Two or three times a week you face the same decision: two players, one spot, and nothing obvious to separate them. The usual help is a ranking. Someone's list has one of them 14th and the other 17th, and you are left to work out for yourself whether three places is a real difference or a rounding error. Most weeks it's a rounding error — and the list almost never says so.
StartSense is built around the part a ranking leaves out: how confident the call is. Give it two players and it gives you three things — what each is expected to score in your league's scoring (based on Las Vegas odds), the chance the one it likes actually outscores the other, and a transparent "I don't know" when the information to answer isn't there yet.
It reads the betting market, not the experts
Every week, sportsbooks post player lines for nearly every starter: passing yards, rushing yards, receptions, the chance of a touchdown. Those numbers are not opinions. They are prices, with real money behind them, set by a business that pays when they are wrong and moves them the moment the news changes. StartSense turns them into fantasy points under your league's rules and compares the players you are choosing between.
I tested that idea before building anything on top of it. Across three seasons of real start/sit decisions between drafted players, starting the player the lines preferred was right 66.6% of the time in standard scoring — about five calls in every 100 better than going by draft position or by the last three weeks, and ahead in 46 or more of the 48 weeks tested. On close calls, where both players look the same on paper, the lines were still right six times in ten, while draft position and recent form were coin flips. The full test is here: the sportsbooks already did your start/sit homework.
The odds are the point
Anyone can put two players in order. The harder promise is the number next to the call: when StartSense says 62%, that should happen 62% of the time. I ran a check: when I built StartSense, I held the 2025 season out from the training data. Then, I used Las Vegas' odds lines from about an hour before kickoff in the 2025 season to test StartSense's accuracy. I tested every pair of players the books priced that season and graded the results, 377,105 of them, in standard scoring:
| StartSense said | It happened | Pairs |
|---|---|---|
| 52% | 52% | 45,084 |
| 58% | 57% | 44,352 |
| 62% | 62% | 45,454 |
| 67% | 67% | 46,744 |
| 72% | 72% | 52,883 |
| 77% | 78% | 42,055 |
| 82% | 82% | 40,492 |
| 87% | 87% | 34,276 |
| 93% | 92% | 20,045 |
| 96% | 96% | 5,720 |
The worst row is off by seven tenths of a percentage point. The tool itself never shows more than 90%, even for calls like the last two rows.
A 62% call you can trust to be 62% tells you something a ranking cannot: that this one is close, that you will lose it nearly four times in ten, and that it is not the decision worth agonizing over on Sunday morning.
What a 55% is for
A probability is only worth having if it changes something. Three ways it changes how I set a lineup:
- Over about 70%, stop thinking. Start the player and close the tab. The real cost of start/sit agonizing is the hour spent on decisions that were never close in the first place.
- Between 50 and 60%, the call is nearly free either way. That is when it is fine to let something else decide: the player you would rather watch, the one you would regret benching, the hunch you cannot shake. A near-coin-flip is the one place a hunch costs you almost nothing in expectation.
- Low numbers all across your lineup are telling you about the roster, not about Sunday. That is a waiver or trade question, and no start/sit call will fix it.
When it says it doesn't know
In about one pair in 50, the two measures disagree: one player has the higher expected score, the other the better chance of finishing ahead. It is always a boom-or-bust player against a steady one, and always within two points. StartSense calls those 50/50 rather than picking a side it cannot defend.
A player with no usable line gets a reason instead of a number:
- Touchdown price only — the books priced his scoring chance but not the yardage his position lives on. Those lines post through the week.
- No yardage line — some lines, but not the one that matters for him.
- Not priced — the books left him alone entirely, which is usually a sign of injury doubt.
- On bye, or his game has not been priced yet.
How often does that bite? In 2025, 91% of drafted players' games had a line by late morning on game day, and 96% by an hour before kickoff. Lines for the coming week first arrive on Thursday and fill in through the weekend, so check again on Saturday.
What it is not good at
The edge is real but it is not magic, and it is thinner in some places than others.
- Quarterbacks and tight ends. These are the two calls where the lines help least, probably because both turn on touchdowns more than on yardage, and a touchdown is the hardest thing for anyone to price.
- A third of the calls are still wrong. Two in three is a good hit rate, not a guarantee. Losing a 72% call is not evidence of anything except football being football.
- Two players in the same game are treated as independent, which isn't true.
- "Not priced" is a hint, not a diagnosis. It means the market has doubts. Go and read the injury report; don't let a missing line make the decision for you.
Where it is
StartSense is on the site now, free, at fantasysense.app/start-sense. There is nothing to sign up for and nothing is stored: you ask, it answers, it forgets.
It answers the upcoming week only, which is the week with a game still to kick off. Once Monday night's game kicks off, it moves on to the next week. If it gets a call wrong in a way you think the method should have caught, that is worth an email: hello@fantasysense.app.
Player lines: The Odds API. Draft positions: Fantasy Football Calculator ADP (fantasyfootballcalculator.com). Box scores and schedules: nflverse (CC BY 4.0). Nothing here is betting advice.