Kelly Criterion is the first thing anyone hits when they start reading about position sizing. Bet a fraction of your bankroll equal to your edge divided by the odds, and long-term growth is maximized. The math is clean and hard to argue with. The catch is that it assumes you know your edge to the decimal, and in live trading you basically never do.
Why full Kelly gets people hurt
Kelly's optimal bet is only optimal when your edge estimate is correct. In practice, traders overestimate their edge, and they overestimate it badly. You backtest something, see a 60 percent win rate at 1.5-to-1 reward-to-risk, feed it into the formula, and size up with confidence. That backtested edge is almost always puffed up by overfitting, survivorship bias, and a failure to account for slippage and fees. So you are sizing off a number that was never real to begin with.
Full Kelly is also violently volatile. Even with a genuinely accurate edge, the Kelly-optimal portfolio runs 50-percent-plus drawdowns often enough to become a real problem. Most people cannot sit through that, and most allocation mandates will not let them, no matter how pretty the long-run math looks.
The standard fix is fractional Kelly, usually a half or a quarter. Half-Kelly buys you roughly 75 percent of the growth rate at about half the variance. Quarter-Kelly gets you around 50 percent of the growth with a far smoother curve. You give up some compounding and you purchase a lot of drawdown protection, and for most people that is a trade worth making every time.
Risk-based sizing is what I actually reach for
For most traders, risk-based sizing wins on practicality alone. Decide the most you are willing to lose on a single trade, usually 1 to 2 percent of the account, then size the position so a stopped-out trade loses exactly that and no more.
The arithmetic is trivial. Position size equals your risk amount divided by the distance to your stop. On a 100,000 dollar account risking 1 percent, so 1,000 dollars, a stop 5 percent below entry gives you a 20,000 dollar position. Move the stop to 10 percent below entry and the position shrinks to 10,000.
What I like is that it tunes itself to the trade. Volatile setups with wide stops get smaller positions, tight setups get larger ones, and your dollar risk per trade stays flat across all of them. That flat risk matters more than it sounds, because it keeps performance attribution honest. When a trade blows up, you know it was the idea that failed and not the sizing.
Correlation is where sizing frameworks quietly break
This is the spot where most sizing rules fall apart. Five positions in correlated assets, each sized to 1 percent risk, is not 5 percent portfolio risk. Depending on the correlations it might be 3 percent, or it might be close to the full 5. And in a correlated crash, all five stops trigger at the same moment, so you learn the real number the hard way.
The fix is easy and almost nobody bothers. Shrink your per-position size when you are holding several names that move together. A rough rule I use is to divide per-trade risk by the square root of the number of correlated positions. Four correlated positions, each gets about half its normal size. It is not precise, but it is close enough to stop one ugly day from taking the whole account with it. On Blockcircle this is the part people skip most, because staring at five separate charts hides how much they actually rhyme.
Scaling in and out
Fixed-size entries and exits leave both money and safety on the table. Scaling in, adding as the trade proves itself, keeps your average risk low on the ideas that fail and lets you build real size on the ones that work. A common version is a half-size initial entry, with the rest added once price moves your way by a set amount.
Scaling out runs on the same logic in reverse. Peeling off partial profits at preset levels banks gains while leaving something on to keep riding. Someone who takes a third at each of three targets ends up with a smoother equity curve than someone holding the full position to one target, even when the single-target version carries a higher expected value per trade. Smoother curves are easier to stay disciplined on, and staying disciplined is most of the job.
Spend more time here
Sizing is the highest-leverage decision in trading and it gets almost none of the attention. Most people pour 90 percent of their effort into entries and maybe 10 percent into how much to bet. Flip that ratio. Put more thought into how much to trade than into exactly when, and your risk-adjusted results improve across essentially every style. Start with a fixed 1 percent per trade plus a correlation haircut, then tighten from there once you have watched it run for a while.