Kelly
Optimal Sizing Formula
2-3
Duan's Ideal Stock Count
f*
Kelly Fraction

How Much to Bet on Each Stock?

Most investing education focuses on WHAT to buy. But equally important — and often neglected — is HOW MUCH to buy. Position sizing can be the difference between mediocre returns and exceptional ones.

"Diversification is protection against ignorance. It makes little sense if you know what you are doing."

— Warren Buffett

The Kelly Criterion Explained

The Kelly Criterion, developed by John Kelly at Bell Labs in 1956, gives the mathematically optimal bet size to maximize long-term wealth growth: f* = (bp - q) / b

Where: f* = fraction of capital to allocate, b = odds received (potential gain/loss ratio), p = probability of winning, q = probability of losing (1-p).

Win ProbWin/Loss RatioKelly FractionInterpretation
60%2:140%Strong conviction — large position
55%1.5:118%Moderate edge — medium position
50%2:125%Coin flip with good odds — decent position
50%1:10%No edge — don't bet!
40%3:110%Low probability but high payoff — small position

Half-Kelly: The Practical Approach

Full Kelly is mathematically optimal but emotionally brutal — it leads to huge drawdowns. Most professional investors use Half-Kelly (f*/2) or even Quarter-Kelly for smoother returns.

💡 Kelly in Practice

  • Full Kelly maximizes growth but creates extreme volatility (50%+ drawdowns)
  • Half-Kelly achieves 75% of Kelly's growth with significantly less risk
  • Quarter-Kelly sacrifices more growth but is emotionally sustainable
  • Never exceed Full Kelly — over-betting guarantees eventual ruin
  • Reassess your edge regularly — Kelly fraction changes as conviction changes

Duan Yongping's Concentrated Approach

Duan Yongping famously said: "2-3 stocks is enough." His portfolio has historically been extremely concentrated — sometimes 70%+ in a single position (Apple). This aligns with Kelly theory: when you have high conviction, concentrate.

Position Sizing Rules of Thumb

💡 Practical Sizing Framework

  • Core positions (highest conviction): 20-40% each, maximum 3
  • Supporting positions (good but less certain): 5-15% each
  • Exploration positions (learning/testing): 1-3% each
  • Cash reserve: always maintain 10-20% for opportunities
  • Never let any single loss exceed 10% of total portfolio value
  • Rebalance when a position grows beyond 50% of portfolio

💡 Kelly Criterion — Key Summary

  • Kelly Criterion = mathematically optimal position sizing formula
  • f* = (bp - q) / b — bet proportional to your edge
  • Use Half-Kelly in practice — full Kelly is too volatile for most people
  • Concentrated portfolios (2-3 stocks) work when conviction is high
  • Never over-bet — exceeding Kelly guarantees eventual ruin
  • Position sizing matters as much as stock selection

一个赌场数学家的发现

1956年,贝尔实验室的物理学家约翰·凯利(John Kelly)发表了一篇论文。他原本是在研究电话线路中的信号噪声问题,却无意中发现了一个可以应用于赌博和投资的最优下注公式。

这个公式后来被称为「凯利公式」(Kelly Criterion),它回答了一个极其重要的问题:当你有优势时,应该押多少?

f* = (bp−q)/b
凯利公式
1956
发表年份
最优
长期财富增长率

用抛硬币理解凯利公式

假设有一个「作弊」的硬币:正面朝上的概率是60%,反面是40%。每次下注,正面赢你就赚1倍,反面输你就亏光注额。你有1万块钱,应该每次押多少?

你的直觉可能是「有优势就全押(All-in)!」但这是错的。全押几次后,只要输一次就归零。正确答案是凯利公式:

f* = (bp − q) / b,其中 b=赔率(1:1),p=赢的概率(60%),q=输的概率(40%)

f* = (1×0.6 − 0.4) / 1 = 20%

答案是每次押总资金的20%。不是10%(太保守,浪费优势),不是50%(太激进,波动太大),更不是100%(等于自杀)。

不同下注比例的长期结果

模拟100次下注(60%胜率、1:1赔率),看看不同下注比例的最终结果:

下注比例100次后预期资金波动程度破产概率
5%(太保守)~1.9万非常低~0%
10%~4.5万~0%
20%(凯利最优)~22万中等~0%
30%~15万~2%
50%~3万极高~15%
100%(All-in)~0~100%

20%的下注比例(凯利值)产生最大的长期财富增长。低于凯利值是安全但低效的;高于凯利值不仅增长变慢,还引入了破产风险。超过凯利值2倍(即40%+),长期期望反而变成负数。

从赌场到投资组合

在投资中,凯利公式需要一些调整。因为投资不是简单的「赢一倍/输全部」,而是有不同的上涨空间和下跌空间。简化后的投资版本:

仓位比例 = 预期收益率 / 最大可能亏损率

例如:你研究了一只股票,认为有60%的概率涨50%,40%的概率跌30%。预期收益 = 0.6×50% − 0.4×30% = 18%。凯利仓位 = 18% / 30% = 60%

但实际操作中,几乎所有专业投资者都会使用「半凯利」或更低——即凯利值的50%。原因很简单:你对概率的估算本身就有不确定性。

段永平的集中投资实践

"好公司有就2-3个,没必要持有20个股票。持有20个说明你对每一个都不确定。真正懂的、有安全边际的,2-3个就够了。"

— 段永平

段永平的投资组合通常只有2-3只核心股票(苹果、茅台、腾讯等)。他不用凯利公式计算具体比例,但他的做法本质上和凯利思维一致:当你有非常高的信心时,集中下重注。

💡 集中 vs 分散:真正的取舍

  • 集中投资:高信心 + 深度研究 = 高回报潜力,但要承受波动
  • 分散投资:承认无知 + 分散风险 = 稳定但平庸的回报
  • 段永平/巴菲特的选择:5-10只核心持仓,前3只占60-80%
  • 普通投资者的建议:如果你研究不深,宁可选指数基金
  • 凯利公式的启示:仓位大小应与信心成正比,但永远留有余地

仓位管理的实操框架

信心等级了解程度建议仓位对应凯利比例
A级(极度确信)研究3年+,深度理解25-40%半凯利到全凯利
B级(较有信心)研究1年+,基本理解10-20%1/3凯利
C级(有待验证)初步研究,仍有疑问3-5%观察仓
D级(纯粹好奇)只是听说了一下0%不买
现金储备等待机会10-30%永远保留

💡 仓位管理核心总结

  • 凯利公式回答的核心问题:有优势时押多少
  • 全凯利太激进,实操中用半凯利(×50%)更安全
  • 仓位大小应与你的研究深度和信心成正比
  • 永远保留现金——大机会来时你需要子弹
  • 宁可仓位太小(赚少一些)也不可仓位太大(一次出局)
  • 如果你不确定该多大仓位——那仓位就该更小

本文简化了凯利公式的数学推导。如需深入了解,推荐阅读 William Poundstone 的《Fortune's Formula》。