Buffett's own record implies a bigger Kelly bet than Li Lu's
The Kelly criterion applied to Li Lu's 14 disclosed trades already implies betting 62 percent of a portfolio on one position, a number Tenachine has called extreme. Applied to Warren Buffett's record instead, the formula outputs 65.8 percent, higher than Li Lu's. Applied to Michael Burry's, it outputs 23 percent, because his payoff ratio sits below 1.
Tenachine's guide on Li Lu's Kelly criterion output already calls a 62 percent formula-implied bet size extreme. Run the identical formula on Warren Buffett's disclosed record and the output is higher, not lower: 65.8 percent of a portfolio in one position, kelly_f on Tenachine's own page for Buffett, 0.6583. The investor usually held up as the cautious, long-horizon example produces a more aggressive formula output than the one already flagged as an outlier.
The same win rate, a very different formula output
Burry's win rate, 64.1 percent, sits within two tenths of a point of Li Lu's 64.3 percent, a gap Tenachine's guide comparing the two investors' expectancy already covers from a different angle. Feed the same near-identical win rate through the Kelly formula and Burry's output, 23.0 percent, kelly_f of 0.2305, is not even half of Li Lu's 62.0 percent. Win rate alone does not set the formula's answer. The payoff ratio, kelly_b on Tenachine's investor pages, does most of the remaining work: Li Lu's reads 15.42, Buffett's reads 11.48, and Burry's reads 0.8737, below 1.0.
A payoff ratio below 1 means the formula's own measure of a typical win is smaller than its measure of a typical loss, even though most of Burry's disclosed trades close positive. The Kelly formula, f equals the win probability minus the loss probability divided by the payoff ratio, still returns a positive number here, 0.6412 minus 0.3588 divided by 0.8737, because the win rate is high enough to clear that unfavorable ratio. It returns a much smaller positive number than either Li Lu's or Buffett's case, both built on payoff ratios in the double digits. Tenachine's guide on Burry's largest position already shows his 170-trade record concentrated in very few current holdings. A payoff ratio this much smaller than the other two investors' is a second, independent signal pointing at a different kind of trading record, not a restatement of the concentration point.
Half Kelly, all three
Nobody actually bets full Kelly, for the reasons the Li Lu guide already lays out: the formula's inputs are estimates from a limited trade history, and full sizing amplifies a wrong estimate into an extreme drawdown. Half Kelly halves the formula's own output rather than fixing the underlying uncertainty. Applied here, Buffett's half Kelly comes out around 32.9 percent, Li Lu's around 31.0 percent, and Burry's around 11.5 percent. Even Burry's more modest half-Kelly figure, on a formula this sensitive to a small sample, is still a large single position by most standards, and all three of these formula outputs are built on trade counts, 140 for Buffett, 14 for Li Lu, 170 for Burry, that are treating a win rate and payoff ratio estimated from that count as if they were fixed facts rather than a read on one investor's history so far.
None of the three figures is a recommendation, and a bigger or smaller Kelly output is not a verdict on which investor's approach is better. The formula answers a narrow question, what bet size would have maximized long-run growth if these exact inputs held for the whole track record, and all three inputs come from disclosed 13F trades that lag real positions by up to a quarter. Past performance does not predict future results for any of the three, and a formula this sensitive to its inputs is the last place to treat a single number as settled. What the comparison does show is that Li Lu's 62 percent, flagged as extreme in isolation, is not even the largest of the three once the same formula runs on two other investors' full records.