surrounding the stock. A pure valuator who refuses to consider appraisal will typically underestimate how long mispricing can persist, while an appraiser who refuses to make their assumptions explicit through a valuation will typically underestimate how violently mispricing eventually unwinds.

What I do is treat appraisal as a pricing lens that I only use for 1) reconciling against a valuation with explicitly stated assumptions, 2) producing a rough shortcut during the filtering stage, and 3) referencing transaction multiples only when dealing with a special situation where control value is directly at stake.

As for (1), this is straightforward. If you have your valuation and forecast on hand, all you have to do is divide your estimated value of the operating assets (or the equity value, depending on what you’re analyzing) into whatever measure of earnings or sales you want for the trailing twelve months, and that gives you your justified multiple. Do the same using next year’s forecast, and you have your justified forward multiple. So if your valuation, built on conservative assumptions, suggests the company is worth 12-14x EBIT, you shouldn’t be surprised if the market prices the closest comps at 10-15x. But if the market is pricing them at 20-25x, you now have a clear point of disagreement. This approach is an important nuance, because a pure appraisal typically starts with the market multiple and lets that anchor the entire story.

As for (2), using appraisal as a shortcut in the filtering stage is simply a practical concession to the fact that you can’t model every company on earth. You want a quick appraisal to narrow your universe but not dictate your conclusion. So you need some sort of heuristic and experience to approach the task as a handicapper, not just someone doing relative valuation.

Here’s what I focus on:

• The balance sheet. I look at its fungibility and safety, and I calculate NCAV and tangible book value to compare against the price.
• EV/sales and normalized margins. Like in screens, I move up the income statement to get the cleanest measure. I then attempt to gauge the normalized operating margin, and if it’s not a deteriorating business, I roughly assume the enterprise value will eventually converge to 10x normalized EBIT. In practice, this means the EV/sales multiple becomes the normalized margin with the dot moved one decimal point, so a 15% normalized margin becomes a justified 1.5x sales multiple at some point. So if you’ve got a 15% margin business trading at 6x sales, you know the market already prices in a lot of growth or margin expansion at the current price. This little exercise will make you realize how it’s usually a good idea to throw anything trading at >10x sales into the too-hard pile.
• Free cash flow, but calculated the right way. I don’t rely on data providers to do this job for me because FCF is not a standardized figure, but a constructed number, and its meaning depends on what you subtract, what you count (think lease payments and share-based comp), and whose cash you’re trying to measure. So I make a quick calculation for the current FCF and what I believe a normalized level would be if you strip off volatile items like changes in working capital. The FCF yield then sets my starting yield, which I adjust up by whatever sustainable level of growth I think the business can deliver to get to a prospective very-long term return. If that prospective return isn’t comfortably above the cost of capital, I might move on, though not always. Some stocks will have no free cash flow in any given year, which is fine, as long as I have a high degree of confidence in gauging the cash flows in the future.

When I get to the later stage of diving deeper into a stock, I calculate what I consider the three valuation layers. And as I wrote in part 3, for each of these layers, I keep my focus on enterprise value. I look at the whole business holistically and let the cap stack determine the level of risk I see in the equity slice.

The layers are:

1. Liquidation value. This is the most conservative layer, asking what would be recovered by the owner if operations were shut down. I’ve written a longer post on how I do this here. In almost all cases, liquidation value isn’t the highest use for a business, because even bad businesses are usually worth more than they can be immediately liquidated for. But liquidation value can act as a redundancy and put a firm floor on the downside.
2. Reproduction value. This layer is about understanding whether the market has underpriced something that would be expensive, time-consuming, or structurally impractical to rebuild from scratch. I’m asking, “How much would a competitor or new entrant have to invest to get to the level of this business?” This means I go over each balance sheet item and attempt to restate each to its probable true value while ignoring the goodwill. When reproduction value exceeds market value, you have an additional layer of redundancy. Not only are the assets underpriced, but any rational competitor would be discouraged from entering the space at current prices (and for an entrant to acquire an existing company would mean paying a control premium that would either partly of fully erase the discount to reproduction value). This one is a nice example. In many cases, I’ll just use tangible book value for a quick proxy, and dependent on the business, I might capitalize R&D spend and brand advertising for the number of years that asset would likely need to be amortized (that is, my guess at the asset’s useful life). For some businesses, intangible investments matter hugely and can give you a sense of the moat. I wrote about it in this post using Coca-Cola as an example.
3. Earnings power value. Earnings power value is what really matters for the vast majority of businesses, so this is, of course, the part of the valuation that relies most on your understanding of the economics, reinvestment opportunity, durability of the margin, competitive positioning, and long-term demand. As previous writeups have shown, I don’t shy away from DCFs (at all), but I stay aware of their mathematical fragility. Any model that concentrates so much weight in the distant future should be handled with suspicion. You’ll only find a couple of handful opportunities in your lifetime where you’ll be able to project economics far into the future.

