Who's Actually Sharp on College Football Mainlines?
Mapping who leads, who follows, and where the real value is across twelve books on 2026 college football spreads, totals, and moneylines.
- Author
- Alex Bow
- Published
- Nov 16, 2026
- Read
- 16 min

College football is a weird market to price. There are over a hundred games on a big Saturday, lines go up as early as six days before kickoff, and the gap between a nationally televised SEC game and a Thursday night MAC game is enormous in terms of how much money and attention each one gets. That makes it a great place to ask a simple question. When a CFB line moves, who is actually moving it, and who is just following?
We pulled line by line odds data across twelve books for the 2026 college football season, from the end of August through October 31, a little over 1,000 games in all. We looked at moneylines, spreads and totals, and ran a handful of different tests on how each book behaves. Some of it confirmed what most bettors would already guess. Some of it didn't.
The books, and why we picked them
Pinnacle. The long standing gold standard for sharp lines, and the book everyone else gets measured against.
Bookmaker and BetOnline. Two of the most widely used offshore books. Neither is supposed to be completely soft, and Bookmaker is known for posting early and taking real limits.
Kalshi and Polymarket. The two biggest names in prediction markets right now.
Novig and ProphetX. The two newer prediction markets with fresh CFTC recognition, now operating under the same rules as Kalshi and Polymarket.
4Caster. A global crypto exchange with real liquidity behind it, and a reference point outside the US regulatory framework.
FanDuel and DraftKings. The two biggest recreational sportsbooks in the country.
Fanatics and Caesars. Our soft book sanity checks. If the method works, these should look soft, the same way most bettors already expect.
How We Got Here
A few quick notes on how we set this up. Every line is devigged, so a book's juice doesn't make it look sharper or softer, and the one exception is crossed markets, which use the actual posted prices. Any price with a reported max bet under $100 doesn't count, and on the exchanges we used the best price you could actually bet at that moment. We also had to account for refill bots. When someone takes a resting offer on an exchange, a bot will often post the same price again right away, which makes the price bounce back and forth. So a new exchange price had to hold for five minutes before we counted it as a move.
Line Accuracy Through the Week (Brier)
A Brier score is a simple way to grade a forecast. You take a book's devigged win probability, subtract the result (1 if the team won, 0 if it lost), and square it. If Alabama is 90% against East Carolina and wins, that's a score of 0.01. A true 50/50 game scores 0.25 no matter who wins. Lower is better.
For mainlines, Brier isn't always that useful. Spreads are tight, liquidity is massive, lines react quickly, and by kickoff almost every book is sitting in the same place. It also has nothing to do with line movement. It's just snapshots of each book's line at different points in the week, so it tells you more about who you can devig against than who is moving the market. It's more useful in football because lines go up five or six days before kickoff, unlike the NBA where a game might only have a line for a day or two.
Through the week there was no clear and obvious leader. The differences between books were small. From using it ourselves, Bookmaker has some good lines early, and it posts them with limits of $5,000 or more. Pinnacle's early limits are much smaller. Its median max bet six days out was under $800, rising to around $10,000 by kickoff, and it only had a line on about 17% of games that far out. The exchanges are thin early too, with median size of about $93 on ProphetX, $249 on Polymarket and $439 on Novig six days out, and less liquidity usually means less confidence in a price.
Here's how to read the table below. For every game, at each point in the week, we measured how far a book's price was from where the market ended up closing. Then we subtracted the average distance for the other books on that same game. A negative number means that book was closer to the close than the others, and a positive number means it was further away. So FanDuel's −0.15 at 144 hours means that six days before kickoff, FanDuel was 0.15 points closer to the close than the other books on those games. These are points of win probability and the average book is a few points off that far out, so differences this small are why there's no clear leader.
