Shannon's Demon: how rebalancing turns volatility into return
Last updated 2026-10-29
Shannon's Demon is a thought experiment attributed to Claude Shannon, the father of information theory, from his lectures at MIT in the 1960s. The idea fits in one sentence: even an asset that only swings up and down and goes nowhere in the long run can make you money if you hold it alongside cash at a fixed ratio and keep restoring that ratio.
The coin-flip example
Imagine an asset whose price doubles on heads and halves on tails, every day.
- Just hold it: one double and one halving gives 2 × 0.5 = 1. You go nowhere (geometric mean 1.0).
- Hold 50% in the asset and 50% in cash, rebalancing to 50:50 daily: a heads day leaves you at 0.5 + 1.0 = 1.5×, a tails day at 0.5 + 0.25 = 0.75×. Since 1.5 × 0.75 = 1.125, you gain 12.5% every two flips — about 6.1% per flip on average.
Selling some after a rise and buying more after a fall forces you to sell high and buy low, automatically. This is called volatility harvesting or the rebalancing bonus.
Up 10%, then down 10%? (volatility drag)
If a stock rises 10% one day and falls 10% the next, you do not break even — you lose. $100 becomes $110, then drops by 10% of $110 to $99 (1.1 × 0.9 = 0.99, -1%). The bigger and more frequent the swings, the larger the loss. This is volatility drag, and it is why leveraged ETFs are hard to hold for the long run.
| +10% then -10%, repeated 10 times | Result |
|---|---|
| 100% in the asset | -9.6% |
| 50% asset + 50% cash, never rebalanced | -4.8% |
| 50% asset + 50% cash, rebalanced to 50:50 every time | -2.5% |
When the ups and downs are equal and the expected return is zero, the demon cannot erase the loss — it only shrinks it. For rebalancing to turn swings into actual profit, the asset needs a positive average (arithmetic) return plus large volatility, as in the coin-flip example. Put simply, rebalancing claws back part of the volatility drag.
When it works and when it hurts
| Condition | Result |
|---|---|
| Large swings but no lasting trend | The demon works (rebalancing bonus) |
| Assets move differently (low correlation) | Bigger bonus |
| One asset trends strongly in one direction | It costs you (you sell the winner early and keep buying the loser) |
| Trading too often | Fees and slippage eat the bonus |
How this site's bot applies it
The bot holds BTC 25%, SOXL (tracking the 3x US semiconductor ETF) 25% and USDT cash 50%, checks every hour, and rebalances whenever any asset drifts 2 percentage points from target. It is a live-market experiment in Shannon's Demon: two very volatile assets paired with cash.
Tested on real data: did the demon work?
Starting on the same day (2026-06-30) with the same 25/25/50 weights, we computed what would have happened with no rebalancing at all, using actual Bybit prices, and compared it with the live bot.
| Month | Rebalanced (bot) | Never rebalanced | Gap (pts) |
|---|---|---|---|
| 2026-06 | +4.24% | +4.24% | +0.01 |
| 2026-07 | -13.72% | -13.21% | -0.52 |
| 2026-08 | +4.99% | +6.16% | -1.17 |
| 2026-09 | +9.34% | +6.44% | +2.91 |
| 2026-10 | -8.96% | -6.22% | -2.74 |
| 2026-06-30 – 2026-10-29 | -6.49% | -4.17% | -2.32 |
So far, rebalancing is behind by 2.32 points. BTC climbed steadily while SOXL fell by more than half — exactly the one-directional trend in the table above. The bot kept trimming the rising BTC and kept buying the falling SOXL. The demon earns in sideways markets and charges a fee in trending ones. This page is updated weekly to track how the gap evolves.
Takeaways
- Shannon's Demon earns from a rule, not a forecast. You never need to call market direction.
- It is not free: during long trends it can lag a simple buy-and-hold.
- Leveraged assets such as SOXL carry their own long-run volatility decay and add significant risk.