A Price Can Move Six Cents Without Anybody Changing Their Mind

Seventy-five percent of the money bet at DraftKings on tonight's game in New York is on Washington. The Nationals opened at plus 104 and they are plus 105 this morning.

Read those two sentences together and something is wrong. Money is supposed to shorten a price. Three quarters of the dollars sit on the Nationals, the Nationals got a cent cheaper, and the standard reading of that is a market ignoring its own sharpest customers.

The standard reading is wrong, and the reason is a third number nobody prints.

The margin, and why it makes a price lie to you#

Convert a price into the probability it implies. Washington at plus 105 means a hundred dollars wins a hundred and five, so the implied probability is 100 divided by 205, or 48.78 percent. The Mets at minus 113 implies 113 divided by 213, or 53.05 percent.

Add them. One of those two teams wins tonight, so an honest pair would sum to 100. They sum to 101.83.

That extra 1.83 is the margin, also called the overround or the vig, and it is the sportsbook's cut. It sits inside both prices at once, in the same direction, making each side look likelier than the book actually thinks it is. Divide both by 1.0183 and you get the book's real read: Washington 47.90 percent, the Mets 52.10.

Now run the identical arithmetic on the opening price. Washington plus 104 and the Mets minus 126 implies 49.02 and 55.75, which sums to 104.77. The opener carried 4.77 points of margin. Strip it and the real read on Washington at the open was 46.79 percent.

So the honest number on the Nationals went from 46.79 to 47.90. It moved a full point toward the side holding three quarters of the money, which is exactly what should happen, while the posted price went the other way by a cent.

Here is the same fact stated as a price, because that is the form a reader actually meets it in. Washington's current honest probability of 47.90 percent, quoted inside the opener's fat 4.77 percent margin, would be posted at minus 101, a hair the wrong side of even money. Quoted inside this morning's 1.83, it is posted at plus 105. Same opinion, six cents apart, counting even money once on the way through. The opinion moved the price five cents shorter and the margin moved it six cents longer, and the six won.

Nobody changed their mind about the Nationals by six cents. The market simply got closer to first pitch, and a market gets thinner and cheaper as it matures, because more books are up, limits are higher, and the early wide number nobody could bet in size has been competed away.

Four games qualify this morning, and the two scoring methods disagree#

I check DraftKings' published splits every Friday for games where one moneyline side holds a handle share (the weight of cash) at least 25 points above its ticket share (the count of bets, one per customer regardless of size). Four of the ten games carried on that page this morning clear the bar. All figures below were read at 10:23 a.m. Pacific.

Seattle at Houston is the widest: the Mariners hold 77 percent of the money on 30 percent of the bets. Baltimore at Tampa Bay is next, the Orioles at 60 and 19. Cleveland holds 83 percent of the money against San Diego on 53 percent of the bets, and Washington sits at 75 and 50.

Score those four on the posted price against the opener and the money side is right once. Seattle shortened from plus 108 to plus 106. Washington lengthened a cent, Cleveland got three cents cheaper at minus 128, and Baltimore lengthened twelve.

Score them on the honest number and it is a different morning.

One of four becomes three of four. The four margins collapsed from 4.53, 4.56, 4.77 and 4.79 percent at the open to between 1.73 and 1.83 this morning, which is remarkably uniform across four unrelated games and is the whole reason the two methods disagree.

Four games on one morning is not evidence of anything about baseball. It is a demonstration that the measuring instrument matters more than the measurement, and that part generalizes.

This breaks a call I made in public two weeks ago#

On July 31 this desk filed a scoreable call on exactly these games: that on the eight Friday cards through September 25, the money side's consensus closing moneyline would finish shorter than the price printed beside its split in at least 60 percent of qualifying cases.

That standard is a posted-price standard. It measures the margin as well as the money, and this morning's arithmetic says the margin can be the larger of the two.

The honest size of the problem, which is smaller than the paragraph above implies. The damage runs from the open to the read, where 2.9 points of margin came out. From a 10:23 a.m. read to a closing number the remaining collapse is on the order of a point, and a point of margin is worth roughly two cents on a plus 105 underdog. So the call is not broken. It is biased against itself by a couple of cents per game, in the direction of scoring FALSE, on a threshold of 60 percent.

I am not touching it. A call whose method was fixed in advance and then adjusted by its author when the arithmetic got inconvenient is worth nothing, and the whole value of writing these down is that they grade as written. It stands, and if it scores FALSE by a hair I will say in public that the standard rather than the thesis is what failed.

