Drawdown survival — what a normal losing streak looks like
Most bettors who quit do not quit because their edge died. They quit during a losing run the math said was coming. This session puts real numbers on what "normal" looks like, shows how deep the hole gets at different Kelly fractions, and gives you the test that separates a drawdown from a dead edge.
What this session covers
The drawdown profile of a genuinely winning bettor: how long streaks run, how deep the peak-to-trough hole gets, and how much of that depends on stake fraction rather than skill. Then the diagnostic — CLV holding while results lag means variance; CLV degrading means the edge is gone. Then the survival protocol.
Reference bettor throughout: 55% at −110, an edge of 5.0% per unit staked. Season length 500 bets.
The streak table
Losses arrive in clumps. For a 55% bettor, each bet loses 45% of the time, and over 500 bets the probability that somewhere in the season you hit a run of at least N straight losses is:
| Consecutive losses | Chance it happens in a 500-bet season |
|---|---|
| 5 in a row | 99.5% |
| 6 in a row | 90.3% |
| 7 in a row | 64.4% |
| 8 in a row | 36.9% |
| 9 in a row | 18.6% |
| 10 in a row | 8.8% |
Read the middle of that table again. A seven-bet losing streak is more likely than not for one of the best bettors you will ever meet. The single most likely value for the longest streak of the season is exactly 7 (27.5% of seasons), with 6 behind it (25.9%) and 8 at 18.3%. Halving the sample barely helps: over 250 bets, a run of 6 still shows up 68.6% of the time.
The honest baseline: if you bet a full season and never lose seven in a row, you got lucky. The streak is not a signal. It is the price of admission.
How deep the hole gets
Streak length is what people notice. Peak-to-trough drawdown is what ends careers, and it is driven almost entirely by your stake fraction.
For our 55% bettor at −110, full Kelly is 5.5% of bankroll per bet, half-Kelly 2.75%, quarter-Kelly 1.375%. Across 200,000 simulated 500-bet seasons, the worst drawdown in a season looks like this:
| Fraction | Median worst drawdown | 90th percentile | Chance it exceeds 50% | Season ends below start |
|---|---|---|---|---|
| Full Kelly (5.5%) | 66% | 84% | 88% | 28% |
| Half-Kelly (2.75%) | 39% | 56% | 20% | 20% |
| Quarter-Kelly (1.375%) | 21% | 33% | 0.3% | 15% |
| Flat 1% | 16% | 25% | under 0.1% | 15% |
Same edge, same bettor, same season. The only variable is stake size, and it moves the typical worst moment of the year from down 16% to down two thirds.
Two things fall out of that table. First, full Kelly here is 5.5% of bankroll — it breaches Sharp's 5% bankroll cap before you place a single bet. The cap exists because Kelly's math, taken literally, hands you stakes no human sits through. Second, half-Kelly still shows a 39% median drawdown: the recommended default, correctly executed, on a winning edge.
The recovery asymmetry
Drawdowns are not symmetric, which is why stake discipline is not optional. Down 20% requires +25% to reach the old peak. Down 39% requires +63%. Down 50% requires +100%. Down 66% — the full-Kelly median — requires +192%.
Full Kelly does produce more growth in the median case, roughly 2.0× versus 1.7× for half-Kelly over 500 bets. It also demands you sit through a two-thirds drawdown without touching anything. Half-Kelly captures about 75% of that long-run growth at roughly half the volatility, which is why it is the house default.
Drawdown or broken edge?
Results cannot answer this. A 55% bettor and a 50% bettor look identical over 100 bets. The test is closing line value.
Variance looks like this: average CLV holding at its usual level, still beating the close on a majority of bets, losses spread across markets rather than concentrated in one. Your prices are good; the outcomes have not caught up. Nothing to fix — keep sizing exactly as before.
A dead edge looks like this: average CLV drifting toward zero or negative, or collapsing in one market or at one book while holding elsewhere. That is not bad luck, it is the market pricing in the thing you were pricing first. Stop betting that market rather than betting it harder. A bet log with a closing-line column and a market tag builds the scorecard for you, so the diagnosis takes minutes.
The survival protocol
- Pre-commit the sizing before the season, in writing. Fraction, cap, and maximum bets per day. A rule set while calm is the only rule that survives a bad week.
- Hold the 5% cap absolutely. No single bet, ever, regardless of what the model says. It is what makes a bad stretch survivable instead of terminal.
- Never increase stakes to recover. Moving from half-Kelly to full Kelly takes your median worst drawdown from 39% to 66% and raises the chance of ending the season below where you started from 20% to 28%. Sizing up during a drawdown does not shorten it. It deepens it.
- If you change anything, change it downward. Reducing to quarter-Kelly cuts the median worst drawdown to 21% for a little growth. That is legitimate. Sizing up is not.
- Cut bet count before you cut standards, and watch yourself as well as the bankroll. If a drawdown is costing you sleep, or you are reaching for money earmarked for something else, stop and step back rather than re-size. That one is not a math problem.
What the live session adds
- Live simulation on screen: 500-bet seasons at each Kelly fraction, equity curves crawling, so the drawdown table stops being abstract.
- A worked CLV diagnosis on an anonymized member ledger — a real bad month, diagnosed as variance, and what happened next.
- Q&A on the stretch members are currently in, including how to set a reduce-size trigger that is not a panic button.
- A walkthrough of the daily bet-count cap, weekly bankroll cap, per-bet maximum, and cool-down timer, so the protocol is enforced by software rather than willpower.
Session status: Replays post to this page after each live run. The next live date is announced in Discord and in the weekly Sharp Report.