Bankroll Management: How Advantage Players Make Positive EV Compound

By Charlie Nguyen · Published Nov 1, 2026 · 10 min read

Advantage players don’t think about money the way most people do. To us, a bankroll isn’t savings and it isn’t a score. It’s a tool: the inventory we use to keep placing positive expected value (+EV) bets. As long as every bet has an edge and no single bet can knock us out, the math compounds in the long run.

I learned that at the tables and now apply it to options selling. Here’s the framework, with the real math: expected value, the law of large numbers, variance, Kelly sizing and risk of ruin, and how it shapes the way I size SPX put credit spreads.

Money is a tool, not a scoreboard

A card counter doesn’t celebrate a winning hand or mourn a losing one. A hand is one sample from a game where the player has the edge. What matters is getting the money in at the right times, in the right size, as many times as possible. The bankroll is what buys those repetitions.

That changes how you feel about money. You stop asking “did I win today?” and start asking “was that a good bet, and was it the right size?” A good bet can lose. A bad bet can win. Over enough bets, only the quality of the bets shows up.

Positive EV: the only bets worth making

Expected value is the average result of a bet if you could make it over and over: (chance of winning × average win) − (chance of losing × average loss). A +EV bet makes money on average; a −EV bet loses on average, and no betting system or bankroll can fix that. Every casino game has negative EV for the regular player, which is why the house wins. An advantage player only plays when the edge flips to their side.

Options have the same structure. When I sell premium, my edge is the gap between what the options market expects and what usually happens. From Jan 2, 1990 to Oct 29, 2026, SPX’s implied volatility (VIX) averaged 19.4%, while the volatility SPX actually delivered over the following month averaged 15.4%; implied ran above realized 85.3% of the time. SPX closed inside its one-day expected move 81.7% of days (Cboe data, 9,270 days). That gap, the volatility risk premium, is the option seller’s house edge (more in selling options outside the expected move).

It isn’t free money. Sellers collect small amounts often and pay large amounts rarely. In the model backtest on our SPX Desk (Jun 2, 2009 to Oct 23, 2026, 726 trades of a 25-point SPX put credit spread), the spread won 96.0% of the time, but the average loss (−$1,622 per spread) was about 13 times the average win ($121). Net, it averaged $51 a spread, about 2.16% of the money at risk. Small, positive, and only useful if you can play it hundreds of times.

The law of large numbers and variance

The law of large numbers says that as the number of independent bets grows, your average result moves toward the expected value. The catch is variance: over a handful of bets, luck swamps edge. In the backtest, the result per trade had an average of 2.16% of risk but a standard deviation of 14.9%, about 7 times bigger.

I resampled the backtest’s 726 trade results 4,000 times to see how often a trader at a fixed size would be behind after a given number of spreads:

Ten trades in, almost a third of runs were losing money with the exact same edge. By a few hundred trades, the edge had nearly always shown up. That’s the advantage player’s mindset in one table: volume turns edge into results, and the bankroll’s job is to survive until the volume arrives.

Kelly sizing: how big is too big

In 1956, Bell Labs scientist John L. Kelly Jr. published “A New Interpretation of Information Rate” in the Bell System Technical Journal, showing how to size bets to maximize the long-run growth of a bankroll. Edward O. Thorp later applied it to blackjack in Beat the Dealer (1962), and then to markets. For an even-money bet, the Kelly fraction is simply your edge.

Illustrative example (not market data): a coin that wins 60% of the time at even money has a 20% edge, so Kelly says bet 20% of the bankroll. Here’s the long-run growth per bet at different sizes:

Same coin, same positive edge. Bet 40% and the bankroll shrinks in the long run; bet 50% and it shrinks faster. A +EV bet sized too big becomes a losing strategy. Under-betting costs a little growth; over-betting can cost everything.

Run Kelly on the SPX backtest and it says to risk about 55% of the bankroll per spread. That’s a red flag, not a plan. It assumes the backtest’s edge is exactly right, that trades are independent and that the future looks like the past. Half Kelly keeps about 70% of full Kelly’s growth rate in this model, with far smaller swings, and I size well below even that.

Risk of ruin: the one rule

Risk of ruin is the chance your bankroll falls so far you can’t keep playing. It’s the number advantage players care about most, because a ruined player can’t collect on any future edge. Here is the same 726-trade backtest, resampled 4,000 times, at different sizes (share of the bankroll at risk per spread; bankroll multiples, not dollars):

At 2% to 5% of the bankroll per spread, none of the resampled paths had a 50% drawdown. At half Kelly, 66% did. At full Kelly, every path did, and going nearly all-in turned a positive-EV strategy into a near-certain wipeout: 82% of paths lost 90% or more. These are model numbers, and real markets have fatter tails than a resample of the past, which is one more reason to stay small.

Applying it to SPX put credit spreads

Here’s an example from the Oct 30, 2026 SPX chain, with SPX near 7,639 and a 149-point expected move to Nov 5, 2026. The short strike sits just outside the expected move, as on Open Lines:

POP is the model chance the spread expires worth its full credit, POT the chance SPX touches the short strike before expiry, and P50 the chance of being able to close at half the credit (POP vs POT vs P50 explained). The bankroll question is the max loss: 20.75 points, or $2,075 per spread. I size so that losing the full width on everything open at once is a bad month, not the end of the account. Defined risk is what makes that math possible; it’s the same reason I prefer index options (why I trade indexes instead of single stocks).

My bankroll rules

1. Only +EV bets. If I can’t explain the edge, there isn’t one. For selling premium, that means strikes outside the expected move and decent IV rank.
2. Size by the worst case. A spread’s max loss is its width minus the credit. I add up the max loss on everything open, not just one trade.
3. A small fraction of Kelly. My edge estimates are uncertain, so I bet a fraction of a fraction. Slower growth, far lower risk of ruin.
4. Defined risk. Spreads, covered calls and poor man’s covered calls, no naked short options. Longer-dated ideas like the Slingshot Trade get sized for the puts being tested.
5. Judge the bet, not the result. I review whether the trade was +EV and the right size, not whether it won.
6. Never blow up. Every rule above serves this one. The edge only compounds for players who are still at the table.

Sources and method. SPX history: Cboe daily data for _SPX and _VIX, Jan 2, 1990 to Oct 29, 2026; realized volatility is the following 21 trading days’ close-to-close volatility, annualized. Backtest: the SPX Desk’s model of a 7-day, 25-point SPX put credit spread with a stop, priced from VIX9D with fees and slippage, Jun 2, 2009 to Oct 23, 2026; no historical option quotes, so it is a model, not real fills. Bankroll tables: each trade’s result as a share of its max risk, resampled with replacement (4,000 paths of 726 trades), which assumes independent trades. Coin example: exact math, illustrative. Kelly: J. L. Kelly Jr., Bell System Technical Journal, 1956. Option example: Cboe delayed quotes, Oct 30, 2026, bid/ask mids; POP, POT and P50 are model estimates.

FAQ