In March 2020, Bitcoin dropped 50% in 48 hours. Most retail traders either got liquidated or stopped out with nothing left to trade. The ones who didn't? They weren't better at reading charts. They sized their positions so that a 50% move — something that happens in this market — wouldn't wreck them.
That calculation is position sizing. Almost nobody does it properly.
The Problem With "Risk 2% Per Trade"
You've heard it: "Never risk more than 2% on a single trade." It's the standard advice from every trading book written since 1985. It's also almost useless without knowing three things: your win rate, your average win-to-loss ratio, and whether your positions are actually independent events.
Risk 2% on a trade where you have 70% conviction looks the same as risk 2% on a trade where you're guessing. That's not risk management. That's cargo-culting.
Here's what actual position sizing math looks like. If your strategy wins 40% of the time with an average gain of 3x your loss, you can stomach much larger position sizes than someone winning 55% of the time with a 1.2x average ratio. Same "2% risk," completely different expected outcomes.
The second problem is correlation. If you're long Bitcoin, Ethereum, and Solana simultaneously, you're not running three positions — you're running one concentrated bet dressed up as diversification. When correlation is high (which it almost always is in crypto during stress events), your effective risk multiplies. A 2% position in each becomes a 6% effective position when they all move together, which they will.
The Kelly Criterion: Where It Actually Comes From
John Kelly worked at Bell Labs in the 1950s. He wasn't a trader. He was solving a problem about transmitting information over noisy telephone lines. The formula that bears his name optimizes the rate of capital growth assuming you know your edge and can keep betting indefinitely.
The formula: f = (bp - q) / b*
Where:
- f* = fraction of capital to bet
- b = odds received on the bet (profit/loss ratio)
- p = probability of winning
- q = probability of losing (1 - p)
For a trade with 50% win rate and 2:1 reward-to-risk, Kelly tells you to bet 25% of your bankroll. Full stop. That's what the math says if you want maximum long-term growth.
Here's the problem nobody tells you: Kelly is a volatility bomb. It assumes you know your actual win rate (you don't), that your trades are independent (they're not), and that you can stomach the drawdowns (you can't). Half-Kelly — betting 12.5% instead of 25% — is the practical version most professionals use. Quarter-Kelly for the paranoid.
I use a modified version. I call it "conviction-adjusted Kelly." Instead of plugging in theoretical win rates, I weight my position size by how confident I am that my thesis is correct, scaled against how much I'd lose if I'm wrong.
Crypto's Volatility Problem Changes Everything
Standard position sizing models assume volatility is roughly constant or normally distributed. Bitcoin is neither. A 10% daily move that would be a black swan event in the S&P 500 happens six times a year in crypto, on average.
This means your stop loss needs to be placed based on actual volatility, not arbitrary support levels. If Bitcoin's 20-day average true range is $3,200 (it's been around there recently at these prices), a "safe" 3% stop is actually two standard deviations away — which gets hit constantly.
Here's what I actually do: I calculate my position size based on where I'd be wrong, not where I'd be uncomfortable. If I'm long Bitcoin at $66,430 and my thesis is wrong if price breaks below $58,000 (meaning macro conditions have changed, not just noise), then my stop is $8,430 below entry. If I'm willing to lose $1,000 on this trade, I can buy $1,000 / $8,430 = 0.119 BTC. That's roughly 0.12 BTC, or about $8,000 in notional value.
Notice I didn't start with "how much do I want to invest?" I started with "how much am I willing to lose if I'm completely wrong?" That's the correct order.
The Conviction Ladder: How I Actually Size
Most traders size positions like they're ordering food — they pick what feels right and rationalize it afterward. I use a framework I call the conviction ladder.
Level 1 (Low conviction, testing a thesis): 1-2% of portfolio This is money I'm willing to lose completely. I'm not adding to this if it goes against me. I'm either right and build a position, or I'm wrong and I take the loss. No averaging down.
Level 2 (Medium conviction, defined catalyst): 3-5% of portfolio I have a specific reason to be here — an upcoming protocol upgrade, a regulatory decision, a technical breakout with volume confirmation. If the thesis doesn't play out within my timeframe, I'm out regardless of P&L.
Level 3 (High conviction, asymmetric bet): 8-12% of portfolio This is where I size up. These are positions where the downside is limited (in my assessment), the upside is multiples, and I have some visibility into the timeline. The Luna collapse would have been a Level 1 or Level 2 for me at most — the downside was theoretically infinite, which violates my asymmetry rule.