Because I write up a lot of special situations in the newsletter, some people tend to think I always look for catalysts. That’s not true. I don’t actively look for catalysts. I look for mispricings and some form of business momentum. Catalysts simply make the possibility of mispricings resolving faster. The paradox is that the absence of a catalyst sometimes makes a stock even more mispriced. And if it’s mispriced enough, you’ll often find that a catalyst will show up around the corner to force the crystallization anyway when you least expect it. The market is partly efficient after all, and certain market participants like strategic buyers do recognize value if it’s unmistakable. “Cheap” ends up the catalyst by itself, because the control market — as opposed to the OPMI, or the market for outside passive minority investors — eventually arbitrages those gaps away.

Of course, there are companies out there that’ll probably never be for sale. In those cases, the control market won’t bail you out, and the value you ultimately realize will be amplified, for better or worse, by the capital-allocation skill of the people running the business. But that doesn’t change the underlying principle: if you find yourself always needing a catalyst for an idea to work, it’s usually a signal that the underlying business isn’t good enough, the valuation isn’t attractive enough, or your conviction isn’t strong enough to let time do the compounding for you. A great investment doesn’t require a scheduled event to unlock value when the economics should do that for you. Catalysts are coveted because investors dislike not having an “exit strategy.” But a true long-term investor should be someone who doesn’t need an exit strategy, and a true value investor should be willing to accept a very low price in place of a catalyst.

This is important to note because, as you know if you’re not new here, my investments tend to fall into two distinct buckets:

1. Special situations (those with a catalyst)
2. Generals (those without a catalyst)

The point is that the presence of an event never substitutes for the presence of a mispricing. I won’t buy a special sit unless it’s mispriced enough to stand on its own without the event, so that if the catalyst fails to materialize, the investment still has a shot (even if a small one) at working on its own merit. Event-driven investments, by themselves, usually don’t offer enough profit potential to justify taking any significant form of risk.

An event could be any sort of resource conversion activity: a spin-off, a divestiture, a recap, a refinancing, a liquidation, a settlement, a merger, a go-private, a tender offer, a court filing, a regulatory approval, a forced asset sale, a covenant breach — anything that changes who controls the assets and what they can do with them. The key thing all special sits share is that value is being unlocked by action, not by operating results alone. The reason they’re a lucrative pursuit is that they’re usually easy to frame in a risk/reward proposition. Special sits tend to have boundaries. You get a good feel for the predictability of the investment case, and what is usually a wide probability distribution of outcomes becomes a decision tree with narrower, discrete branches.

The generals I invest in, on the other hand, are usually more mispriced than the special sits, not because they have an event attached to them, but because they don’t. Generals don’t need anything to “happen.” They’re the kinds of stocks where people say, “This stock will never move,” until one day it does. They work because the underlying economics are strong enough, the valuation conservative enough, and the margin of safety wide enough that time alone will do the heavy lifting. So it goes without saying that in generals I stick strictly to businesses with satisfactory returns on unlevered tangible capital. Because the timeframe/holding period could stretch, I need the comfort of knowing that every year the business continues to add to intrinsic value. Over the very long term, it’s hard for a stock to earn a much better return than the business underlying it earns.

In a neutral market — that is, a market where prices are neither euphoric nor depressed — my portfolio might end up in a 40–60 split between special sits and generals, respectively. But the proportions swing dramatically depending on the opportunity set. When the market is handing out high-quality businesses at silly prices, I naturally gravitate toward generals. In frothy markets, I tend to find more to do in special sits, which offer attractive absolute returns without moving to the beat of the market.

And the opportunity set very much determines my position sizing too. In special sits, you need a more constant flow of ideas, where when one closes, the proceeds must be reinvested. The position sizing here tends to be smaller because even if the downside is protected, the path is less smooth. You can underwrite the value, but you can’t to the same degree underwrite the mechanics with the same degree of confidence you can underwrite a long-term, solid business.

Generals are different. When the mispricing is severe, the margin of safety solid, and both the economics and people running the business trustworthy, you’ve got time on your side rather than against you, and it makes sense to bet more concentrically, sometimes extremely concentrated. A nice example is Fairfax Financial, which in 2024 had grown to >70% of my personal portfolio. It was a no-lose investment at a very opportune time (August 2020) when the downside was anchored, the people were exceptional, and the gap to value was enormous.

At a certain point, position sizing becomes a question of how much you should bet when the odds are skewed so far in your favor. This is where the Kelly criterion becomes a useful mental model, not as something to follow mechanically (which would be reckless in the stock market when truly great decisions are spaced far in time), but as a way to think proportionally. Kelly’s basic insight is simple: when the expected value is high and the probability of permanent loss is low — which determine your odds and edge — your position size should rise. And the right position size for a genuinely superior proposition will almost always surprise you to the upside. So I don’t “run” Kelly as a formula, but use it as a directional compass.

