Trade Your Way to Financial Freedom Ch. 4: The Marble Game

Investment Background

To show audiences what position sizing does, Tharp ran a simple game in his seminars. A bag holds marbles of different colours, each representing a trade outcome: some win a multiple of the amount risked, most lose it. Marbles are drawn one at a time and replaced, so every player in the room experiences exactly the same sequence of wins and losses. Each player starts with the same stake and decides only one thing: how much to risk on each draw.

The outcome is always the same. At the end of the game, players' results range from large gains to total ruin, although every one of them traded the same system with the same results. Some go broke on a game with positive expectancy. Tharp's point is that the sizing decision alone can decide whether a good system makes or loses money.

The library's Asymmetric Bet Sizing covers sizing from the side of conviction and thesis, and Way of the Turtle covers volatility-based unit sizing. This chapter concerns Tharp's narrower and more psychological lesson: the same trades, sized differently, produce different fates, and most people size badly in predictable ways.

The Wall Street Translation

A Worked Example: Twenty Draws

Take a bag where 60% of marbles win 1R and 40% lose 1R. The game has a clear positive expectancy of +0.2R per draw. The group draws 20 marbles: 12 winners and 8 losers, including a streak of five losses in a row.

Each player risks a fixed percentage of current capital on every draw.

1. Risking 2% per draw: the five-loss streak costs about 10%. The final result is about +8%.
2. Risking 10% per draw: the streak costs about 41%. The final result is about +35%.
3. Risking 25% per draw: the streak costs about 76%. The final result is about +46%.
4. Risking 40% per draw: the streak costs about 92%. The final result is about −5%.

The trades were identical. The player risking 40% of capital on a system that won 60% of the time lost money. The player with the best final result endured a 76% drawdown along the way, which almost no real person would sit through without quitting at the bottom.

Why Overbetting Loses Even When You Are Right

The arithmetic behind this is the asymmetry of losses. A 50% loss requires a 100% gain to recover. When risk per trade is large, a losing streak digs a hole so deep that later wins, even plentiful ones, cannot refill it. Beyond a certain point, raising the bet size lowers the long-run result, and at an extreme it turns a winning system into a losing one.

The Behavioural Patterns Tharp Saw

In the game, Tharp observed the same mistakes again and again:

1. Raising size after wins. Players who felt confident bet more, just before the losing streak arrived.
2. Raising size after losses to "win it back". This is the fastest route to ruin in the game.
3. Betting everything on one draw after a few successes, from the conviction that they had "figured out" the bag.
4. Ignoring the objective. Players with a stated objective of avoiding large losses still sized for maximum gain when the room became excited.

None of these errors has anything to do with the system. All of them are about the person.

Sizing as the Bridge to Objectives

This is why Tharp places position sizing directly after objectives. The same system can meet an objective of "steady growth, never down more than 15%" or "maximum growth, whatever the drawdown", depending only on size. Choosing a size is choosing which objective you are pursuing, whether you intend to or not.

Executable Rules

1. Fix the risk per trade in advance, as a percentage of current capital, and derive it from your drawdown objective (Chapter 2).
2. Never raise size to recover losses. If anything, reduce it during a drawdown until the record shows the system is behaving normally.
3. Do not raise size because of recent wins. Raise it only on a schedule and only because capital has grown, keeping the percentage fixed.
4. Test your sizing against the worst streak you can imagine, not the worst you have seen. The marble game's five-loss streak was modest.

Relevance to a Retirement Portfolio

The marble game is the clearest demonstration of why retirement portfolios must be sized for survival rather than for maximum return. A retiree drawing income from a portfolio experiences losses like a player who cannot add new chips, so a deep drawdown early on may never be recovered.

For the portfolio as a whole, the stock allocation plays the role of risk per trade, and it should be chosen so that the worst plausible sequence does not force a sale at the bottom. If any trading happens alongside a low-cost index core, it must be sized so that a streak far worse than any in the record still leaves the retirement plan untouched.