Full-Kelly and you

There seems to be a few misconceptions regarding full-Kelly wagering:

Let’s back up a bit. Last week I began tracking the Massey-Peabody bowl picks:

Team Line Risked To Win Result
Arizona St. 5.5 23161 21055 -23161
Middle Tennessee 6.5 27004 24549 -27004

When I alerted them to this, a spirited Twitter discussion on the merits of full-Kelly wagering broke out. Some highlights:

Cade was swiftly informed that his hypothesis was in fact, false:

Data Niner shared his thoughts from the trenches. Diggs was gracious.

Lou dropped the truth:

More truth from Lou:

Rufus suggested that the magnitude of the disagreement isn’t key, so long as the direction is correct. In short, he just has to be “on the right side.” Plus EV clarified:

Diggs broke things down:

Diggs was on point:

Rufus offered this all-timer:

So, now would be a good time for me to make one point clear:

The problem isn’t the Kelly-Criterion.

PFF isn’t getting their asses handed to them in my tracking due to the Kelly-Criterion:

Current Tab -$786,199.98
% of Bankroll -393.10%
Standard Dev 438076
Two-tailed Z -1.79
ROI -10.93%

Same with Ed Feng:

Current Tab -$161,201.79
% of Bankroll -161.20%
Standard Dev 84577
Two-tailed Z -1.91
ROI -38.18%

The problem is their numbers are absolute garbage. If your internal numbers say you have a 75% edge vs. a highly liquid market, your numbers are wrong. If you think an underdog should be favored by a touchdown, you’re missing something:


Regressing to market will lower the magnitude of the error, but it will not remove its existence. If you don’t want to get your pecker caught in Chesterton’s Fence, it behooves you to figure out what you are missing and incorporate that into your model. In short, if your numbers are regularly largely off market, don’t fucking bet them.

There’s another issue in terms of what you believe in regards to your numbers. There’s nothing wrong with regressing to market. The market is the collective wisdom of profit-seeking agents. It’s not perfect, and it’s not unbeatable. It’s just better than anything else publicly available.

I began my tracking career with the Vampire Squid itself, Goldman Sach 2014 World Cup projections. I opened with this art commission:


That’s “Snorting Bullshit” from Penelope Love.

Don’t believe your own bullshit. Don’t believe anyone else’s bullshit either. Heck, my best gambling advice is very simple: Don’t bet.

But also, don’t blame the Kelly-Criterion. If I think I have a 3.45% edge at even money, I’ll risk 6.9% of my bankroll. If you are going quarter-Kelly, in practice that means you really believe your edge is 0.8625%. Accept that as your internal reality.

TLDR, Norm sums it up perfectly:

Oh, and go Broncos:

Team Line Risked To Win Result
Arizona St. 5.5 23161 21055 -23161
Middle Tennessee 6.5 27004 24549 -27004
Western Michigan 10.5 7836 7392 0
  1. evo34 said:

    Do you bet full Kelly with any of your models? If not, why not?

    • sethburn said:

      Yes. In general, the exceptions would be limit related, with futures or props. That will happen all the time in March.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: