Thinking in Bets Ch. 3: The Buddy System — Decision Groups

Investment Background

Professional poker players, Duke observes, do something almost no individual retail trader does: they review hands in groups, specifically to have their reasoning attacked.

The group's explicit norm is not agreement. It is finding the flaw. A player who presents a hand and hears only "nice job" has wasted the group's time. A player who presents a hand and gets told "you had no business calling there — here's what you missed" has just gotten something valuable for free.

The Wall Street Translation

Why Most Trading Communities Do the Opposite

Search any trading forum, Discord, or social media thread built around a stock, and the dominant incentive is visible immediately: agreement is rewarded, dissent is punished.

Someone who posts a bullish thesis and gets fifty replies saying "🚀🚀🚀" has received zero decision-quality information. Someone who gets one reply saying "your revenue assumption is 40% too high — here's the filing" has received something that could save real money.

The second reply feels worse and is worth more. Most communities filter for the first kind, because agreement is pleasant and disagreement is not — which means the average trading community actively degrades decision quality while feeling supportive.

The Four Design Rules for a Working Decision Group

Duke lays out what separates a real accountability group from a chat room that happens to talk about trades:

1. Small. Three to five people, not a crowd. A crowd produces consensus theater; a small group produces actual scrutiny.
2. Selected for honesty over agreement. The right member is the one who will say your thesis is weak to your face, not the one who always likes your posts.
3. Reviews decisions, not just outcomes. The group discusses why a trade was taken, using only what was known at entry — never "well, it worked out" as a substitute for that discussion.
4. Explicitly rewards finding flaws. The norm must state, out loud, that catching an error in someone's reasoning is a contribution, not an attack — otherwise the group drifts back to politeness by default.

A Worked Contrast

Consider two versions of the same event: a trader posts a thesis for a leveraged short position ahead of an earnings report.

	Agreement-seeking forum	A working decision group
Typical response	"Great trade, I'm in too"	"What's your thesis if guidance beats — have you sized for that?"
What the poster learns	Nothing new; feels validated	A gap in the plan, before it costs money
Cost of finding the flaw	None — it wasn't found	None — found for free, before the position was live
Cost of not finding the flaw	Realized only after a bad fill	Avoided

The forum did not fail to help by accident. Its incentive structure makes flaw-finding rare, because flaw-finding is socially costly there and free in a real decision group.

Why This Differs From "Following a Guru"

A decision group is not a signal service, and the distinction matters. Following a guru's calls outsources the decision entirely — you inherit their conviction (or their resulting-driven confidence) without seeing their reasoning tested. A decision group keeps you as the decision-maker and uses the group only to stress-test your own reasoning before you commit capital. The output is a better version of your thesis, not someone else's trade idea.

Division of Labor With the Rest of the Library

Book	Owns
Misbehaving ch5	Why an individual, alone, rarely out-thinks their own bias — knowing a bias exists does not neutralize it in your own head
Six Secret Teachings (institutional flow)	Reading what large external players are doing — a different kind of outside information
This book, ch3	A structured, reciprocal peer process for testing your own reasoning before capital is at risk — not information about the market, but scrutiny of your read on it

Executable Trading Rules

1. Build a group of three to five, selected for willingness to disagree, not for being pleasant. One well-chosen skeptic is worth more than twenty people who upvote your thesis.

2. Present decisions before outcomes are known, when possible — or immediately after, using only entry-time information. The moment an outcome is known, the group's own hindsight bias contaminates the review; discuss the decision on its own terms first.

3. Make "finding the flaw" the group's explicitly stated highest-status contribution. State this norm out loud at the group's formation. Without saying it, groups drift toward the comfortable default of validation.

4. Treat a group that only agrees with you as a cost, not a resource. If a review group has not identified a real weakness in your reasoning in the last ten sessions, it is not testing you — replace or restructure it.

Relevance to a Retirement Portfolio

This chapter's mechanism applies with less urgency but real value to the far larger, far slower decisions of retirement planning: savings rate, allocation, and withdrawal strategy.

A financial decision reviewed only by sources that already agree with you — a forum that shares your risk tolerance, a search history that returns your own beliefs back to you — never gets stress-tested until a real market event does it involuntarily. A spouse, a fee-only fiduciary advisor, or even a single honest friend willing to ask "why do you believe your withdrawal rate is safe?" plays the same role this chapter assigns a poker player's decision group.

The goal is never to find someone who will talk you out of a low-cost, diversified core — that allocation does not need defending against scrutiny, because the evidence for it is unusually strong across the whole library (see Winning the Loser's Game, A Random Walk ch5, The Psychology of Money). The goal is to find someone who will ask the uncomfortable question about the parts of your plan that are genuinely uncertain — your withdrawal rate, your cash buffer size, your assumed longevity — before a real shortfall asks it for you.

Chapter 4 gives the individual half of this same discipline: a structured way to interrogate your own decision before and after it happens, when no group is available.