The Kelly Criterion gives you a mathematically defensible answer to "how much of my account should I put into this trade?" — and for wheel traders, it's more useful than the vague "never risk more than 5%" advice you've probably heard. The short version: Kelly tells you to size your position based on your edge, not your gut.

What Kelly Actually Says

The classic Kelly formula is: f = (bp - q) / b

Where f is the fraction of your capital to risk, b is your net odds (profit/loss ratio), p is your probability of winning, and q is your probability of losing (1 - p).

For a cash-secured put, let's say you're selling a 30-delta put on NVDA at $800. You collect $12 in premium, and your max loss if NVDA goes to zero is $788 (the strike minus premium). Your realistic max loss — to a stop at $720, say — is $68. So b = 12/68 = 0.176. Your probability of winning (staying above $800 at expiration) is roughly 70%, so p = 0.70, q = 0.30.

Kelly says: f = (0.176 × 0.70 - 0.30) / 0.176 = (0.123 - 0.30) / 0.176 = -1.0

Negative Kelly. That's a position you shouldn't take at all by pure Kelly logic. That stings to hear, but it's exactly why this exercise is valuable — it forces you to confront whether your edge is actually there.

Why Wheel Traders Usually Get a Negative or Near-Zero Kelly

The wheel is a low-b strategy. Your wins are small (premium collected) and your losses can be large (stock drops 20-30%). Kelly punishes that asymmetry hard. This doesn't mean the wheel is a bad strategy — it means Kelly in its raw form was designed for gamblers with symmetric payoffs, not options sellers.

The fix is to think about Kelly differently for the wheel. Instead of using max loss as your denominator, use your managed loss — the loss you'd actually take if you followed your trading plan. If you close losers at 2x the premium received, your b changes dramatically.

Back to NVDA: you collected $12, and you'll close if the position goes against you by $24 (2x premium). Now b = 12/24 = 0.50. With p = 0.70, Kelly becomes: f = (0.50 × 0.70 - 0.30) / 0.50 = (0.35 - 0.30) / 0.50 = 0.10

Ten percent of your capital. That's a real, actionable number.

Half-Kelly Is Your Friend

Most professional gamblers and traders use Half-Kelly, not full Kelly. Full Kelly maximizes long-run growth but creates brutal drawdowns — you can lose 50% of your account even when you have a real edge. Half-Kelly cuts that volatility significantly while giving up only a small amount of expected growth.

So if full Kelly says 10% on that NVDA put, you're actually putting 5% of your account to work. On a $100,000 account, that's $5,000 in risk — not $5,000 in notional. Your notional on a $800 put is $80,000 per contract, but your risk (your 2x stop) is $2,400 per contract. So Half-Kelly at $5,000 risk means you can sell roughly 2 contracts.

This is the part most traders skip. They think about notional exposure, not risk-adjusted exposure. Kelly doesn't care about notional. It cares about what you can actually lose.

Accounting for Correlation

Here's where wheel traders specifically need to pay attention. If you're running the wheel on AAPL, MSFT, and NVDA simultaneously, you don't have three independent positions. You have three highly correlated tech positions. Kelly assumes independence between bets — violate that assumption and you're taking on more risk than the math suggests.

A practical workaround: treat correlated positions as a single position for Kelly sizing purposes. If AAPL and NVDA move together 70% of the time (their 1-year correlation has been around 0.65-0.70), you're not getting the diversification benefit you think you are. Size your combined tech exposure using one Kelly calculation, then split it across tickers.

Running the wheel on SPY alongside NVDA is a better diversification story — SPY is the market, NVDA is a single name. Even then, in a real drawdown, everything falls together.

The Practical Setup

Here's how I'd actually implement this. Before entering any wheel position, run this quick check:

Figure out your managed loss — what you'll close at, not theoretical max loss. Calculate b as premium / managed loss. Estimate p from your historical win rate on similar setups (if you don't have data, use 65% as a conservative baseline for 30-delta puts). Plug into Kelly, take half. That's your risk allocation.

On a $100,000 account running the wheel on AAPL around $170, selling a 30-delta put for $2.50 with a 2x stop at $5.00 loss: b = 0.50, p = 0.65, Kelly = (0.325 - 0.35) / 0.50 = -0.05. Negative again. That's telling you the premium isn't rich enough relative to your stop. Either find a week with higher IV, move to a closer strike to collect more premium, or skip the trade.

That's the real value of Kelly — not a formula that tells you how much to size, but a filter that tells you when the math doesn't support trading at all.

Your action today: Take your last five wheel trades, reconstruct the Kelly calculation using your actual managed stops, and see how many were negative Kelly. If more than two were, your position selection process needs work before sizing even matters.