Gap to the closing line against the other books on the same games, in points (negative is closer to the close)
| Book | 144h | 120h | 96h | 72h | 48h | 24h |
|---|---|---|---|---|---|---|
| Pinnacle | −0.04 | 0.09 | 0.07 | 0.01 | 0.02 | 0.01 |
| Bookmaker | 0.07 | 0.10 | 0.10 | 0.15 | 0.08 | 0.14 |
| Novig | −0.01 | −0.07 | −0.07 | −0.04 | −0.08 | −0.08 |
| ProphetX | −0.05 | −0.07 | −0.06 | −0.09 | −0.11 | −0.10 |
| 4Caster | −0.13 | 0.03 | −0.06 | 0.01 | −0.02 | −0.10 |
| Kalshi | 0.11 | 0.05 | 0.08 | 0.05 | 0.07 | 0.12 |
| Polymarket | 0.21 | 0.15 | 0.05 | 0.05 | 0.08 | 0.12 |
| FanDuel | −0.15 | −0.05 | −0.03 | −0.03 | 0.01 | −0.11 |
| DraftKings | −0.01 | −0.07 | −0.06 | −0.09 | −0.04 | −0.02 |
| Fanatics | −0.02 | 0.06 | 0.04 | 0.04 | 0.04 | 0.02 |
| Caesars | 0.03 | −0.05 | −0.03 | 0.02 | 0.01 | 0.04 |
| BetOnline | 0.01 | −0.08 | −0.04 | −0.03 | −0.04 | −0.04 |
Crossed Markets
A crossed market is when two books disagree enough that you could bet one side at one book and the other side at the other book and come out ahead on paper whichever side wins. When that happens, at least one of those prices has to be +EV. The question is which one.
Everything in this section uses the actual posted prices. A cross only counts if the two prices really add up to under 100%. For each one we asked a simple question. If you bet that book's side, did you beat where the market closed? The closing number is the only thing that's devigged, because every price needs a fair number to be compared against. We also threw out any cross showing an edge above 5%, since the biggest ones were clearly bad prints, and that removed less than 2% of them. The more negative a book's number, the sharper it is, because you'd lose money betting into it.
Every cross by book on posted prices, toughest to beat first. Profit is points of closing line value
| Book | Crosses | Profit betting its side (pts) | Std error | Its side was the better bet (%) |
|---|---|---|---|---|
| Pinnacle | 5,012 | −0.56 | 0.05 | 35 |
| DraftKings | 3,848 | −0.45 | 0.06 | 40 |
| ProphetX | 7,733 | −0.10 | 0.03 | 42 |
| 4Caster | 8,409 | −0.08 | 0.03 | 43 |
| Caesars | 2,543 | −0.01 | 0.06 | 46 |
| FanDuel | 6,260 | 0.01 | 0.05 | 45 |
| Kalshi | 6,466 | 0.13 | 0.05 | 46 |
| Novig | 11,911 | 0.21 | 0.03 | 49 |
| Bookmaker | 8,450 | 0.40 | 0.04 | 51 |
| Fanatics | 6,575 | 0.67 | 0.05 | 52 |
| Polymarket | 10,723 | 1.19 | 0.03 | 62 |
| BetOnline | 8,436 | 2.02 | 0.04 | 71 |
To no surprise, Pinnacle is the toughest book here. If you bet Pinnacle's side of every cross you'd lose about 0.56 points against the close, and the other side was the better bet about 65% of the time. DraftKings was next at −0.45. ProphetX and 4Caster were just under zero. On the other end, betting BetOnline's side of a cross returned about 2 points and Polymarket's returned 1.2. Fanatics returned 0.67, Bookmaker 0.40 and Novig 0.21. Caesars and FanDuel were about break even.
The market type matters. BetOnline returned over 2 points on spreads and totals, and nothing on moneylines. Bookmaker has the same pattern, slightly positive on spreads and totals and slightly negative on moneylines. Fanatics and Polymarket were profitable in every market. Pinnacle was negative in every market, and toughest on moneylines and totals. Kalshi is the most interesting split. Betting its side returned 1.14 points on moneylines, but it lost money on spreads and totals. There are also a lot more crosses on spreads and totals, about 17,000 each, than on moneylines, about 9,000.