If a full-Kelly bet tells me to bet 40% of my capital, I know that 1) the expected value is unusually favorable, and 2) I should probably size well below that in practice, because the real world contains correlation, liquidity risk, estimation error, and my own psychological limits.

This also clarifies why special sits rarely justify huge position sizes. Even with a well-defined event path, the payoff is capped, the timeline constrained, and the branches of the decision tree contain more embedded process risk. The “edge” is real, but it doesn’t scale. Generals, when they’re truly mispriced, do scale, because the upside isn’t bounded by an event but by the economics of the business and the patience of the owner.

It’s for this reason that readers know — as I’ve said before — that I buy just a fraction of the stocks I write up on this newsletter (and some that aren’t written up), because I’m a concentrated investor. My portfolio rarely holds >10 stocks and usually <7 (which, of course, is dependent on the opportunity set as there are periods with an abundance of special sits, and there are situations where it might make sense to buy a basket, like foreign net nets). Any serious investor should allocate more to their best idea than their tenth, a recognition that the distribution of opportunities in markets is wildly uneven and that great ideas are few and far between.

If you’re a concentrated investor behaving like an owner rather than a trader, portfolio activity will automatically become de minimis. You can do the math: If your typical holding period is, say, three years, and you on average hold eight stocks, you’re turning over two or three stocks per year. And contrary to the hyperactive mythology surrounding markets, two or three good ideas per year is plenty of work to find.

And once you’ve built a position in a concentrated portfolio with long holding periods, doing nothing is often the correct action. If you find yourself trading constantly, it’s almost always a sign that you’re either failing to size your best ideas properly (if liquidity allows it, I usually buy my intended full position in one lump sum and never think more about it) or failing to buy businesses you’re willing to sit with for years. A portfolio built on mispricings that take time to converge is a portfolio that compounds not through frantic movement, but through stillness and equanimity.

Which brings me to when to sell and let go of a position. My discipline here is simple but, like almost everything else, not mechanical:

• I never sell because a stock has gone up. Price isn’t evidence of correctness, but evidence of changing market opinion. What matters is whether the gap between price and value has closed (or has closed enough).
• I sell when the thesis is broken, not when the share price is down. A stock that drops 30% right after I buy it doesn’t bother me. Cracks in the investment case that form after I’ve entered, even if they force me to recalibrate on the go and ultimately lose money, don’t bother me either. But misanalyzing the situation from the beginning bothers me a lot.
• I also sell when the opportunity cost becomes undeniable. A new idea must not be marginally superior but clearly superior to displace an existing holding, even if the existing holding is fine. It has never made sense to me to swap an idea I know intimately for something I don’t, just because it looks slightly better.
• In special sits, I sell either when the event has crystallized or the potential IRR is no longer attractive. If you’ve got a good special sit, the market tends to front-load your thesis into the share price, and your IRR often ends up higher even as the event has yet to materialize.

Here are two important points to end this part of the series:

1. You will always be wrong, which is okay, because valuing stocks is an art. I rarely know why a stock is cheap when I’m buying it, and I don’t waste much time trying to guess why others avoid it. My focus is simple: do I like the business, and what is it worth? Labels like “contrarian” don’t matter. The tricky part in investing is that you almost never approach an idea neutrally. Before diving in, you already know the multiples the stock trades at, and that primes you to see problems that “explain” the price. Once you’ve seen a cheap-looking price, you can’t undo that kind of bias, and your mind absorbs the market’s view before you’ve formed your own. The nature of valuation is that you’re basing a view of an uncertain future on incomplete information, so it’s natural to look for guidance in the wisdom of the crowds. But, as counterintuitive as it is, that instinct is the wrong process to apply to stock picking.
2. Focus on the process, not the outcome. Investing is a probabilistic game, and probabilistic games punish anyone who judges decisions by short-term results or low sample sizes. One of the most common errors in the stock market is confusing a good outcome with a good decision. A stock going up doesn’t validate your thesis any more than a stock going down invalidates it. The entire point is that you don’t need to be right often, you just need to be right well. It’s entirely possible to be very profitable over time even if the majority of your investments are losers. It’s the belief that you must always be right that leads you to lose discipline, rationalize sloppiness, and let emotion dictate your decisions so that you end up riding a bad idea for longer than you should.

Thanks for reading, and stay tuned for part 5 — the final of the series — which will cover how I train my brain every day so the whole process we’ve talked about in this series slowly compounds over time, and so information doesn’t go in one ear and out of the other. I hope you’ve enjoyed the series thus far.

Cordially,
Oliver Sung