Why Consensus Odds Aren't Always the True Price
Averaging odds across DraftKings, FanDuel, BetMGM, and a few others feels like a reasonable way to find the "fair" probability. It's cleaner than trusting any single book. But consensus has built-in distortions that make it a flawed proxy for true price — and understanding those distortions is what separates sharp line-reading from recreational line-reading.
What Consensus Actually Is
Consensus odds are the average (or median) of available market prices across multiple sportsbooks. If DraftKings has a prop at -120 and FanDuel has it at -115 and BetMGM has it at -125, the "consensus" implied probability is roughly 54.1%.
That sounds like signal averaging — combining multiple estimates to get closer to the true value. But there's a core problem: you're averaging prices set by books whose primary goal is profit margin, not accuracy. Every input to the average is already biased in the same direction.
Problem 1: Hold Percentage Is Baked In Everywhere
Every sportsbook builds a margin into their lines. On a two-sided prop (Over / Under), the implied probabilities on both sides add up to more than 100% — often 105–110%. That gap is the book's take, called the "hold" or "juice."
The extra 7% has to come from somewhere. It comes from you — the bettor. When you average implied probabilities across books, you're averaging numbers that are all systematically inflated. The consensus doesn't cancel the margin out; it inherits it.
A true price is a no-vig price. To get from a book's line to an estimated true probability, you have to remove the juice. Averaging juiced lines just gives you a noisy juiced number, not a fair one.
Problem 2: Books Are Not Independent
The theory behind averaging multiple estimates is that independent sources with different errors will cancel each other out. That works for weather forecasts from different models. It doesn't work as cleanly for sportsbooks.
Most retail sportsbooks — DraftKings, FanDuel, BetMGM, Caesars, PointsBet — all track the same sharp market makers (Pinnacle, Circa, the offshore sharp books) and copy their lines with a delay. When sharp money hits a line at one book, the others follow within minutes. By the time a line appears at all five retail books, it's often the same number repeated five times with slightly different juice on each side.
Averaging those five isn't like averaging five independent opinions. It's more like asking the same person five times and treating each answer as a separate data point.
Problem 3: Public Action Shades Lines Away from True Value
Books don't just respond to sharp money — they also respond to public betting patterns. If 80% of bettors are hammering the Over on a high-profile prop (say, a famous slugger's HR or a star pitcher's K line), books shade that line a few points toward the public to balance their liability. The price drifts away from what the market actually thinks the probability is.
Sharp pricing on a K prop: 52% chance of going Over.
80% of public dollars are on the Over. Book shades to -130 on the Over to limit exposure.
Implied probability at -130: 56.5%. Consensus across five books doing the same thing: also ~56%.
The true edge here is on the Under — but consensus says the Over is the "expected" outcome at 56%.
This effect is strongest on the most popular props — top-tier players on national TV games, props with high public awareness. The more recreational interest a prop attracts, the more the consensus line drifts toward what the public wants to be true rather than what the market thinks is likely.
Problem 4: Stale Books Drag the Average
Not all books update their lines at the same speed. A sharp line move might hit Pinnacle at 10:45am. DraftKings adjusts by 10:47am. BetMGM might not update until 11:15am. If you're computing consensus during that window, you're mixing a current line with a stale one.
Stale lines are almost always stale in the wrong direction — they're the old price that existed before new information moved the market. Including them in a consensus calculation drags the average toward yesterday's truth.
When Consensus Is Useful
None of this means consensus is worthless. It has legitimate uses:
- Sanity checking your model: If your projection says 65% and every book in the market is priced at 50%, that's a signal to check your inputs — not to blindly bet the edge.
- Identifying line shopping opportunities: If one book is 5+ points off consensus, that outlier might offer value — either the book is stale, or it's shading differently than the rest.
- Detecting market-wide moves: If consensus shifts 3 points over 90 minutes, something happened — lineup news, weather, sharp steam. Consensus movement is informative even if the absolute level isn't.
What to Use Instead
The closest thing to a true market price comes from no-vig markets. These are books or exchanges that charge no margin — they make money on volume, not on inflated lines. The price they show is a cleaner reflection of what the market actually thinks the probability is.
Novig is the most accessible no-vig reference for MLB props in the current US market. When Novig prices a K prop at -105 / -105, the implied probability on each side is 51.2% — and you can trust that number as an estimate of true value far more than a -115 / -115 retail consensus.
PropPrizm's approach: Novig prices are shown as a book column and factored into edge calculations, but they're explicitly excluded from consensus computations. Averaging a no-vig price with juiced retail prices would corrupt both signals. The no-vig reference stands alone as the best available market estimate of true probability.
Reading the Edge Scanner with This in Mind
When you look at a prop in the PropPrizm edge scanner, the edge percentage is calculated against the best available book price — not the consensus. The model's projected probability is compared to what you can actually bet at, not a theoretical average.
A few things to look for:
- If the best book price matches Novig closely, the retail market is fairly priced and you're getting a real read on true edge.
- If the best book is significantly better than the rest of the market, you may be seeing a stale price or a shaded line — worth acting on quickly before it corrects.
- If the model and Novig agree but retail books are significantly lower (Over is cheaper at retail), the sharp market has already found the edge and retail books haven't caught up yet. That's often the best entry window.
Understanding the difference between a consensus price and a true price won't change what prop you're betting — but it will change how you evaluate whether the edge is real. The Edge Scanner is built around this distinction: model probability vs. the actual price you can get, not the average of what everyone else is offering.