Head to head on posted prices. Row book's profit minus column book's profit, in points (negative means the row book was tougher to beat)
| Book | Pinnacle | Bookmaker | Novig | ProphetX | 4Caster | Kalshi | Polymarket | FanDuel | DraftKings | Fanatics | Caesars | BetOnline |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Pinnacle | n/a | −1.01 | −0.72 | −0.32 | −0.21 | −2.63 | −3.29 | −1.40 | −0.86 | −2.66 | −0.55 | −3.82 |
| Bookmaker | 1.01 | n/a | −0.44 | 0.24 | 0.06 | 0.50 | −0.92 | 0.21 | 0.95 | −0.33 | 0.75 | −2.36 |
| Novig | 0.72 | 0.44 | n/a | 0.49 | 0.62 | −0.73 | −0.57 | 0.01 | −0.61 | −1.08 | −0.31 | −1.71 |
| ProphetX | 0.32 | −0.24 | −0.49 | n/a | 0.32 | −0.75 | −2.25 | −0.24 | −0.07 | −0.98 | −0.75 | −2.51 |
| 4Caster | 0.21 | −0.06 | −0.62 | −0.32 | n/a | −0.84 | −1.66 | −0.23 | −0.18 | −1.12 | −0.34 | −2.18 |
| Kalshi | 2.63 | −0.50 | 0.73 | 0.75 | 0.84 | n/a | −1.97 | 0.29 | −0.32 | −1.56 | 1.94 | −3.89 |
| Polymarket | 3.29 | 0.92 | 0.57 | 2.25 | 1.66 | 1.97 | n/a | 1.66 | 3.71 | 1.08 | 4.18 | −1.96 |
| FanDuel | 1.40 | −0.21 | −0.01 | 0.24 | 0.23 | −0.29 | −1.66 | n/a | −1.34 | −2.40 | 0.54 | −4.22 |
| DraftKings | 0.86 | −0.95 | 0.61 | 0.07 | 0.18 | 0.32 | −3.71 | 1.34 | n/a | −3.26 | n/a | −7.51 |
| Fanatics | 2.66 | 0.33 | 1.08 | 0.98 | 1.12 | 1.56 | −1.08 | 2.40 | 3.26 | n/a | 1.71 | −4.50 |
| Caesars | 0.55 | −0.75 | 0.31 | 0.75 | 0.34 | −1.94 | −4.18 | −0.54 | n/a | −1.71 | n/a | −4.62 |
| BetOnline | 3.82 | 2.36 | 1.71 | 2.51 | 2.18 | 3.89 | 1.96 | 4.22 | 7.51 | 4.50 | 4.62 | n/a |
The head to head grid is where it gets interesting. Pinnacle beats everyone, though only by 0.21 against 4Caster and 0.32 against ProphetX. 4Caster loses only to Pinnacle, and is basically even with Bookmaker. ProphetX loses only to Pinnacle and 4Caster. Novig loses to those three and to Bookmaker. So at the top, the order is Pinnacle, 4Caster, ProphetX. DraftKings has the second lowest profit overall, but it gets there by beating BetOnline, Fanatics and Polymarket by a lot. Against the sharper books it mostly loses. BetOnline loses to everyone, and Polymarket loses to everyone except BetOnline. Kalshi was a disappointment, losing to Pinnacle by 2.6 points and to 4Caster, ProphetX and Novig by 0.7 to 0.8.
One thing worth being upfront about. BetOnline's data feed is sparse. In the crosses it showed up in, the feed typically hadn't confirmed its price in about 16 hours, and 73% of the time it had been over an hour. Fanatics had the same issue on about half of its crosses. So some of those crosses may be prices that weren't actually still up. These are softer books, but how soft is probably overstated here.
Who Blinks After a Cross
This is an extension of the crossed market study. Grading a cross by where the market ends up works over a big sample, but it carries a lot of variance. Say there's a Pinnacle and BetOnline cross on a Big Ten game, and you take the BetOnline side. That's probably a good price. But if the starting quarterback breaks his leg on Friday, after you placed the bet, the market is going to move against you anyway. Playing that bet is still likely +EV over time, but variance like that can make too big of an impact. So we looked at where the market moved right after the cross. Which book moved, which one stayed still, or did both move a little closer to each other?
