The Week I Ran Eighth Kelly and It Still Felt Terrifying
Everyone talks about Kelly sizing until they hit a twelve-bet losing streak. Here's what it actually feels like, and why I didn't deviate from the plan.
David Koval
Senior Advantage Betting Contributor
9 min read·Published 18 Jan 2026
The eighth day of the losing streak I was lying awake at 2am running numbers on my phone trying to figure out whether my entire process was broken. I'd gone 19 bets without a win. My bankroll was down 22%. My average closing line value over those 19 bets was +3.4%, which is to say I was genuinely picking the right side of the market and the market was picking the right side of the result, and we were both losing to variance.
I was betting eighth Kelly at the time. A quarter of quarter Kelly. For context, eighth Kelly on a +4% edge puts about 0.5% of bankroll on a typical bet, which is genuinely small. I had deliberately under-sized to give myself emotional headroom during drawdowns, and even at that size, the drawdown was unbearable enough that I was questioning my process at 2am on a Tuesday.
This piece is for anyone who's read about Kelly sizing and thinks they know how they'll behave during variance. You probably don't. I didn't. Here's what actually happens.
The sizing that was supposed to protect me
A bit of context. I'd been betting advantage markets for about eight months when this hit. My tracked average edge was +3.1%. My CLV had stabilised around +2.7% after a volatile first quarter. My bankroll was $18,000, built up from an initial $12,000 plus profits.
I'd read every word I could find on bankroll management before I started. I understood that full Kelly was theoretically optimal but practically a nightmare. I'd landed on quarter Kelly as my operating size, which for a +3% edge puts roughly 1.4% of bankroll on a typical bet. At $18,000 bankroll, that's about $250 per bet. Comfortable.
Two months before the drawdown I'd voluntarily dropped to eighth Kelly because I'd had a hot streak and the bankroll was up about 8%. Stepping down felt paranoid at the time - things were going well, why would I reduce stakes? - but I did it because the bankroll guide I was following suggested it as a way to pre-commit to lower sizing before a drawdown hit. Smart-past-me, as it turned out.
The streak
The streak started on a Tuesday. Lost a disposal prop on an AFL game by one disposal. Lost a run-metres prop on NRL by four metres. Lost an A-League H2H that went to a penalty shootout at 89 minutes. Individual bad beats, all normal.
By Friday I was 0-7. I remember thinking this was unusual but probably fine. Variance runs happen.
Saturday was 0-4. Sunday was 0-3. I was 0-14 entering Monday and my internal monologue had shifted from "this is variance" to "is it variance, though? Am I sure my edge model isn't broken?"
Monday lost two more. Tuesday lost three. I was 0-19 entering Wednesday, down about $3,900 off the peak, and spending evenings re-running my CLV calculations on every losing bet trying to find the hidden process failure.
There wasn't one. Every bet had positive CLV. The market was moving in my direction on nearly every placement. The games were just landing on the 48% side of my 52% bets, over and over.
The maths says this is normal
A 19-bet losing streak at a 52% win rate has a probability of (0.48)^19 = 0.00014, or about 1 in 7,000. Sounds rare. Sounds like something you'd never hit.
Except you're not placing one sequence of 19 bets. You're placing hundreds of bets across a year, and any 19 consecutive bets within that sample can form a streak. The probability of hitting at least one 19-bet losing streak within 500 bets at 52% is substantial - something like 20%.
In other words, if you bet 500+ times a year at a modest edge, you should expect to see a 19-bet losing streak. It's not a rare event. It's a normal feature of the variance landscape. Knowing this did not help me at 2am on day eight.
What kept me in the seat
Three things, in order of how much they mattered.
The CLV column. Every losing bet had positive CLV. I was picking the right side of the market and getting beaten by variance. The alternative hypothesis - that my model was broken - would have shown up as negative CLV, and it wasn't showing up. I trusted the CLV number more than I trusted my 2am gut feeling, and the CLV number kept telling me to keep going.
The eighth Kelly sizing. If I'd been on quarter Kelly or half Kelly, the 22% bankroll drop would have been 44% or 88%, and there is no world in which I hold discipline at 88% down. Eighth Kelly was small enough that the pain was survivable. The reason you pre-commit to small sizing before variance hits is specifically so that when variance hits, you can actually sit through it.
The pre-written rule. I'd written down six months earlier what I would do at -20%, -30%, -50%. The -20% rule was "continue sizing as planned, audit bets for process errors, do not change strategy." At -22% I had a rule to follow, which removed the decision. Making the decision while I was down was not something I was capable of doing well, and the pre-written rule protected me from having to.
The recovery
Day 10 of the streak I hit a three-bet winning run. Day 11 was 2-1. Day 12 was 4-0. Over the following three weeks I went 28-12, which at eighth Kelly sizing meant the bankroll recovered to roughly flat against the prior peak.
The recovery didn't feel triumphant. It felt like the variance had just stopped punishing me. Which is exactly what it was. The picks in the recovery weren't better than the picks in the streak. The market moves weren't more favourable. The dice just started landing on the 52% side for a while.
The psychological effect was significant though. Sitting through a 0-19 run and coming out the other side taught me more about process discipline than any amount of reading did. I trust my CLV now in a way I didn't before, because I've seen it correctly predict recovery when I was panicked.
What I'd tell myself on day 8
Three things, if I could go back to 2am that Tuesday.
First, the 2am spreadsheet audit is not going to find a bug in your process. If your CLV is positive, there is no bug. The losing streak is weather, not climate. Close the laptop.
Second, the eighth Kelly sizing is working exactly as designed. The reason you dropped stakes when the bankroll was at peak was for this moment. You haven't been punished by it - you've been protected by it. The person deciding to bet quarter Kelly through a streak like this is not a more disciplined version of you. They're someone who hasn't had this experience yet.
Third, 19 bets is nothing. You will place thousands of bets across the life of this operation. A stretch of 19 that lands wrong is six days of your year. Stop treating it as a statement about your future.
For the Kelly maths and fractional-Kelly logic in detail, the bankroll guide walks through it carefully. For why raw win-rate deceives you in the middle of a streak like this, the CLV guide is the read. And if you've just started advantage betting and haven't yet hit your first real drawdown, don't assume it won't come. It will. Pre-commit to the rules you'll follow when it does.
About the author
David Koval
Senior Advantage Betting Contributor
David has been running advantage betting strategies across Australian bookmakers since 2023 and contributes long-form retrospectives, case studies, and operational pieces drawn from years of running real bets in AU markets. His writing focuses on the realities of running a sustainable AU advantage operation — what works, what fails, and the operational details most blogs gloss over.
Story11 min
How I Made $11,400 Arbing in My First Six Months (And Lost Most of It to Gubbings)
Fourteen bookmaker accounts, 847 arbs, one very expensive lesson about account longevity. The honest numbers and what they taught me.
