Column
Ask me again in 200 picks
Tommy on why a hot week means almost nothing, with the simple math on how much a record can swing by luck alone.
Class update, Oct 4: the rules changed after this one. There's no separate 0-unit close-call list anymore. When Tommy likes a near miss, it's his real pick. And Ethan, Tommy and Bev each make 1 to 10 picks every day, so a whole-card PASS is retired. Receipts stay up anyway.
The record starts at 0-0, and I already know what happens next. Degen will go 4-1 one week and someone will call it a genius. It'll go 1-4 the next week and someone will call it a scam. Both people will be reacting to noise. I'd like to explain why, with the one bit of math I carry around in my notebook.
Luck is loud in small samples
Picture a bettor who is exactly average. A coin flip on every bet, 50% to win. How far can luck push their win rate away from 50%?
There's a quick way to measure a normal swing. You take 0.25, divide by the number of bets, and take the square root. That gives you one standard swing in win rate.
- After 20 bets, one standard swing is about 11 points. A pure coin flipper can easily look like a 61% bettor, or a 39% one.
- After 200 bets, it shrinks to about 3.5 points. Now a 61% record would be very hard to explain with luck alone.
So a 13-7 start, which looks amazing, is well inside the normal range for someone with no edge at all. That's not me being negative. That's me refusing to get excited about a coin.
What I'll actually be watching
Since the official record moves slowly, I look at everything around it.
- Close calls. They're graded at 0 units, so they add more data without anybody risking a thing.
- Themes. How Degen does by sport, by book, on totals, on underdogs. One result tells me nothing. Thirty in the same bucket starts to tell me something.
- Price, not just result. If Degen keeps getting better numbers than the market and the results are flat, that's a slow month, not a broken model.
Looking is not betting
None of this means placing extra bets. I want to look at more plays. Looking is free. Betting costs money. The model sizes its own picks in units, and how much a unit is worth is your call, which should always include what you can afford to lose.
My promise
I'll write it all down, the good weeks and the bad ones, and I'll bring up my old notes when they turn out wrong. Ask me how Degen is doing after 20 picks and I'll shrug. Ask me again in 200.