What happened in the 30 minutes after a cross
| Book | Crosses | Book moved (%) | Other book moved (%) | Both moved (%) | Neither (%) | Book's price gone (%) | Book moved first (%) | Other book moved first (%) |
|---|---|---|---|---|---|---|---|---|
| Pinnacle | 3,712 | 11 | 24 | 4 | 42 | 4 | 17 | 33 |
| Bookmaker | 6,479 | 34 | 11 | 10 | 27 | 0 | 52 | 14 |
| Novig | 6,656 | 15 | 18 | 3 | 32 | 22 | 25 | 29 |
| ProphetX | 2,907 | 9 | 12 | 3 | 25 | 44 | 19 | 32 |
| 4Caster | 4,166 | 4 | 10 | 1 | 13 | 70 | 10 | 42 |
| Kalshi | 4,134 | 17 | 12 | 4 | 44 | 7 | 27 | 19 |
| Polymarket | 6,126 | 6 | 17 | 1 | 48 | 8 | 12 | 23 |
| FanDuel | 4,658 | 23 | 16 | 5 | 34 | 0 | 35 | 21 |
| DraftKings | 2,664 | 14 | 15 | 6 | 44 | 6 | 22 | 23 |
| Fanatics | 4,524 | 14 | 9 | 3 | 52 | 0 | 24 | 13 |
| Caesars | 1,680 | 9 | 19 | 3 | 41 | 0 | 15 | 26 |
| BetOnline | 6,174 | 3 | 9 | 1 | 66 | 0 | 5 | 13 |
Here we see Pinnacle showing its muscle. Pinnacle very rarely moves after a cross. In the 30 minutes after one, Pinnacle moved toward the other book 11% of the time, while the other book moved toward Pinnacle 24% of the time, about twice as often. Bookmaker is the opposite. It moved toward the other book 34% of the time against 11% for its opponent. FanDuel leans the same way, 23% against 16%.
The exchanges mostly held their price, but something else happened to them. The offer disappeared. Within 30 minutes of a cross, 4Caster's price was gone 70% of the time, ProphetX's 44% and Novig's 22%. Most likely someone bet into it. So on the exchanges, a cross usually gets closed by the liquidity getting taken, not by the price ‘moving’ the same way traditional sportsbooks do.
How often the row book moved within 30 minutes of a cross with the column book (%)
| Book | Pinnacle | Bookmaker | Novig | ProphetX | 4Caster | Kalshi | Polymarket | FanDuel | DraftKings | Fanatics | Caesars | BetOnline |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Pinnacle | n/a | 14 | 21 | 12 | 14 | 7 | 11 | 14 | 4 | 5 | 13 | 4 |
| Bookmaker | 47 | n/a | 28 | 22 | 15 | 35 | 43 | 37 | 29 | 35 | 56 | 37 |
| Novig | 38 | 14 | n/a | 13 | 17 | 8 | 18 | 17 | 19 | 7 | 11 | 9 |
| ProphetX | 23 | 11 | 15 | n/a | 6 | 5 | 10 | 15 | 16 | 3 | 4 | 2 |
| 4Caster | 10 | 6 | 6 | 4 | n/a | 3 | 2 | 5 | 4 | 1 | 1 | 2 |
| Kalshi | 39 | 14 | 33 | 14 | 14 | n/a | 11 | 24 | 26 | 10 | 19 | 4 |
| Polymarket | 10 | 6 | 8 | 5 | 3 | 6 | n/a | 9 | 8 | 4 | 6 | 4 |
| FanDuel | 45 | 20 | 28 | 24 | 16 | 22 | 27 | n/a | 29 | 15 | 37 | 13 |
| DraftKings | 32 | 19 | 21 | 14 | 16 | 7 | 10 | 22 | n/a | 5 | n/a | 2 |
| Fanatics | 20 | 14 | 18 | 21 | 6 | 7 | 16 | 22 | 10 | n/a | 20 | 7 |
| Caesars | 13 | 7 | 12 | 7 | 3 | 9 | 7 | 18 | n/a | 7 | n/a | n/a |
| BetOnline | 3 | 4 | 6 | 1 | 1 | 4 | 4 | 4 | 2 | 2 | n/a | n/a |
In the head to head, Polymarket did well, and so did 4Caster. Both held their number far more often than the books they crossed with. Against Polymarket, Bookmaker moved 43% of the time and Polymarket 6%. The books doing the moving after a cross were FanDuel, Fanatics and Bookmaker.
+EV Outliers
Here is one final extension of crossed markets, and this time we're adding a constraint. There's a cross, but one book's price is also at least 1% +EV against two or more other sharps. Say Novig is crossed with 4Caster, and Novig's price is also +EV against Pinnacle, Kalshi and ProphetX. Now Novig looks like the true outlier. You see this a lot on the exchanges, Novig/4caster especially due to their incredibly tight spreads, and the question is simple. Is that book super early to the market, or super late?
Our sharps here are Pinnacle, Novig, ProphetX, 4Caster, Kalshi, Polymarket and FanDuel. For each case we looked at the 30 minutes before, to see how the gap opened. Either the book moved into that price, or the sharps moved and the book stayed put. Then we looked at the 30 minutes after, to see who closed it. If the sharps moved toward the book, it was early. If the book moved back toward the sharps, it was late or just wrong. That gave us about 8,300 cases. The before columns don't add up to 100, because some cases didn't have enough price history to tell.
