The Bankroll Starter Kit: Unit Sizing and Quarter Kelly Without the Math Degree
Hey, it’s Turtle. Quick question before your next bet: what is your unit size? Not what felt right last night. The number you decided in advance. If you do not have one, you are in the majority. Almost every bettor we talk to is guessing their stakes, and stake guessing is how people with decent reads still lose money. This post is the whole fix: one lookup table, one cheat sheet, three rules. Ten minutes, set for the season.
What you’ll learn in this post:
- The unit-size lookup table: the exact stake for a $200, $500, $1,000, or $10,000 bankroll
- Quarter Kelly in plain words: the formula once, a worked example at -110, and why sharps bet a quarter of it
- The honest drawdown table: the losing streaks and drawdowns a genuinely good bettor should still expect over 100 to 1,000 bets
- The 3 tilt rules that protect the bankroll from its owner
- Who honestly does not need any of this, and the 4 mistakes that undo it
1. Step zero: ring-fence the money
A bankroll is not "whatever is in my betting apps right now." It is a number you pick on purpose, with three properties:
- It is separate. One place, countable in ten seconds. If you cannot say what your bankroll is, you do not have one yet.
- It is allowed to reach zero. Not "would sting." Allowed. If losing this money would touch your bills or your housing, the correct bankroll is zero and the honest advice is do not bet at all. No unit table makes that money safe to stake. That is a stop sign, not a sizing problem.
- Pending bets are spent. Money on open slips is not balance. Your bankroll is settled cash only.
2. The unit table (copy-paste this)
One unit is your default stake, and the boring, battle-tested answer is 1% to 2% of your bankroll:
| Bankroll | Careful unit (1%) | Standard unit (2%) |
|---|---|---|
| $200 | $2 | $4 |
| $500 | $5 | $10 |
| $1,000 | $10 | $20 |
| $10,000 | $100 | $200 |
Any other size: divide your bankroll by 100 for the careful unit, by 50 for the standard one. That is the entire system. Flat one-unit stakes on everything until you have a strong, specific reason to vary.
Two reactions people have here. "That’s tiny." Yes, and section 4 shows exactly why. "Which column am I?" If this is your first season running a real bankroll, take the careful column. Moving up later is easy. Moving down usually happens for sad reasons.
3. Quarter Kelly without the math degree
The Kelly formula gets treated like a physics exam. Here it is once, in plain words:
Your edge, divided by what the book pays per dollar of stake. That fraction of your bankroll is the full Kelly bet.
Worked example, the whole thing:
- A bet at -110 pays about 91 cents of profit for every dollar staked.
- Say you genuinely believe the bet wins 55 times out of 100. Breakeven at -110 is 52.4%, so that belief is a real edge.
- The edge: 55 wins pay about 0.91 each, 45 losses cost 1 each, so per 100 bets you clear about 5 units. Your edge is about 5 cents per dollar.
- Divide that edge by the 91-cent payout: full Kelly says stake about $5.50 of every $100 in your bankroll.
- Quarter Kelly: about $1.40 of every $100. Notice where that lands: right inside the 1% to 2% band from the table above. The unit table is quarter Kelly with the math pre-done for a realistic edge.
Why bet a quarter of the "optimal" number? Because the formula’s input is a guess. Kelly assumes you actually know your win probability, and you do not, you estimated it. If the true number is 52.4 (breakeven) and you size like it is 55, full Kelly has you staking several times too much, forever. Overbetting an overestimated edge is how bettors with genuinely good reads still go broke. Quarter Kelly gives back some theoretical growth in exchange for being very hard to ruin with an honest estimation error. That trade is why the sharps you respect size closer to a quarter than to full.
Our free Kelly calculator does this for you: enter your win probability and the odds, and it has Full, Half, and Quarter buttons plus a dollar output. One more rule from it worth tattooing somewhere: if Kelly outputs zero or negative, you have no edge, and the correct stake is nothing.
4. The honest drawdown table (read this before you size up)
Here is the section nobody selling picks wants you to read. Take a bettor who truly wins 55 bets out of every 100 at -110, long run. That is a genuinely strong rate (breakeven is 52.4%). Here is what that winning record still does along the way, simulated over two hundred thousand runs at flat one-unit stakes (full math in the master file’s appendix):
| Bets placed | Longest losing streak to expect | 1 in 10 runs see | Typical worst drawdown | 1 in 10 drawdown |
|---|---|---|---|---|
| 100 | 5 in a row | 7 or more | 8 units | 15 units |
| 250 | 6 in a row | 8 or more | 13 units | 22 units |
| 500 | 7 in a row | 9 or more | 17 units | 27 units |
| 1,000 | 8 in a row | 10 or more | 21 units | 34 units |
And the row that stings most is not in the table: this same good bettor is still down money after 100 bets about 3 times in 10. After 1,000 bets it is closer to 1 in 20. None of that means a broken model or a cold gut. That is what winning slowly looks like while it is happening.
