How the Vegas start/sit test worked

This page has the details behind the Vegas start/sit test. The FantasySense Forecast turns lines into points the same way. The page starts with how to use the lines yourself: how to read a prop line, what the lines say about kickers, and what to do about it. After that comes exactly how I ran the test.

How to read a prop line

You do not need a spreadsheet to use the lines. You need two facts about how they are set.

1. A line is the typical game, not the average one. Over the three seasons, players finished under their rushing-yards, receiving-yards and catches lines 52–54% of the time. Yet their average finish was above the line: 14% above for receiving yards and 8% above for rushing yards. Both are true because yardage is lopsided (statistically, it is positively skewed). Most games land a little short, and a few monster games pull the average up. Passing yards are the exception: on average they came in almost exactly at the line.

So, to turn a line into a projection I made the following adjustments:

• Receiving yards: add about 14%.
• Rushing yards: add about 8%.
• Passing yards: use the line as it is.
• Catches: add about a tenth of a catch.

(None of this is betting advice. It is about reading the numbers, not beating them.)

2. Touchdown odds are too generous on long shots. The anytime-touchdown price includes the sportsbook's cut, and most of that cut sits on the unlikely scorers:

To turn odds into a percentage: for plus odds, divide 100 by the odds plus 100, so +150 is 40%. For minus odds, divide the odds by the odds plus 100, ignoring the minus sign, so −150 is 60%.

One wrinkle: a player with a real 36% chance to score is worth a little more than 0.36 touchdowns, because the players who score sometimes score twice. It works out to about 0.43.

Put it together. Then add it up the way your league scores: yards divided by 10 (passing yards by 25), 6 points for each expected touchdown, 4 per passing touchdown, and the catches if your league pays for them. For a receiver with a 64.5-yard line, a 5.5-catch line and +150 to score:

• Yards: 64.5 plus 14% is about 74, so 7.4 points.
• Catches: about 5.6, so 5.6 points in PPR.
• Touchdowns: +150 looks like 40%, is really about 36%, and is worth about 0.43 touchdowns, so 2.6 points.
• Total: about 15.6 PPR points. Read at face value, the same lines say about 14.4.

Potential shortcut: if you are choosing between two players at the same position, you can skip the corrections. Add up the raw lines and compare, because the raw numbers put players in almost exactly the same order. The corrections matter when you compare a back with a receiver for your flex spot, and whenever you care about the actual point total.

What to do about it

• Check the lines before you set your lineup. Every major sportsbook publishes player props, and so do plenty of free odds sites. You don't have to bet to read them.
• Same position: add up each player's lines and start the bigger number.
• Flex: apply the corrections above before you compare a back with a receiver.
• No line on Sunday morning? Treat the player as doubtful and check the news.
• Kickers: do not overthink it.

The lines will not be right every time. A third of the time, the other player scores more. But across three seasons, none of the habits I tested matched them in a single season, in any scoring format. Neither did the expert projections.

The rest of this page sets out exactly how the test was run.

The lines

Player prop lines came from The Odds API, which gathers them from sportsbooks that take U.S. bets: up to ten books a game. I used the lines for passing, rushing and receiving yards, catches, passing touchdowns, interceptions, anytime-touchdown odds and kicking points, as they stood one hour before kickoff, for all 855 NFL games of the 2023, 2024 and 2025 seasons, playoffs included.

From lines to fantasy points

A prop line is not a projection yet. Each one went through the same steps:

• One vote per book. Two sportsbooks posting identical prices from the same odds feed count once. Where a book offers several lines for the same stat, its main line is the one priced closest to even.
• The sportsbook's cut comes off. The over and under prices are scaled so their chances add up to exactly 100%.
• From the typical game to the average one. A line marks the typical game, and the price shows which way the book leans from it. Each line and its price become an expected yardage or count, which is then adjusted by a factor learned from past seasons. For yardage that factor is a real boost, because a few huge games pull the average above the typical game.
• The middle book. Each stat takes the median across the books.
• Touchdowns. Anytime-touchdown odds are posted for the "yes" side only, so the cut cannot be taken off the same way. Instead, the typical price across the books is matched to how often players at that price really scored in past seasons, and that chance becomes expected touchdowns, allowing for two-touchdown games.
• Stats with no line. A receiver's rushing yards, say, or anyone's fumbles, take the average from past seasons for players at the same position who were priced the same way.

Every setting was learned only from seasons other than the one being graded: 2025 from 2023 and 2024, 2024 from 2023, and 2023 from 2024. For 2025 the yardage boost came out at 14% for receiving yards, 8% for rushing yards and under 1% for passing yards.

