The Lab · Foundations · updated 2026-07-28
Is it better to bet favorites or underdogs?
Neither. At fair odds both break even exactly, and at real sportsbook prices both lose at the same rate. Here's the math with $100 bets, and what actually creates an edge.
It's the first question almost everyone asks, and it has a clean answer that surprises people: neither is better. Not slightly. Not on average. The choice between favorites and underdogs is not where money is won or lost.
Here's the proof, and then the part that actually matters.
Ten bets, $100 each — run it both ways
Take two bettors. One only backs 40% underdogs at +150. The other only backs 60% favorites at −150. Both are exactly right about the probabilities, and both bet $100 ten times.
The underdog bettor wins 4 of 10. Each win pays $150 profit:
4 wins × $150 = $600 — 6 losses × $100 = $600 → net $0
The favorite bettor wins 6 of 10. Each win pays $66.67 profit:
6 wins × $66.67 = $400 — 4 losses × $100 = $400 → net $0
Both start with $1,000 and finish with $1,000.
That isn't a coincidence or a rigged example. It's what "fair odds" means. Bigger payouts arrive less often; smaller payouts arrive more often; they cancel precisely.
What win rate do you need to break even?
The reason they cancel is that the price already contains the probability. A moneyline is not a prediction — it's a break-even threshold, and it is exactly the implied probability:
| Price | Implied probability | Win rate you need |
|---|---|---|
| +220 | 31.2% | 31.2% |
| +150 | 40.0% | 40.0% |
| +130 | 43.5% | 43.5% |
| −110 | 52.4% | 52.4% |
| −150 | 60.0% | 60.0% |
| −220 | 68.8% | 68.8% |
Those last two columns are identical at every row, and that identity is the whole answer. A 31% win rate is excellent at +220 and catastrophic at −220. A win rate means nothing until you say what price it came at.
This is why "I hit 60% of my bets" is not a brag. At −150 it's break-even. If you're mostly betting favorites, 60% is exactly par.
Now add the vig — and watch both sides lose equally
Real sportsbooks don't offer fair odds. A typical baseball game is priced something like −150 favorite / +130 underdog. Add the two implied probabilities:
60.0% + 43.5% = 103.5%
That extra 3.5% is the house fee, baked into both sides. Strip it out and the honest probabilities are 58.0% / 42.0% — but you're charged as if they were 60% and 43.5%.
So what happens if you bet each side at its true rate?
- Backing the favorite at −150, winning 58%: −3.4% per bet
- Backing the underdog at +130, winning 42%: −3.4% per bet
Identical, to two decimal places. The fee is spread so evenly that there is no safer side of the ticket. Both bettors lose at the same speed; the favorite bettor just loses in smaller, more frequent increments, which feels better and isn't.
(If the fee itself is new to you, the vig covers how it works and what it costs over a season.)
Why favorites feel safer anyway
Because winning often and winning money are different experiences, and only one of them is visible night to night.
Backing favorites means most tickets cash. You feel competent. The losses are rarer, larger, and easy to file away as bad luck. Backing underdogs means losing most nights and waiting on the occasional big hit — which feels like failure even when the ledger is identical.
Our own board shows the trap plainly: across 311 graded games, the market's favorite won 54.7% of the time. More than half — and still a losing bet at the prices those favorites carried. We wrote about why that 54% number is what it is separately.
So where does an edge actually come from?
Exactly one place: your probability being closer to the truth than the price is.
If a team's real chance is 45% and the market prices it at +150 (implying 40%), that bet makes money — not because it's an underdog, but because 45% beats 40%. The same logic on a favorite at −150 requires you to believe better than 60%.
Which reframes the original question. It was never favorites or underdogs. It's where is the price wrong, and the honest answer on most games is nowhere, because the market is very good and the fee is charged on top of it. That's why our model publishes a bet on a small fraction of games and stays quiet on the rest.
What this doesn't tell you
The examples assume you know the true probability. Nobody does — you estimate it, and your estimate carries error that easily exceeds the edge you're chasing.
The −3.4% figure comes from one representative price. Hold varies by book, sport and market; shopping for the best number across books is one of the few reliably profitable habits available to a normal bettor, and it works on either side of the ticket.
And breaking even at fair odds is the ceiling of betting without an edge, not the expectation. Real bettors also face limits, timing, and the temptation to chase — none of which appear in a ten-bet table.
Model output is informational and entertainment content, not betting or financial advice. If you bet, bet what you can afford to lose.