What I can do is file the better instrument beside it and let both run. That is at the bottom.

Baltimore is the one that actually moved against its money#

Strip the margin from the Orioles and they still went the wrong way: 40.71 percent at the open, 39.80 this morning. That is a real move against a side holding 60 percent of the moneyline cash on 19 percent of the tickets, and the run line in the same game says the same thing louder, with 90 percent of the money on Baltimore plus 1.5 against 16 percent of the bets.

So the dollars are on Baltimore in two separate markets, the bet count is against Baltimore in both, and the price walked away from the dollars by twelve cents.

I have not established what did that, and I am not going to guess at it in print. A pitching change, a lineup card, weather, or a book unloading last night's exposure will all move a baseball price without anybody sharp having an opinion. The useful thing is that this is the one game of four where the correction does not rescue the reading, which makes it the game worth watching tonight rather than the three that behaved.

The counter-case#

Proportional margin removal is a choice. Splitting the overround in proportion to each side's implied probability is standard and it is not the only method. On a near-even market the alternatives land within a couple of tenths of a point, and all four games here are near even, the longest being Baltimore at plus 147. On a heavy favorite the choice matters much more, and none of this arithmetic should be carried over to a minus 300 board without redoing it.

An opener is not a price most people could have taken. The 4.7 percent margin at the open is partly a real cost and partly an artifact of an early market with low limits and few books posting. Calling the collapse from 4.7 to 1.8 a saving assumes a bettor who could have transacted at the open in size, and most could not.

The splits and the closing number come from different populations. The split describes DraftKings' customers. The consensus price is eight books. A conclusion drawn by comparing one book's cash against eight books' prices is softer than it looks, and that is a flaw in my own method rather than a hedge.

Four is four. Two of the three margins that flipped a verdict are under 1.2 points, which is inside the range that a different de-vig method or a fifteen-minute-later read could produce on its own.

The four games that are not on the page at all#

DraftKings carried splits for ten games this morning. Fourteen are scheduled.

The four missing are the four latest: Kansas City at the Angels, Texas at the Athletics, Milwaukee at the Dodgers, and Colorado at San Francisco, first pitch 10:15 Eastern. Every game starting after 8:10 p.m. Eastern is absent, which is every West Coast game on the card.

A reader in the Bay Area who wants to know what the money is doing on the Giants tonight cannot find out from the page that the rest of the country is quoting. That is not price shading and it is not the fan tax, which is a different subject and needs a real sample before this desk asserts it. It is something plainer: the market information that gets published is published on Eastern time, and a Giants bettor is structurally last in line for it.

It is also a defect in the call I filed, and I would rather name it than have it named for me. The population that call grades against is not the day's card. It is whichever subset of the card DraftKings had posted at the hour I read it, and that subset systematically excludes the two Bay Area clubs. Tonight it is 10 of 14. That count gets printed every Friday from here.

The call#

No plays. This desk's betting record exists and is empty, and it stays empty until the machinery behind it is built, because a play logged without the apparatus to grade it honestly is a tout sheet with better manners.

One call, filed at today's numbers and scoreable by a stranger with a calculator: on the six remaining Friday cards through September 25, in qualifying games, the money side's de-vigged win probability rises from the morning read to the consensus close in at least 60 percent of cases, while the same side's posted American price shortens in fewer than 50 percent of cases.

The gap between those two figures is the entire claim. If both come out the same, the margin correction is decoration and I have spent a column teaching a distinction that does not exist. If the second figure lands above the first, the posted price is a better instrument than the honest one and I have it backwards. Under 20 qualifying games it voids, and the count gets printed either way.

What would make this wrong. If the margin at the morning read is already close to the margin at the close, there is nothing left to collapse and the two methods converge, which would show up as the two percentages landing within a few points of each other. And if the money side simply is not predictive, both methods score under 60 and the correction is a better measurement of a thesis that was never true. I would rather find that out on six Fridays of printed counts than assume it for a season.

Splits, paired prices and the game count: DraftKings' published game splits, read 2026-08-14 at 10:23 a.m. Pacific. Opening and consensus moneylines: VegasInsider's Las Vegas MLB board, page dated 8/13/26. Full schedule and start times: ESPN's MLB schedule for August 14, 2026. Implied probabilities and margin removal use the standard proportional method. The July 31 call referenced above is on this site's Called Shots ledger with its method fixed as filed.

Betting content is for readers 21 and over. If gambling stops being entertainment, the National Problem Gambling Helpline is 1-800-522-4700, free, confidential, and answered 24 hours a day.