The Devil in the Portfolio: What Is Shannon’s Demon?

Meet Shannon’s Demon

Picture a man in a Bell Labs office in the 1960s, the guy who basically invented the mathematics behind every text message, YouTube video, and Wi-Fi signal you’ve ever used, taking a break from information theory to build a mechanical device that could beat the house at roulette. That’s Claude Shannon. And buried in his side projects is a little thought experiment that investors still argue about today: Shannon’s Demon.

(Small aside before we go further: if “Claude” sounds familiar from somewhere else entirely — like, say, the AI that might have helped research this very article — that’s not a coincidence. Anthropic’s chatbot is widely reported to be named after this same Claude Shannon, as a nod to the man whose 1948 information theory paper laid the statistical groundwork that modern language models are built on. Anthropic has never issued a big formal announcement about it, but it’s become pretty much accepted lore in the industry. Kind of fitting that he’d end up lending his name to a machine, given how much of his own career was spent building them for fun.)

It sounds like something out of a physics textbook (and honestly, it kind of is — the name is a wink at Maxwell’s Demon, a famous 19th-century thought experiment about breaking the laws of thermodynamics). But Shannon’s version isn’t about heat and gas molecules. It’s about something almost every investor has felt tempted by: the idea that you can generate real, “free” returns just by rebalancing.

The Setup: A Coin That Can’t Lose (Or Win)

Here’s the classic version of the demon. Imagine a stock that does something wild: every day, it either doubles in price or gets cut in half, with a 50/50 coin-flip chance of either. If you just buy and hold this stock, your expected long-run outcome is… zero. Half the time you double, half the time you lose half, and multiplied out over many days, this stock is a slow-motion trip to worthless.

Now here’s the trick. Instead of holding 100% of your money in this coin-flip stock, you split it 50/50 between the stock and cash. And after every flip, you rebalance back to 50/50.

Something strange happens. Even though the stock itself is going nowhere on average, your portfolio — the stock-plus-cash combo, constantly rebalanced — actually grows. Forever. It’s not a rounding error either; it compounds at a real, positive rate, purely from the act of selling high and buying low, over and over, mechanically, without ever needing to predict anything.

That’s the demon. A source of return that seems to come from nowhere — not from picking the right stock, not from timing the market, just from the geometry of rebalancing a volatile asset against a stable one.

Enter Edward Thorp: The Man Who Actually Cashed the Check

If Shannon dreamed up the theory, Ed Thorp is the guy who turned it into cold, hard cash — twice.

As documented in William Poundstone’s Fortune’s Formula, Thorp was a math professor with a gambling problem—except his problem was that he was simply too good at it. In the early 1960s, using early computers, he figured out that blackjack wasn’t actually a coin-flip game against the house; with card counting, a sharp player could tilt the odds in their own favor. After he published these findings in Beat the Dealer (1962), casinos panicked and rewrote their rules, turning Thorp into a minor legend.

Here’s where it gets fun: Thorp and Shannon actually knew each other. They worked together, took a literal suitcase-sized wearable computer to Vegas to beat roulette, and traded ideas about probability, information, and edge. And when Thorp eventually got tired of casinos banning him from the floor, he did the obvious next move for a man who’d just proven he could out-math a casino: he turned to Wall Street.

Thorp went on to found one of the first successful quantitative hedge funds, applying the exact same mindset — find small, provable, repeatable statistical edges, and let them compound — to markets instead of card tables. Shannon’s Demon wasn’t just an abstract curiosity to these two; it was a live question. Could disciplined rebalancing really manufacture a “free lunch” out of pure volatility?

The Kelly Criterion: The Math That Tells You How Much to Bet

As documented in William Poundstone’s Fortune’s Formula, there’s a third character in this story who deserves equal billing: John Kelly, another Bell Labs researcher and Shannon’s colleague, who in 1956 published a formula that answers a question every gambler and investor eventually runs into — not “should I bet?” but “how much should I bet?”

Bet too little on a good edge, and you leave money on the table. Bet too much, and even a genuinely favorable bet can wipe you out through sheer bad luck along the way — the math of compounding is brutal about punishing overconfidence. The Kelly Criterion is a formula that finds the sweet spot: the bet size that maximizes your long-run growth rate without risking the kind of ruin you can’t recover from.

Thorp took Kelly’s formula out of the casino and into his portfolio construction, using it to size his positions. And Shannon’s Demon lives right next door to Kelly in the family of ideas: both are about the mechanics of compounding under uncertainty — not “what will happen,” but “given that I don’t know what will happen, how do I structure my bets so that volatility works for me instead of against me?”

It’s a genuinely elegant lineage: an information theorist’s thought experiment, a card counter’s formula for bet sizing, and a hedge fund built on both — all pointing at the same seductive idea. Volatility itself, tamed by discipline, can be turned into a source of return.

So… Does It Actually Work?

Here’s a wrinkle worth knowing before : Shannon himself wasn’t just a theorist. According to William Poundstone’s Fortune’s Formula, Shannon ran his own portfolio using ideas adjacent to this thinking and reportedly compounded around 28% annually from 1966 to 1986 — decades that included brutal bear markets. That track record makes the demon’s real-world failure in Part 2 land even harder: the math checked out for Shannon personally, but as you’re about to see, it doesn’t automatically transfer to any volatile asset.

Here’s the catch, and it’s the reason this article exists as part one of a series rather than the whole story: Shannon’s Demon works beautifully on a stock that goes nowhere on average — a pure coin-flip asset with no underlying drift. Real markets, especially something like the Nasdaq-100, don’t behave like that. They have a strong long-term upward drift, punctuated by real volatility.

And that changes everything. Rebalancing between a stock and cash means you’re constantly trimming your winners and topping up during dips — which is exactly the right move when an asset is chopping sideways, and exactly the wrong move when an asset is quietly compounding upward for two decades.

I actually built this out — ran Shannon’s Demon-style weekly and threshold rebalancing on 20 years of real Nasdaq-100 (QQQ) data, both as a lump-sum experiment and layered on top of monthly dollar-cost averaging — and compared it against just… buying and holding, or buying every month and doing nothing clever at all.

I actually built this out — ran the numbers on 20 years of real Nasdaq-100 data.”

The results genuinely surprised me, and not in the direction Shannon’s Demon’s fan club would want. That’s Part 2.

Next up: I ran Shannon’s Demon, a classic trend-following stop-loss system, and a few other “smart money” tricks against 20 years of real Nasdaq-100 data. One of them lost to doing absolutely nothing. Here’s the full breakdown.