Think Like Everyone Else, Lose Like Everyone Else | Brent Donnelly

Brent Donnelly's central argument is that consensus thinking produces consensus results, and since most traders lose money and most investors underperform the index, independent thinking is not optional but a prerequisite for edge. His book Trade Outside the Box extends the four disciplines of his earlier Alpha Trader, combining fundamentals, technicals, behavioral analysis, and quantitative methods, with risk management remaining the dominant theme.

Donnelly uses poker as a structural analogy for trading discipline. The tight-aggressive concept, folding weak hands often but pressing hard with strong ones, corrected his natural tendency to stay in bad positions. He identifies a specific behavioral failure in himself that runs counter to the typical loss-chasing pattern: he over-trades when performing well, taking on more risk precisely when he should be reducing it. His mechanical fix is conditional formatting in his P&L spreadsheet that triggers automatic risk reduction when he is over-earning. He connects this to the euphoria of running up a large gain, which produces the same overconfidence that erodes it. He notes that DSM-5 symptoms of gambling disorder map directly onto poor trading behaviors, and he cites Jesse Livermore as an example of extraordinary skill destroyed by inadequate risk management, resulting in three bankruptcies and eventual suicide. He invokes ergodicity to explain why even skilled traders like Niederhoffer blow up every five to six years by running strategies that are convex in the wrong direction, arguing that approaching ruin repeatedly will eventually produce it regardless of skill level.

On large language models, Donnelly was cautious about writing on the topic because the content risks obsolescence within two years. His core observation is that tools like Gemini Deep Research produce the most vanilla, consensus view on any stock or macro topic by aggregating all available information, which he finds useful rather than limiting. He applies Keynes's beauty contest principle: markets reward picking who everyone else thinks will win, not who you think will win. He gives a concrete example where LLM consensus suggested a 50 percent decline in a biotech stock following a headline, but the stock was only down 9 percent three minutes after the news, creating a tradable divergence with a target of down 30 percent. He also uses Claude to find patterns in time series data for idea generation while acknowledging this constitutes data snooping that would not hold up out of sample as a systematic strategy. He cautions that asking an LLM for the best macro currency trade produces output comparable to a 22-year-old analyst with limited market knowledge, and he recommends running multiple models simultaneously to cross-check outputs.

Donnelly describes his edge as having shifted from predicting central bank policy or macroeconomic data toward predicting what humans will do over the next one to two weeks. He references Jim Grant's principle that the key to making money is thinking what everyone else thinks but just slightly before them, and he sees LLM consensus detection as a tool for identifying narratives that are relevant but not yet fully priced in. He uses humanoid robotics as a current example, arguing that the stocks an LLM identifies as top picks are likely the ones retail will buy, making them worth owning ahead of that flow.

On the Fed, Donnelly's base case is that Fed funds remain unchanged through December, with Powell's hawkish tone being performative rather than predictive. He supports this with the observation that the Fed has missed its inflation target for 62 consecutive months and has been comfortable doing so, and he notes that every new Fed chair tends to start more hawkish than they ultimately prove to be. He acknowledges a real risk that a strong non-farm payrolls print could cause the Fed to tease a September hike at the July FOMC, and he specifies that a payrolls print of 28,000 with unemployment at 4.4 percent would likely cause the Fed to ease off hawkish guidance.

On FX, Donnelly maintains that rate differentials remain the primary driver of currency moves. He describes dollar-yen as effectively a pegged currency with significant jump risk, with MOF intervention capping upside above 162 but insufficient sellers to push it lower. He argues unilateral intervention is structurally ineffective because the underlying policy variables, BOJ hiking, Fed cutting, and fiscal tightening, are not aligned. Coordinated intervention has worked almost every time historically, but he does not expect Treasury or the Fed to participate. He identifies a dangerous feedback loop in JGBs where rising yields lead to simultaneous bond selling and yen weakness resembling an emerging markets loss of fiscal credibility, and notes that the velocity of yield moves matters more than the level.

Donnelly is skeptical of gold and Bitcoin as inflation hedges. He says empirical evidence for gold's inflation-hedging role is inconsistent across time periods and that its reputation derives disproportionately from its 1980s performance. On Bitcoin, he argues the asset has cycled through narratives including store of value, peer-to-peer cash, digital gold, and risky asset, and none have fully panned out. He predicts Bitcoin may not see another major bull run for a very long time and questions what macro outcome a buyer at current levels is actually hedging for, suggesting no clear answer exists. He prefers equities in a nominally higher inflation environment because they are productive assets generating nominal earnings.