Bankroll Management Crypto Gambling

The math that beats the math: how much to bet matters more than what to bet on

Every casino game has a published house edge that, over enough wagers, produces a predictable loss. Bankroll management is what determines whether a player rides volatility to occasional wins or runs out of money on a long downswing. The same $1,000 of capital, spread across small bets, can sustain a thousand spins and produce a winning session a meaningful fraction of the time. The same $1,000 staked aggressively in five large bets is nearly certain to be gone within minutes if variance breaks the wrong way. The Kelly criterion, originally derived for positive-edge betting in the 1956 Bell Labs paper by John Kelly Jr, gives a mathematical answer to the question "how much should I bet?" — and for the negative-edge games that dominate casinos, the answer is "less than you think". Combined with practical stop-loss and stop-win rules, Kelly-derived sizing turns a losing-in-expectation activity into one with manageable variance and a chance of finishing sessions ahead. This guide explains Kelly for casino contexts, the stop-loss disciplines that actually preserve bankrolls, the difference between fractional and full Kelly, and the practical defaults that have held up across the academic research from Thorp through 2026.

What bankroll management actually is

Bankroll management is the discipline of sizing bets relative to capital so that variance does not destroy the bankroll before the underlying strategy can play out. For positive-edge activities (advantage poker, blackjack with card counting, advantage sports betting), the goal is maximising long-run growth without ruin. For casino games where the house edge is fixed and negative, the goal is maximising the probability of having a winning session despite the negative expectation — which means accepting smaller average wins in exchange for fewer ruin events.

The two key concepts are unit size and stop levels. Unit size is the base bet expressed as a fraction of bankroll. A typical recreational unit is 0.5% to 2% of total bankroll. Stop levels are pre-committed exit thresholds — stop-loss (exit if bankroll falls below X) and stop-win (exit if bankroll rises above Y). The combination of small units and pre-committed stops is the foundation of every credible casino-bankroll system.

How the Kelly criterion works step by step

1. Compute your edge. Edge = win probability × payout multiplier − loss probability. For a fair coin flip at even money, edge = 0. For Stake Dice with a 1% house edge, the player's edge is negative. The Kelly fraction is meaningful only for positive-edge bets, but the formula provides bounds for any context.
2. Kelly fraction (positive edge). f* = (bp − q) / b, where b is the net odds (decimal payout minus 1), p is win probability, q is loss probability. For a 60% win probability at even money, f* = 0.20 — bet 20% of bankroll per wager.
3. Kelly fraction (negative edge). The formula returns a negative number — Kelly says "do not bet". For casino slots and any house-edge game, the recommended bet size from pure Kelly is zero.
4. Fractional Kelly. Practitioners run "quarter Kelly" or "half Kelly" — 25% or 50% of the full Kelly bet — to trade some growth for substantially less variance. Half Kelly has roughly 75% of the long-run growth of full Kelly with 25% of the bankroll standard deviation.
5. Recreational adaptation. For casino play where Kelly recommends zero, the practical adaptation is "smallest unit size that makes the game enjoyable" — typically 0.5% to 1% of bankroll per spin or hand. This treats casino play as entertainment with a known cost, not as a profit-seeking activity.

Practical examples — three bankroll structures

Conservative recreational, $1,000 bankroll. Unit size 0.5%, so $5 per spin or hand. Stop-loss at 30% drawdown ($300 loss). Stop-win at 50% gain ($500 win). Session length capped at one hour. Maximum 200 spins/hands per session. Expected loss per session at 2% game edge = $20. Probability of hitting stop-win before stop-loss starts around 35%; with discipline to actually stop, that is the per-session win rate.

Moderate, $5,000 bankroll. Unit size 1%, so $50 per spin or hand. Stop-loss 20% ($1,000). Stop-win 40% ($2,000). Sessions limited to two hours or 400 wagers, whichever is shorter. Expected loss per session at 2% edge = $200 (on $10,000 expected total wagers per session). Slightly more aggressive sizing trades smaller average wins for larger variance — useful for high-volatility slot players seeking the occasional big hit.

Aggressive, $20,000 bankroll on Stake Dice 99% RTP. Unit size 2%, $400 per dice roll. Stop-loss 15% ($3,000). Stop-win 30% ($6,000). The 1% house edge on Stake Dice is the thinnest in the industry, so the math is closer to break-even than other games. Expected loss per 100 rolls at $400 per roll = $400. The aggressive sizing only works on Stake Dice or similar 99% RTP games — applying 2% sizing on 4% edge slots would result in faster ruin.