Crossed and at least 1% +EV against two or more sharps, most often early at the top
| Book | Cases | Median EV (%) | Book moved into the price (%) | Sharps moved, book stale (%) | Gap older than 30 min (%) | Sharps came to the book (%) | Book moved back (%) | Neither (%) |
|---|---|---|---|---|---|---|---|---|
| Pinnacle | 518 | 1.9 | 56 | 3 | 5 | 58 | 10 | 25 |
| ProphetX | 391 | 2.4 | 50 | 6 | 17 | 41 | 23 | 33 |
| 4Caster | 166 | 2.5 | 60 | 10 | 25 | 36 | 24 | 36 |
| Novig | 1,479 | 3.0 | 52 | 9 | 23 | 32 | 24 | 40 |
| DraftKings | 232 | 1.6 | 54 | 16 | 19 | 31 | 28 | 36 |
| Polymarket | 1,891 | 3.9 | 8 | 25 | 63 | 9 | 10 | 80 |
| Caesars | 179 | 0.8 | 16 | 5 | 13 | 15 | 17 | 68 |
| BetOnline | 383 | 1.8 | 15 | 14 | 28 | 13 | 20 | 58 |
| FanDuel | 870 | 3.4 | 50 | 4 | 16 | 24 | 36 | 35 |
| Kalshi | 405 | 2.8 | 21 | 32 | 38 | 16 | 35 | 47 |
| Fanatics | 305 | 2.2 | 24 | 14 | 26 | 11 | 31 | 51 |
| Bookmaker | 1,451 | 3.4 | 63 | 0 | 5 | 16 | 43 | 26 |
Pinnacle is the cleanest result. When it's the outlier, the sharps came to Pinnacle 58% of the time, and Pinnacle moved back only 10% of the time. So if Pinnacle looks +EV against two other sharps, the other sharps are usually the ones about to move. ProphetX, 4Caster and Novig lean early too, with the sharps coming to them more often than they move back. DraftKings was about even.
Bookmaker and FanDuel go the other way. Bookmaker moved into the price itself 63% of the time, and then it was the one to move back 43% of the time, against only 16% for the sharps coming to it. FanDuel moved back 36% of the time against 24%. So when one of those two is the outlier, a lot of the time it looks like a number that's about to get corrected.
Kalshi looks more like a book that's late. The sharps had already moved away from it in 32% of its cases, and Kalshi was the one to move back more than twice as often as the sharps came to it. Polymarket has the biggest EV on paper, a median of 3.9%, but it's mostly not real movement. In 63% of its cases the gap was already over 30 minutes old, and 80% of the time nobody moved in the next 30 minutes either. That's more where Polymarket tends to sit than a signal.
Liquidity changes the picture a lot. A price with $100 behind it isn't the same as one with $10,000 behind it, so we split these by the max bet available at that moment.
Sharps came to the book / book moved back (%), by max bet behind the price, with the number of cases in parentheses
| Book | Under $250 | $250-1k | $1k-5k | $5k+ |
|---|---|---|---|---|
| Pinnacle | n/a | 41 / 8 (73) | 62 / 9 (269) | 51 / 21 (82) |
| Novig | 22 / 27 (386) | 29 / 26 (591) | 42 / 19 (370) | 52 / 14 (132) |
| ProphetX | 32 / 24 (159) | 49 / 23 (152) | 44 / 22 (78) | n/a |
| 4Caster | 42 / 17 (24) | 37 / 36 (67) | 32 / 16 (50) | 32 / 16 (25) |
| Polymarket | 5 / 10 (747) | 7 / 10 (750) | 14 / 12 (284) | 28 / 9 (110) |
| Kalshi | 17 / 28 (147) | 17 / 34 (143) | 14 / 41 (76) | 8 / 54 (39) |
Novig shows it best. Under $250, it's a coin flip, with the sharps coming to it 22% of the time and Novig moving back 27%. With $5,000 or more behind the price, the sharps came to it 52% of the time and Novig moved back only 14%. So a big Novig outlier looks early, and a small one is mostly noise. ProphetX and Polymarket also did better with more size behind them, though even at its best Polymarket's 28% is on only 110 cases.
