Why Shitty Models Are Great

Whenever I talk to people about my betting, the first question is usually something like: “Oh wow, how good is your algorithm?” Most people—especially those without much experience building models—don’t realize just how limited models are. This is especially true in sports betting, where small-data problems are everywhere.

Plenty of smart people have built public-facing prediction models for ESPN, The Athletic, and even Nate Silver for the NBA. Some are better than others, but I’m not aware of any that consistently beat the market in any sport. These models often get dunked on by Gambling Twitter—I’ve probably seen a thousand variations of the same joke: an ESPN FPI prediction side-by-side with ESPN Bet odds, where the win percentages are wildly different.

That criticism is fair. If you’re building a prediction model, you’d be way better off just using the vig-less moneyline percentage. But that doesn’t mean the people building these models are dumb. Some maybe, but Nate Silver is certainly a way better modeler than I am—yet I’d wipe the floor with any of the public-facing betting models he’s built. Why? Because building a model that actually wins in sports is very hard—especially without consistent subjective adjustments and being on top of all the information flow.

Models Still Matter

This isn’t to say modeling isn’t important in betting—it absolutely is. In every sport I bet, I have a model spitting out numbers. But I’m very rarely just blind-betting those numbers into the market.

A few sports betting friends of mine wanted to beat NCAA basketball. Like most new originators, they were obsessed with building the best model possible and thought that was the only way to win. Both are smart guys, but not master modelers. When they asked me for advice, I told them they were solving the wrong problem.

They weren’t going to build an NCAA model more sophisticated than what the market already has. Instead, I told them to make sure their model was functional, but spend most of their time sharpening everything else—information flow, player and team knowledge, recruiting, the transfer portal, etc. That’s how they’ll win in that market, not by trying to out-model Rufus.

Why Shitty Models Are Great

“Shitty” models are great for a ton of reasons. Obviously, if your model is too shitty—if every line is miles off market and you’re showing 70% edges everywhere—it’s not helpful. But a reasonable model, even if it doesn’t beat the market, can still be incredibly useful.

1. Baselines

Models give you baselines, which are critical in betting. Say you’re watching an NFL game and Lamar Jackson sprains his ankle. You pull up FanDuel and his rushing prop is 32.5 (-110). Without a baseline, that number is meaningless. Maybe Fanduel trader moved the line from 50.5 to 32.5, hoping some fish will run to it trying to bet the under, its also possible they missed it entirely and that was his line if he was healthy.

If you have a shitty model, and it spits out 35 as fair, you’re in a much better spot. Normally you wouldn’t touch 32.5. But now you know Lamar’s hurt, so you can confidently hit the under, knowing the adjustment wasn’t big enough.

2. Soft Markets

Another great example is smaller, less hard-to-beat markets. Sure, maybe your NBA spread model isn’t beating the market—that’s a tough challenge. But if it’s decent enough, it can absolutely help you win in spots like season win totals, “to make the playoffs,” or division bets. You can make a lot of money in those markets while ignoring the tougher mainlines entirely.

Of course, you have to use this in conjunction with your brain. You don’t want to bet a team’s win total over just because your model rates them highly without considering they might tank, or trade their best player at the deadline. But the beauty of sports betting is you don’t have to bet anything—yet everything is available to you. The model helps you pick off the spots that are obviously good, and pass on anything that’s close. That way, you’re ready when the opportunity presents itself.

3. “System” Bets

Betting “systems” are another one that gets ripped apart on Gambling Twitter. Again, somewhat fairly. But the truth is, plenty of these angles have been profitable at different points. A classic example is NBA playoff teams going down 0–2 on the road—they’ve historically been much more likely to win that third game. The frustrating part is how these bets often get presented with no context. The market now knows how important that swing game is, and it’s priced in.

I just saw a tweet saying: “0–2 NFL teams entering Week 3 that are priced as an underdog are 38–17 ATS over the last 9 years.” That sounds good, but in reality, most of these are just random start and end points, or small samples dressed up as meaningful trends. If you actually think you’ve found something real, you have no way to capitalize without context. That’s where a shitty model helps—it lets you see if the market is already pricing this in, and if not, gives you the framework to decide whether it’s worth betting.

The Takeaway

If you want to be the best bettor in the world, yeah—you’ll probably need an extremely advanced model. But if you want to scoop up the low-hanging fruit and win a lot of money without battling the sharpest markets every day?

Then shitty models are awesome.

🟢 Check out Kalshi and get signed up today: https://kalshi.onelink.me/1r91/hammer

🔵⚪️ New to FanDuel Sportsbook? Bet $5 and Get $200 in Bonus Bets if your bet wins. Download the FanDuel Sportsbook App or check it out at http://fanduel.com/forwardprogress to get in on the action.

Must be 21+ and present in select states or 18+ and present in DC. Opt in req. Bonus issued as non-withdrawable profit boost tokens. Restrictions apply, including any token expiration and max wager amount. See terms at sportsbook.fanduel.com. Gambling Problem? Call 1-800-GAMBLER or visit rg-help.com.

Trusted by PROs. Edges in Seconds.

Markets move fast. Betstamp PRO moves faster — surfacing real-time edges from the outs that matter most. Trusted by sharp bettors, PRO helps you act quickly, save time, and scale your edge.

Hammer Exclusive: Unlock your free 14-day trial of the Betstamp PRO Main Market Odds Screen → https://betstamp.com/trial

Betting props? Apply for a Betstamp PRO props screen demo  → https://betstamp.com/hammer

Every bettor talks themselves into picks. FTN helps you talk yourself out of bad ones. The platform gives you unbiased team and player context so you are not chasing last week or falling for hype. Start with Stats Pro or upgrade to Stats GOAT. Use HAMMER10 for 10% off. Head over to FTNFantasy.com.