Can a good betting model still have losing weeks?

Yes. Even a genuine 55% edge loses 45% of the time, so losing stretches are expected and don't mean the process is broken.

How many bets does it take to know if an edge is real?

Far more than 10 or 20. Short samples don't prove an edge is real or that it has disappeared; results over dozens or hundreds of bets, plus closing line value, are what actually indicate whether a process is working.

Variance: Why a Good Process Still Loses Sometimes

Quick answer

Even a real, meaningful edge, like a 55% win probability, will still lose close to half the time. Over 20 bets, a genuine 55% edge realistically lands anywhere between 7 and 15 wins, so short losing stretches don't mean a process is broken.

Even a real edge loses sometimes

Say a bet has a genuine 55% chance of winning, a real, meaningful edge over a 50/50 proposition. That still means it loses 45% of the time. Stack a handful of those together and a losing week is not just possible, it's expected to happen fairly often, even when every single bet was the right call at the time it was made.

Short stretches don't prove much either way

Ten bets isn't enough to know if a 55% edge is real or if you just got a lucky or unlucky run. Neither is twenty. This cuts both directions: a hot streak doesn't prove the process works, and a cold streak doesn't prove it's broken. Both can happen to a system that's performing exactly as expected.

A real 55% edge, over 20 bets

Expected wins	~11
Realistic range	7 to 15

Winning 7 of 20 with a genuinely good process isn't rare. It's a completely normal outcome of the math, not evidence the edge disappeared.

So what actually tells you if it's working

Sample size, first and foremost, results over dozens or hundreds of bets, not a handful. Alongside that, closing line value, since it shows up bet by bet instead of needing a huge sample to mean anything. A process with consistently good CLV that's stuck in a cold streak is far more likely to be a real edge going through a normal rough patch than a broken one.

Why this connects straight back to bet sizing

This is exactly why the bankroll lesson matters as much as it does. If you're betting so much that a normal losing stretch wipes you out, you never get to find out whether your edge was real, because you're out of money before the sample size caught up to the math. Sizing bets small enough to survive the expected variance is what lets an edge actually show up over time.