Kalshi goes the opposite way. With $5,000 or more behind it, Kalshi moved back 54% of the time and the sharps came to it only 8%, and in about half of those cases the sharps had already moved away from it first. It's only 39 cases, but it looks like real size sitting on an old number. 4Caster didn't show a clear pattern on a small sample.
One thing to keep in mind on the exchanges. If someone bets into the best offer and the next offer is worse, that shows up here as the book moving back. So on those books, moving back can also mean the price got taken.
Who Actually Moves the Market
Here we're looking at what happens when a book gets 2 or more points away from the rest of the market. We're talking in percentage points, so going from 50% to 52%, which in American odds is about +100 to −108. Who is the leader, the one going away from the rest of the market, and how long does it take the other books to catch up? Or does that book just turn around and revert?
One thing we had to account for is that exchange prices usually can't jump straight from 47% to 50%. They have to stop at each level unless someone sweeps it. So instead of only counting sudden jumps, we counted any time a book was 2 points off the market, then looked back to see whether that book broke away on its own or whether the rest of the market moved and left it there. Novig is a good example of why that matters. Only about a third of its breakaways happened within a minute. Everything below is solo breakaways, where the book did the moving and nobody joined it within 10 seconds.
Solo breakaways where at least one other book reached the number within 5 minutes
| Book | Spread (%) | Total (%) | Moneyline (%) | Solo breakaways, all markets |
|---|---|---|---|---|
| Pinnacle | 60 | 63 | 68 | 948 |
| DraftKings | 40 | 43 | 40 | 658 |
| Caesars | 22 | 10 | 33 | 116 |
| FanDuel | 21 | 25 | 17 | 2,159 |
| Bookmaker | 20 | 27 | 21 | 4,710 |
| Novig | 17 | 21 | 11 | 1,866 |
| BetOnline | 17 | 8 | 37 | 232 |
| ProphetX | 15 | 32 | 8 | 550 |
| Kalshi | 13 | 12 | 2 | 1,450 |
| Fanatics | 9 | 12 | 21 | 1,312 |
| 4Caster | 8 | 13 | 8 | 118 |
| Polymarket | 7 | 9 | 2 | 1,107 |
Here again, Pinnacle is the book being followed. When Pinnacle leaves the market by itself, at least one other book is at its number within five minutes 60 to 68% of the time, and the median wait is under two minutes. Three or more books get there within five minutes about a quarter of the time, which is still well ahead of everyone else. DraftKings was second, with another book reaching its number within five minutes about 40% of the time, usually in two to four minutes.
Solo breakaways where the book went back to the old market first
| Book | Spread (%) | Total (%) | Moneyline (%) | Solo breakaways, all markets |
|---|---|---|---|---|
| Pinnacle | 5 | 9 | 7 | 948 |
| DraftKings | 19 | 13 | 25 | 658 |
| BetOnline | 21 | 13 | 17 | 232 |
| FanDuel | 26 | 22 | 29 | 2,159 |
| Caesars | 26 | 40 | 23 | 116 |
| Kalshi | 33 | 37 | 76 | 1,450 |
| Bookmaker | 45 | 39 | 25 | 4,710 |
| 4Caster | 45 | 43 | 38 | 118 |
| Novig | 49 | 44 | 58 | 1,866 |
| ProphetX | 49 | 31 | 52 | 550 |
| Polymarket | 54 | 43 | 70 | 1,107 |
| Fanatics | 59 | 37 | 31 | 1,312 |
But who cares about the movement itself if it doesn't stick. So how often did each book revert? Pinnacle almost never goes back, only 5 to 9% of the time depending on the market. Every other book reverts much more. Bookmaker went back 45% of the time on spreads, Novig 49% and Polymarket 54%, and on moneylines Polymarket reverted 70% of the time. Bookmaker is also the most active book in the study by a wide margin, with over 2,300 solo breakaways on spreads alone.
The books that most often moved together, to the same number within seconds of each other, were Bookmaker and Pinnacle, then DraftKings and Pinnacle, then Novig and Pinnacle. Pinnacle is still the winner in terms of market movers.
Refill Bots
This is a fairly new phenomenon with the rise of prediction markets. A refill bot puts out liquidity at a price, and when it gets filled, it refills it, sort of a way to feel out whether there's real interest at that number. This only happens on the exchanges, like Novig, ProphetX, 4Caster, Kalshi and Polymarket. The question we wanted to answer is whether a price that keeps getting refilled is ahead of the market. If someone keeps defending a number, does the market end up moving toward it?
