Howard Marks on the Investing Mistakes That Never Go Away

In 2005, Howard Marks looked at the market's new favorites (real estate, private equity, hedge funds, crude oil) and had one reaction: "There they go again." The old favorites, tech and telecom, had wrecked portfolios only six years earlier. Almost nobody seemed to remember.

The Oaktree Capital co-founder's memo is a field guide to investing mistakes that return every cycle, each time with a fresh excuse. The excuses change. The mistakes don't. The one that should worry you most is the one you're certain doesn't apply to you.

This post sorts those mistakes into four families, tests them against Indian market data, and ends with a questionnaire you can run before your next purchase. What you'll get:

• Why smart investors repeat the same errors
• Four families of mistake, in plain language
• What 2008 did to the riskier corners of the Indian market
• What the October 2026 market looks like through the same lens
• Seven questions to ask before you buy

Why the Same Mistakes Keep Coming Back

Marks offers two reasons, and neither involves stupidity. First, few investors have lived through enough cycles to recognize a repeat. The last time this went wrong was twenty or forty years ago. Second, greed beats caution when times are good, and the people who preach the old lessons get dismissed as relics. The economist John Kenneth Galbraith made the same observation decades earlier.

I'd add a third reason. It's mine, not his: following the rules is expensive at exactly the moment it matters. In the last stretch of a boom, the cautious investor trails the reckless one. Trailing is visible, immediate and humiliating. The cost of recklessness arrives later, and humans discount delayed pain. That is why the rule book exists and so few people follow it.

Four Families of Mistake

Rather than march through a list, I find it more useful to group the errors by the faulty belief underneath. Each one swaps a mechanical fact for a comforting story. Table 1 is the map.

Table 1: Four families of mistake

Source: Author's synthesis of Howard Marks, “There They Go Again” (Oaktree Capital, May 2005), paraphrased and regrouped.

Family 1: The Rules Changed

Every boom needs a reason why valuation doesn't apply: new technology, new geography, new demand. Sometimes something real has changed. The mistake is concluding that the change cancels the link between price paid and return earned. Marks's core point here is circular: a strategy that works draws money, the money bids up the cost of entry, and that higher entry price leaves less return and more risk for everyone who arrives afterward.

Family 2: The Trend Is Destiny

Trends have ceilings. Buyers, sellers and competitors respond to whatever is working, and their response reshapes it. Investors who assume today's favorable conditions will persist, or that past returns forecast future returns, ignore regression to the mean: the biggest recent winners slow down, and the laggards catch up. Marks argues that the longest-surviving investors expect this, and read a long run of gains as a reason to be wary.

Regression cuts both ways. The Nifty 50 returned +36.95% over five years to 5 November 2026, about 6.5% a year, against a long-run price CAGR of roughly 11–13% (Equities India). A weak stretch isn't a verdict on the asset, any more than a strong one is.

Family 3: Price and Risk Don't Matter

Marks treats price as arithmetic. When more buyers chase a fixed supply, the price climbs, and the climb cuts what the next buyer can expect to earn. Add the myth that more risk buys more return (it only offers more promised return), the habit of borrowing on the assumption that assets will always out-earn the interest, and contempt for cash, and you have a recipe for leverage at the top.

Table 2 makes the price rule concrete with a hypothetical. Same business, same growth, same exit multiple. Only the entry price differs.

Table 2: The price paid decides the return

Source: Author's illustration. Hypothetical numbers, before tax and dividends.

Family 4: I'll Be Fine

The last family is the most human. Investors often sense a trend can't last but join it anyway, because the discomfort of being left out outweighs their doubts. Then comes the comforting assumption that they'll sell before the turn. Marks flags the question they skip: what tells you to leave before the crowd does, and who takes the other side when the crowd leaves too?

On 8 October 2026, 2,268 NSE stocks fell and 303 rose, roughly 7.5 to 1, and 72 Nifty 500 stocks closed at 52-week lows (Oquilia). A normal bad day, not a crash. But it shows how a crowded exit looks from inside: far more sellers than takers.

The Indian Scorecard: What 2008 Did to Risk

Marks wrote his memo in 2005. The Nifty's January 2008 peak (a 6,357 intraday high, per Business Standard) came about two to three years later. That gap matters. Anyone who heeded the warning in 2005 had to watch the market keep rising for years while being told they were wrong. Being early feels identical to being wrong, which is the same trap we saw in our analysis of the Nifty's recent 29-month flat stretch.

When the cycle turned, riskier segments did not reward the risk. Table 3 has the numbers.

Table 3: What 2008 did to the riskier segments

Source: Abakkus Mutual Fund analysis via Cafemutual (July 2026). Recovery multiples are the author's calculations.

Reading Today's Market Through the Same Lens

I'm not predicting a crash. The memo's lens is simpler: ask what's priced in, where the crowd is, and what everyone assumes will continue. Two data sets from 2026 are worth a look.

Table 4 compares where each index stood against its previous peak at the end of July 2026.

Table 4: Distance from the previous peak, 31 July 2026

Source: Abakkus Mutual Fund via Cafemutual (as of 31 Jul 2026).

Table 5 shows a month where the headline hid the story.

Table 5: October 2026, month to date (27 Oct)

Source: Oquilia Market Pulse, month-to-date as of 27 Oct 2026. Percentage fell is the author's calculation (2,147 of 2,611).

Neither table proves a bubble. One month of narrow breadth happens without a crash. They are the kind of evidence the memo asks you to check before you decide the current trend is destiny.

Seven Questions to Ask Before You Buy

Knowing the list of mistakes doesn't stop you from making them. A habit does. Table 6 turns each family into a question with a red-flag answer.

Table 6: Seven questions before you buy

Source: Author's framework, built on the mistake families in Table 1. Education, not personal advice.

What If I'm Wrong?

Five reasons to hold this analysis loosely:

• These are heuristics, not timing tools. The memo came in 2005. The Indian market peaked in January 2008. Acting on the warning in 2005 would have felt wrong for years.
• Sometimes it really is different. Interest-rate regimes shift and new industries do justify higher valuations. The mistake is paying as if the change is certain and permanent.
• The memo is a US-era document. The Indian data overlay and the questionnaire are my additions, not Marks's.
• One month isn't a pattern. Narrow breadth in October 2026 happens without a crash following.
• Reading about a bias is not the same as beating it. You can agree with every point here and still do the opposite when the market is running.

Closing Thoughts

Last time we looked at what a flat market does to patience. This is the other half of the cycle: what a rising one does to judgment.

The mistakes outlive every cycle because they are not intellectual errors. They are emotional ones wearing intellectual cover stories. "This time is different" is a feeling that arrives with a footnote. So the defence is deliberately boring: write the questions down before the story starts, and answer them when you're calm.

Which of the four families is yours?

This article is educational and is not personalized financial advice. Investing involves risk of loss, and past performance does not predict future results. It paraphrases and comments on a publicly available Oaktree Capital memo; analysis beyond that is the author's own. Data as of 6 November 2026.