Now read those drawdowns against the unit table:
- At 1% units you hold 100 units. Nothing in that table threatens you. Even the ugly 1 in 10 run over 1,000 bets costs about a third of the roll, and you keep playing.
- At 2% units you hold 50 units. The bad runs hurt. You survive all of them.
- At 5% units you hold 20 units. The TYPICAL worst drawdown over 500 bets nearly ends you, and the 1 in 10 version does end you. While winning.
That is the entire argument for small units. The table in section 2 is not timid, it is sized to the variance a real edge actually carries.
5. The 3 tilt rules
The table sets the size. These three rules keep it:
- The unit changes on a schedule, never after a night. Recompute it from your current bankroll on the first of the month. Won a lot? It grows on the first. Lost a lot? It shrinks on the first. No same-day resizing in either direction.
- Down 3 units in a day means the day is over. Not "one more to get it back." Over. Section 4 says losing nights are a certainty, so the job is making sure any single night stays a data point instead of a crater.
- The stake never doubles to chase. Doubling after losses feels like conviction and is mathematically the fastest route from a bad night to a dead bankroll. If a bet only makes sense at double size, it did not make sense.
6. Who should skip this (honestly)
- You bet a few times a year, for fun, with money you would have spent on a night out anyway. A formal bankroll is optional. Enjoy the games.
- Losing the money would touch bills or housing. Then this is not a sizing question, and no table fixes it. The honest answer is do not bet, and if stopping is hard, 1-800-GAMBLER is free and confidential.
- You already run a staking system that survived a real losing month. Keep it. Switching systems mid-variance is its own kind of tilt.
- You only play DFS pick’em apps. Sizing still applies (an entry is a stake, use the same table), but the odds math above matters less; your multipliers are fixed.
7. The elephant in the room
Bankroll discipline means smaller bets than you want to place. Every table above points the same direction: down. We know that is not the exciting answer, so here is our incentive, stated plainly: a bettor sized small enough to survive variance is a bettor who is still around in two years, and a subscription business makes its money from people who stick around, not from people who flame out in a month. Our incentives point at your solvency. Take that for exactly what it is.
8. The 4 mistakes that quietly kill bankrolls
- Unit creep after wins. Three green nights and the $10 unit is quietly $25. That is emotional resizing with extra steps; rule 1 exists for this.
- Doubling after losses. Chasing converts a normal, survivable losing streak (section 4: they are a certainty) into ruin at record speed.
- Counting pending bets as balance. Open slips are not money. Size from settled cash, or you are quietly betting more than your system says.
- Full Kelly on an estimated edge. Kelly’s output is only as good as the probability you fed it, and yours is a guess. Full Kelly on a guess is overbetting with a formula for cover. Quarter it.
9. Why are we giving this away?
Fair question, so here is the honest answer. We sell a board that finds mispriced lines across 30+ books, next to a model whose calls are graded in public, 84,000+ of them and counting as of July 14, 2026 (check the live count), wins and losses both. Bankroll discipline does not require our product. But only people who last long enough for the math to play out can get anything from what we sell, so a solvent, correctly-sized bettor is exactly who we want reading us. The kit above works with or without us. We would rather you bet small and stick around.
Your one next step
You now have a unit. If you also want the math delivered every morning: The Daily Receipt. Yesterday’s graded record (losses included) plus the boldest calls graded, win or lose, across 30+ books, free, in your inbox at 9am ET.
Data note: the Kelly and drawdown figures on this page are teaching math, computed at authoring with a seeded two-hundred-thousand-path simulation at flat one-unit stakes and -110 pricing. They describe variance mechanics, not our model’s performance.
21+. This is analytics and education, not betting advice. Past results don’t guarantee future results. If betting stops being fun, step away: 1-800-GAMBLER.
Related reading
Related Posts
The 10-question capper audit with what fake answers sound like, the napkin math that breaks the classic 7-of-10 win claim, the 2-minute version for skimmers, and the 5 mistakes that undo an audit.
Measured weekly from live boards: this week's hold ladder ranking every book by what it actually charges, the 8 copy-paste routines that stop the leak, who should honestly skip line shopping, and the 5 mistakes that undo the habit.
The exact math behind parlay odds, why standard parlays carry compounded vig, when DFS-style and SGP parlays make sense, and how to verify any ticket.
Want picks like these?
Turtle +EV scans thousands of player props every day across 30+ sportsbooks and surfaces only the +EV opportunities backed by our sport-specific probability models.