We counted a price as defended if it got taken and came back at the exact same price within five minutes, at least twice. Then we tracked where the market went over the next 30 minutes and three hours, and compared it to prices that got taken once and didn't come back.
Exchange prices that kept getting refilled against prices taken once. Market move is in points toward the price over the next 30 minutes
| Book | Price vs market | Refilled prices | Taken once | Market move, refilled | Market move, taken once |
|---|---|---|---|---|---|
| Novig | At market | 5,102 | 41,123 | 0.03 | 0.05 |
| Novig | Better than market | 232 | 2,940 | 0.38 | 0.21 |
| Novig | Worse than market | 4,448 | 38,395 | 0.06 | 0.11 |
| ProphetX | At market | 4,992 | 40,370 | −0.03 | 0.01 |
| ProphetX | Better than market | 37 | 736 | 0.18 | 0.44 |
| ProphetX | Worse than market | 11,054 | 81,767 | 0.02 | 0.06 |
| 4Caster | At market | 3,813 | 157,169 | 0.01 | 0.03 |
| 4Caster | Better than market | 32 | 3,048 | 0.32 | 0.37 |
| 4Caster | Worse than market | 4,020 | 158,374 | 0.03 | 0.06 |
| Kalshi | At market | 1,638 | 5,147 | −0.04 | −0.01 |
| Kalshi | Better than market | 49 | 146 | 0.16 | 0.29 |
| Kalshi | Worse than market | 15,772 | 51,426 | 0.00 | 0.06 |
| Polymarket | At market | 1,632 | 11,439 | −0.04 | −0.02 |
| Polymarket | Better than market | 45 | 405 | 0.13 | 0.48 |
| Polymarket | Worse than market | 31,739 | 230,302 | 0.00 | 0.01 |
Most of the time, there's nothing to see. Refilled prices almost never sit better than the market. Even on Novig it was only 232 of about 9,800. Nearly all of them are right at fair or a little worse, which is where you'd expect a market maker to be. And on 4Caster, ProphetX, Polymarket and Kalshi, when a refilled price was better than the market, the market moved toward it less than it did for prices that got taken once and pulled.
Novig was the exception. When Novig keeps refilling a price that's better than the rest of the market, the market ends up moving toward that price. So if the market consensus is at 50 cents on a team and Novig is offering 48 cents, meaning the trader on the other side is taking 52, the market tends to move toward that 52 being fair value. The market moved about 0.38 points toward those refilled Novig prices in the next 30 minutes, compared to 0.21 for Novig prices that got taken once. It's a small move, but it's a consistent one. So a best in market price on Novig that keeps getting refilled may not always be the best thing to take, because the market tends to move toward the bot right after.
Where the Actual Value Is
The soft spots are pretty consistent across every test. BetOnline spreads and totals are the most profitable to bet into in the study, with the caveat about its feed. Fanatics is profitable to bet into on crosses in every market. Polymarket's side of a cross was profitable in every market, and Kalshi's moneylines are worth betting into while its spreads and totals aren't.
Bookmaker is slightly profitable on crosses, and it's the book most likely to move afterward, usually within about a minute when it does. So if you see a Bookmaker number crossed with Pinnacle, don't count on it being there long.
The Closing Line
Pinnacle is still the book that sets the market in college football. It doesn't post many games early and its early limits are small, but when it breaks away, another book is usually at its number within a couple of minutes, and it almost never backs off. It's also the toughest book to beat on crosses. 4Caster and ProphetX hold up well there too, but they don't lead moves the way Pinnacle does. DraftKings was the biggest surprise, sharper on CFB than its reputation, especially as a mover.
So when a CFB number moves, start with Pinnacle. If it broke away on its own, it's very unlikely to come back. If it's Bookmaker or FanDuel out there alone, there's a real chance that book is the one that moves back.
Alex Bow
Alex Bow trades tennis and baseball markets on Kalshi and Polymarket and writes about prediction market structure for SmartStake.
Every issue of The Closing Line →Subscribe / Weekly
Get the next issue in your inbox, before the line moves.
Free, unsubscribe anytime. By subscribing you agree to receive emails from SmartStake. See our Privacy Policy.