Level 4 (Thesis-level conviction): 15-20% of portfolio I have maybe two of these at any given time. This requires the trade to pass three tests: I can articulate exactly why I'm right in a way that would survive scrutiny, I can name the specific event that would prove me wrong, and I can hold through 50% drawdown without selling. If I can't pass all three, it's not a Level 4 position.
Notice what's missing: leverage. Size your position with leverage and you're changing the math in ways that don't show up on your position size calculator. A "small" 2% position with 10x leverage becomes a 20% position. The math doesn't care about your intentions.
The One Mistake That Destroys Accounts
Let me be specific about the most common failure I see: adding to losers.
The logic feels sound: you're lowering your average entry price, and if you were right to buy at $66,000, you're more right to buy at $60,000. But this is only true if your thesis hasn't changed. Most people add to losers because they're emotionally unable to accept the loss, not because the thesis improved.
When you add to a losing position, you're not just lowering your average — you're increasing your exposure to the thing that's already losing. Your position is now larger AND the market is showing you it disagrees with you. The combination is toxic.
Here's the rule I follow: I can add to a position only if I would open that larger position fresh today. Not "would I average down" — would I buy this amount at current prices with fresh capital? If the answer is no, I'm not averaging down. I'm doubling down on a mistake.
This is where the conviction ladder matters. If I'm at Level 2 and the price drops to a level where I'd enter at Level 3, I add — not because I'm averaging down, but because new information has increased my conviction. The distinction matters because it forces you to think about whether the world actually changed or whether you're just in pain.
Correlation: The Hidden Position Size Killer
In 2021, the "DeFi summer" narrative meant that DeFi protocol tokens moved together with eerie precision. You could have held 10 different positions across different protocols and thought you were diversified. When the sector turned, you lost on all 10 simultaneously.
Before I size any position, I ask: if everything I'm holding goes down 30% in a week, what happens? If the answer is "my account is blown," I need to reduce size or cut some positions. I don't care if the positions are "different." I care about correlation during drawdowns.
Practical check: calculate the 30-day correlation of your positions. Anything above 0.7 correlation is essentially the same bet. If you have five positions all correlated at 0.8, you're running 20% of portfolio in one effective position, not 5% in five positions.
This is why I keep a "correlation journal" alongside my trade journal. Every week, I note how my positions actually moved relative to each other. In bull markets, everything correlates to 1. In bear markets, it goes to 1 faster. Knowing your actual correlation exposure keeps you from lying to yourself about diversification.
The Framework in Practice: A Real Example
Let's say Bitcoin is at $66,430 and I'm considering a swing trade. My thesis: inflation data comes in hot next week, Fed holds or raises rates, crypto dumps 8-12%, and I buy the dip with a 20% target.
My analysis:
- Probability of thesis playing out: 35% (based on current sentiment and historical reactions)
- If wrong: Bitcoin continues higher, maybe breaks $70K
- My stop: $71,000 (thesis is wrong if price breaks with volume above resistance)
- My loss on the trade: $4,570 per BTC if stopped out
- Amount I'm willing to lose: $500
- Position size: $500 / $4,570 = 0.11 BTC ($7,300 notional)
That's a 0.11 BTC position on a $10,000 account — about 7.3% of portfolio. My "intuition" might have said "buy 0.5 BTC because I'm confident." The math said 0.11 BTC. I'd rather be right with small size than wrong with large size.
Now here's what the math actually means for my account: if I'm right 35% of the time on 2x reward-to-risk trades, I have a positive expected value strategy. Over enough trades, I grow. But I need to survive long enough to be right — and I do that by never letting one trade cost me more than I can afford to lose.
The Takeaway
Position sizing is the only math in trading that matters more than your edge. You can have the best strategy in the world and still blow up if you bet too much on any single outcome. You can have mediocre analysis and survive for years if you never risk ruin.
Three things to do this week:
Calculate your actual position sizes using the "stop-loss from thesis" method, not arbitrary percentages. Your thesis should determine where you're wrong; your loss tolerance should determine how much you bet.
Map your correlation exposure. Write down every position and estimate how it would behave if Bitcoin dropped 30% tomorrow. If you can't hold through that mentally, your size is too large.
Stop adding to losers. Average down only if you'd open the larger position fresh today. If you're holding because you're in pain, you're already making the mistake. The trade doesn't care about your feelings.
The traders who survive long enough to be right aren't the ones with the best analysis. They're the ones who never bet so much that being wrong once changes everything.