The expected stats were then scored like real games: a point per 10 rushing or receiving yards and per 25 passing yards, 6 per rushing or receiving touchdown, 4 per passing touchdown, minus 2 per interception or lost fumble, 2 per two-point conversion, and 0, 0.5 or 1 per catch, depending on the format.

Which calls counted

A start/sit call is any two players at the same position in the same week, or any running back, receiver or tight end against one of the others for a flex spot. A pair counted only if:

• both players were on Fantasy Football Calculator's 12-team draft board for that scoring format;
• both recorded a stat that week;
• the books had priced both on the yardage stat their position lives on: passing yards for quarterbacks, rushing yards for backs, receiving yards for receivers and tight ends;
• the week fell between week 2 and week 17. Week 1 has no "last week" to go on, and week 18 is when teams rest starters.

A call counts as right when the pick outscored the other player, and ties count half. That makes 282,031 calls in standard scoring, 250,505 in half-PPR and 337,793 in PPR. The books had lines for 94–98% of those players' games.

Close calls are the pairs drafted within one round of each other (12 picks on these 12-team boards) whose last-three-weeks averages were within 3 points. They are defined only by what a manager can see without the lines, and make up 5% of the calls in standard scoring, 4.5% in half-PPR and 4% in PPR. As a check, the same comparison was run with tighter and looser limits (half a round and 1.5 points; two rounds and 5 points), and on the pairs that the combination of draft position and the last three weeks (below) put within 1 point of each other. On every version the lines picked right between 58.8% and 60.9% of the time, and nothing else reached 55%.

The habits

• Last week and last three weeks use the player's team's most recent games, with byes skipped and a missed game counted as zero, the way a box score shows them.
• Draft position is Fantasy Football Calculator's average draft position for 12-team leagues in that scoring format, as it stood in late August, before the season. It never changes during the season.
• Draft position + last three weeks, the combination, is a weighted average of two numbers: the points a player's draft position usually brings at his position, read off a curve fitted to past seasons, and his last-three-weeks average. Draft position's weight is k divided by (k plus the number of weeks already played), so it starts at about 95% in week 2 and falls to between a half and two-thirds by week 17. The curves and k (between 16 and 30) were learned only from seasons other than the one being graded. K is selected such that it calls the most start/sit pairs right. It is graded on exactly the same calls as everything else.

I also tested adjusting that combination for the opposing defense. Each defense's points allowed per game (last season's rate, moved toward this season's games before the week) was divided by the league average, and the combination was multiplied by that ratio raised to a power. How much last season counted, how hard the adjustment pushed (including not at all) and the combination's own k were all set from other seasons. Last season's numbers stood in for preseason defense rankings, because the draft boards rank only 15 to 23 of the 32 defenses. Two versions were tested: fantasy points allowed to the player's position, and points allowed on the scoreboard. Neither moved the results by as much as a tenth of a call per 100, and when the adjustment changed a pick it was right between 48% and 51% of the time. So the scoreboard shows the combination without it.

How confident is this?

Calls from the same week share the same games, so they are not separate pieces of evidence. Every comparison was therefore also checked week by week, and a lead counted only if it held up across the weeks, not just in the grand total. Against each single habit, the lines were ahead in at least 47 of the 48 weeks, and against the combination in at least 46.

The other figures

The figures above on how lines compare with results, such as how often players finish under them and how touchdown odds hold up, cover every quarterback, back, receiver and tight end who had a line an hour before kickoff and played, across all three seasons. A player counts as having played if he was on the field for at least one snap, which is also when sportsbooks settle a prop.

Kickers

The kicker figures above use standard kicker scoring: 3, 4 or 5 points for a field goal depending on its length, 1 per extra point, and minus 1 per missed field goal. They cover the same weeks 2 to 17 of the regular season. Every two kickers in the same week make a call, counting kickers who had a kicking-points line, a team total and at least one earlier game that season: 19,584 calls in all.

The three forecasts are the kicking-points line (three per field goal, one per extra point) adjusted for its price, the points the betting lines expected the kicker's team to score, and the kicker's average so far that season. The first two were turned into fantasy points with a straight-line fit on other seasons.

Sources

Player lines: The Odds API. Draft positions: Fantasy Football Calculator ADP (fantasyfootballcalculator.com). Box scores, schedules, game lines and snap counts: nflverse (CC BY 4.0). Nothing here is betting advice.