MLB Parlay Prices

The best-price ledger

Ranking prices is easy to do and easy to fool yourself with, so the top three combinations from the list above are written down every day before the games, kept in a log that is never edited, and graded on the finals. The three are chosen so that no two of them share a game, which makes them three separate results rather than one result counted three times. The record below prints whatever it says, including the stretches where it says nothing good.

How the price comparison is made

Every book that posts both sides of the same number on a bet gives us one estimate of that bet's probability: take the two prices, convert them to probabilities, and divide each by their sum so they add to 100% instead of the 104ish% the prices imply. That removes the margin from that book's own pair using nothing but its own numbers. The consensus is the median of those estimates across all the books posting that number, and a bet needs at least 4 of them before it gets a consensus at all.

Multiply the legs' consensus probabilities and you have the probability of the whole ticket. One divided by that is the fair price. The percentage on every row above is the offered price measured against that fair price: −8.4% means the ticket pays 8.4% less than the consensus probability corresponds to, and +1.2% means it pays 1.2% more. The comparison is always against the market's own devigged numbers, never against a forecast of ours.

Where the prices come from, and what is left out

• Game markets come from our board capture, which runs ten times a day. Player props are captured once per event in the morning and are never refreshed, so every prop leg carries the time it was captured and is labelled as an opening price. A price nobody has updated since breakfast is still a real price, but it is not today's price.
• Home run props are absent and cannot be added. Books post the over side of the home run rungs and no under side at all, so there is no second price to remove the margin against and no consensus to compare anything with. Measured on our own capture of one slate: 1,528 over quotes, zero under quotes.
• One-sided quotes are not used. If a book shows only the side you want and not its opposite at the same number, we cannot confirm from our own capture that it is a live two-way market, so its price is left out of both the consensus and the ticket pricing.
• A book that has stopped updating is not compared. Any game-market quote more than three hours behind the newest capture on the board is dropped. Without that guard a book that simply went quiet would top this list every single day, and the difference would be the clock rather than the market.
• Same-game combinations are refused everywhere on this page, in the list and in the builder. Two outcomes in one game are correlated, which breaks both the payout arithmetic and the consensus comparison.

The one thing to keep in mind about a ranked list

This page ranks hundreds of thousands of combinations, and the top of a ranking that large is a maximum. A maximum is won by whichever book has the widest scatter around the consensus, not by the book that prices best on average, and those are often different books. The table below exists so the ranking cannot be read as a house being generous: it shows how each book's own prices sit against the consensus across every leg it quotes on this board.

Margin by market

The same comparison run on single legs rather than tickets, which is why the ranked list is mostly game markets: prop legs sit further from the consensus than moneylines do, so they rarely reach the top on price alone.

Why a parlay pays less than it looks like it pays

A single bet at −110 is priced so that both sides together imply about 104.8% of the outcomes, not 100%. That extra 4.8 points is the margin built into the price structure. Take the margin out and a −110 side is a coin flip: 50%, not the 52.4% its price implies.

Now multiply. A parlay's price is the product of its legs' decimal prices, and its true probability is the product of the legs' true probabilities. So the margin does not add up across legs, it compounds. If each leg returns 95.45 cents of a fair dollar, then two legs return 91.11 cents and three legs return 86.97 cents. In shorthand, a ticket of n legs each priced at decimal d against a true probability p returns (p × d)ⁿ of a fair payout. Every leg you add multiplies the shortfall by another factor.

The part that inverts

A +600 ticket breaks even at a 14.29% hit rate, and the three −110 legs above break even at 14.37%. A hit rate of 50% on tickets like that would be a spectacular result. But 50% of three-leg parlays is not a 50% skill level on the legs. It is 79.4% on every single leg, because 0.794 cubed is 0.50. The prices themselves imply 52.4% per leg. The payout looks like it is rewarding a long shot, while the hit rate needed to reach a coin-flip strike rate on the ticket is far above anything the individual prices describe. That is the inversion: the ticket gets easier to be paid well on and much harder to win, at the same time.

Why shopping the ticket matters more than shopping a leg

Because the shortfall compounds, so does its cure. Half a point of price on one leg is worth a fraction of a percent; the same improvement on three legs multiplies. That is the whole reason the list above prices each combination at every book in turn rather than showing one number, and it is why the ranking moves around between books far more than any single leg does. What none of it can do is let you mix books. One ticket, one book every number on this page is a single book's own ticket.

Why parlays get promoted

Not because anything is hidden. The prices are all posted in the open, and the arithmetic above uses nothing but them. It is that the margin per ticket grows with the number of legs while the stake does not. A book holding roughly 4.5% on a straight side holds roughly 13% on a three-leg version of the same sides. That is the whole mechanism, and it is why the tool above prints the offered payout and the fair payout side by side instead of just the payout.

What the compounding measures out to in our own archive

The arithmetic says margins multiply. Our odds archive says how much they multiply by in practice. The figures below come from a committed script over a read-only session on our own lake: every pre-game MLB moneyline our board captured, and then every pair and every triple of different games the same book was pricing at the same instant.

MLB moneylines, pre-game captures only, 2025-02-23 to 2026-08-14 · 11 books · 4,372 games · one slate per book per ET day (that day's earliest board run) · row 1 counts posted two-sided markets, rows 2 and 3 count every pair and every triple of different games a single book was pricing at one instant · median, not mean · committed script: seo/f78_parlay_hold_facts.py

Hold is the share of the payout the price structure keeps. Read down the column: 3.98% on one moneyline becomes 7.86% on two and 11.61% on three, which is what multiplying prices does. It is close to what the arithmetic predicts and not identical to it, because the median of a product is not the product of medians: raising the one-leg median to the third power predicts 112.94 and the measured three-leg median is 113.13.

Those are medians over a book's own posted prices, so they describe the typical ticket rather than the best-priced one. The ranked list at the top of this page is the other end of the same distribution: the same arithmetic, applied to today's board, sorted to find where the shortfall is smallest.

For context, our hedge guide measured the same quantity across every sport we hold and found a median of 4.16% across 711,756 two-way quotes. MLB moneylines are tighter than that average, at 3.98%. Two other things worth stating plainly: 10.3% of the MLB price rows in this window were captured after first pitch, and every one of them is excluded here, because an in-play price is not a price you could have put in a pre-game parlay. And the worked example above uses a −110 pair, which holds 4.55%; real MLB moneylines hold less than that. The compounding is the same either way.