The Kelly Criterion: Too Smart for Your Own Good?

Right, let me set the scene. It’s 2015. I’m sitting in a Mayfair wine bar with this absolutely brilliant quant from Cambridge. PhD in mathematics. Published papers. The whole nine yards. He’s just lost £240,000 on a single GBP/USD trade using something called the “Kelly Criterion”—a mathematical formula that supposedly tells you the optimal bet size to maximize long-term wealth.

I watch him stare into his Châteauneuf-du-Pape like it’s the ocean, and I think: Here we go again. Another genius brought down by his own intelligence.

This is the story of the Kelly Criterion, and why it’s simultaneously the smartest and dumbest thing you could do with your trading account.

The Beautiful Trap

The Kelly Criterion was developed by John Larry Kelly Jr. in 1956. Basically, it’s a formula that calculates the optimal fraction of your bankroll to risk on each bet, given your win rate and risk/reward ratio. The formula is deceptively elegant:

f = (bp - q) / b*

Where:

f* = fraction of bankroll to wager
b = odds received
p = probability of winning
q = probability of losing (1 - p)

On paper? Chef’s kiss. It’s mathematically proven to maximize your long-term wealth growth. Theoretically, if you follow Kelly, you will outperform any other betting strategy. It’s the Holy Grail. It’s the answer. It’s—

It’s also a financial doomsday device wrapped in mathematical ribbon.

Why the Quant Got Destroyed

Here’s what happened with my Cambridge mate. He had a backtested system with a 55% win rate and a 1:2 risk/reward ratio. By the Kelly formula, he should be risking about 5% of his bankroll per trade. Sounds reasonable, right?

So he did. He risked 5% on twenty consecutive losing trades. That’s a 35% account drawdown in about three weeks. Not catastrophic on paper, but here’s the problem: he didn’t have three weeks of mental stability.

On day ten, he started second-guessing his system. By day fourteen, he was revenge trading. By day twenty, he was betting the Kelly Criterion for his entire remaining account on a single “sure thing” that obviously wasn’t. He blew it to £3,000.

The formula was perfect. His execution was destroyed by human psychology.

The Dirty Little Secret

Nobody talks about this in the academic papers, but the Kelly Criterion assumes several things that don’t exist in reality:

One: Your win rate and risk/reward ratios are perfectly known and stable. They’re not. You think your backtest is your future? That’s adorable. Markets evolve. Your edge decays. Murphy’s Law is real.

Two: You have infinite capital. Kelly works beautifully as you approach infinity. With £50,000? You’re one bad streak away from catastrophe.

Three: You can actually execute the formula mechanically without losing your mind. You can’t. No one can. I’ve tried. I’ve seen people try. You get a 30% drawdown, your wife asks why the mortgage payment is late, and suddenly you’re trading emotionally, not mathematically.

Four: There are no black swans, liquidity shocks, or flash crashes. Narrator: There are.

The Practical Problem

Let’s say you’re trading EUR/USD with a decent 52% win rate and a 1.5:1 reward-to-risk. Pure Kelly says you should risk 2% per trade. Sounds fine, right?

Now imagine you get a 10-trade losing streak. (Spoiler: this will happen. Variance is a bitch.) You’re down 20%. Your equity curve looks like a ski slope. Your confidence is obliterated. You’re emotionally destroyed.

If you’d been using half-Kelly (1% per trade), you’d be down 10%. Slightly annoying. Completely manageable. You could actually stick with the system.

That’s the real edge in trading: Consistency over optimization.

What Actually Works

After fifteen years of this, I can tell you the uncomfortable truth: the traders making serious money aren’t using pure Kelly. They’re using a fraction of it. Half-Kelly. Quarter-Kelly. Sometimes even lower.

They’re not trying to maximize growth. They’re trying to survive variance long enough for compound returns to work. It’s unsexy. It’s not mathematically optimal. But it works.

The best traders I know use a simple mental model:

They risk 1-2% per trade (not the Kelly-optimized 5%)
They accept slower growth
They never experience account-destroying drawdowns
After ten years, they’ve crushed the “optimal” Kelly traders because they’re still in the game

The Real Edge

Here’s what kills me about this: everyone obsesses over the formula when the real game is discipline. You could hand the Kelly Criterion formula to a thousand traders, and I’d bet £100,000 that 999 of them blow up within five years.

Why? Because mathematics doesn’t account for the bit between your ears. Your brain is the weakest link in the entire system. Your psychology will betray you. Your emotions will override your calculations. Your fear and greed are more powerful than any formula.

The Kelly Criterion isn’t wrong. It’s just too powerful for humans to handle responsibly—like giving a teenager a sports car and a bottle of vodka. Technically possible. Practically disastrous.

My Honest Take

If you’re obsessed with the Kelly Criterion, use half-Kelly. It’s mathematically proven to still maximize long-term wealth growth (just slower), but it’s psychologically sustainable. You’ll actually stick with the system when the inevitable losing streaks arrive.

Better yet? Use a simple 1-2% risk per trade, build your account like a sensible human, and stop trying to squeeze every last percentage point of theoretical optimization out of a market that doesn’t care about your PhD.

The market rewards survivors, not mathematicians.

Now, if you’ll excuse me, I have a meeting with another “brilliant” trader who just discovered the Kelly Criterion. I’ll bring tissues.

P.S. – That Cambridge quant? He took a job in quantitative analysis. Last I heard, he’s making £300K a year and never trades again. Seems he learned his lesson.