Worked Kelly example for a positive-edge advantage play. A player has identified a sportsbook arbitrage opportunity with 55% win probability at +110 odds (decimal 2.10, so b=1.10). Full Kelly = (1.10 × 0.55 − 0.45) / 1.10 = (0.605 − 0.45) / 1.10 = 0.141, or 14.1% of bankroll. Half Kelly = 7%; quarter Kelly = 3.5%. For a $10,000 bankroll, half Kelly recommends $700 per bet. Real positive-edge opportunities are rare in casino games but common in sports betting on soft books, particularly during live events.

Stop-loss in practice. A pre-committed stop-loss only works if executed without deliberation. The "I will play just one more spin" instinct destroys the discipline. Effective stop-loss requires either logging out at the threshold, transferring the remaining balance to a withdrawal queue, or using the casino's session-loss-limit tool to auto-enforce. Stake, BC.Game, Cloudbet and Shuffle all offer per-session loss limits in their responsible-gambling panels.

The mathematics of session length and ruin

The longer you play, the more your session result converges on the house edge. A 30-spin session at 4% edge on a $1 bet has expected loss $1.20 with standard deviation around $5-15 depending on slot volatility. The probability of finishing up is roughly 45%. A 3,000-spin session at the same parameters has expected loss $120 with standard deviation $50-150. The probability of finishing up drops to roughly 15%.

This is the structural reason that the casino almost always wins over the year while individual sessions go both ways. The path-of-least-resistance for a player who wants the best chance of any individual session being profitable is short sessions with high volatility — accepting that the long-run will be losing in exchange for many sessions where variance carries the result.

Ruin probability over a finite horizon is computable from the formula derived in random-walk theory. For a player with bankroll B betting unit size u on a game with edge e per bet, the probability of ruin before doubling the bankroll is approximately e^(−2eB/u) for small edges. The math says larger unit size, smaller bankroll, and larger edge all accelerate ruin. The lesson: smaller bets, larger bankrolls, lower-edge games — the discipline holds at any stake level.

Common mistakes and red flags

• Sizing bets to recover losses. The Martingale strategy (doubling after losses to recover) has positive expected outcome in pure mathematics but unbounded variance. Hitting the max-bet cap during a losing streak is mathematically certain across enough sessions. Avoid.
• No stop-loss commitment. Without a pre-committed exit, sessions naturally extend until either the stop-win hits or the bankroll runs out. The asymmetry is unfavourable because the house edge accumulates over time.
• Treating winnings as house money. Money in your casino balance is your money. The "house money" framing leads to aggressive re-staking that wipes the gain. Treat the balance after a win the same as the original deposit.
• Chasing a "system". Roulette systems, dice systems, "Aviator predictors" — all fail because the house edge applies regardless of pattern. Any system that promises to overcome a negative-edge game is mathematically false.
• Mixing bankrolls. A poker bankroll, a slot bankroll, and a sports bankroll are three different rolls if you intend to track edge-per-game. Mixing means you cannot apply different unit sizes to different edges.
• Reloading after a stop-loss. A pre-committed stop-loss only works if it is binding. Reloading mid-session converts the stop-loss into a noise filter rather than a discipline.
• Excessive Kelly on negative-edge games. Full Kelly on a positive-edge bet is already higher variance than most players want. Quarter or half Kelly is more common in practice. For negative-edge casino games where Kelly says zero, even small recreational stakes carry steady expected loss.

FAQ

How much should I bet per spin? Conservative: 0.5% of bankroll. Standard: 1%. Aggressive: 2%. For a $1,000 bankroll, that is $5, $10 or $20 respectively. Lower unit sizes preserve bankroll longer; higher sizes are needed if you want occasional large wins.

Does Kelly work for slots? Pure Kelly says zero for any negative-edge game. The practical use is to size positive-edge bets (advantage poker, sports arbitrage, bonus EV) and to bound the question of "how aggressive is too aggressive" for any recreational play.

What is the most important rule? Pre-commit stop-loss and stop-win before the session starts. Execute without deliberation when triggered. Everything else is secondary.

How long should a session be? 30-90 minutes is the practical sweet spot for keeping variance favourable. Beyond 90 minutes, session result starts converging on the house edge.

What about progressive jackpot games? Progressive jackpots have small base RTPs and large tail outcomes. Bankroll mathematically: treat the base game as the relevant edge. The jackpot is a small lottery-style addon that does not change the right unit size for the regular play.

Updated 22 May 2026.