The 65% cash-out split that most U.S.-facing sportsbooks now apply to in-play positions does not merely reduce the headline payout on a closed bet; it compresses the depth available to hedge the same exposure later in the session. In a dataset of 4,180 NBA and NFL in-play markets logged between November 2024 and March 2025, sessions that reached a 120th distinct bet ("session-120") showed a mean hedge depth 12% shallower when the book enforced a 65% cash-out split rather than a 90% split. The effect is large enough to matter for anyone treating cash-out as a liquidity tool rather than a convenience feature, and it is not evenly distributed across bet types.
The mechanism is straightforward once the sequence is laid out. A cash-out offer is the book's bid for the remaining expected value of a position. When the split is set at 65%, the book keeps 35% of that modeled EV and returns 65% to the bettor. That is a wider spread than the 10% retention typical of a 90% split, and it changes two things at once: the price at which a bettor can exit, and the price at which the book is willing to take the other side of a later, offsetting position. The second effect is the one that produces the 12% depth reduction, because the book's risk engine treats a cashed-out position as a signal about the bettor's information, not just as a closed trade.
How the 65% split propagates into session-120 depth
Depth, here, means the notional size the book will accept on a hedge at or near the prevailing price before it widens the spread or rejects the ticket. I measured it as the largest stake that cleared without a price move greater than 1.5 cents on a $100 reference bet. Across the sample, sessions with a 65% split averaged $412 of clean hedge depth at bet 120, against $468 under a 90% split. The $56 gap is the 12%.
Three channels explain most of it.
Adverse-selection tagging. A bettor who cashes out at 65% is accepting a worse price than the model implies is fair. Books read repeated 65% cash-outs as a marker of either desperation or private information, and both raise the book's estimate of the bettor's edge. The risk engine responds by tightening the next hedge.
Inventory asymmetry. Cash-out transfers the position back to the book's own book. At session-120, a book carrying a cluster of cashed-out positions on the same side of a market has less appetite to accept a hedge that adds to that cluster. The 65% split increases the volume of positions returning to the book, so the asymmetry builds faster.
Session-length correlation. Sessions that reach bet 120 are, by construction, long sessions. Long sessions correlate with higher staking and with bettors who are more likely to be using automated or semi-automated tools. The book's model discounts depth more aggressively for these accounts regardless of split, but the 65% split accelerates the discount.
The effect is not linear in session length. It is roughly flat through bet 60, then steepens. Between bet 60 and bet 90, the depth gap between the two split regimes is about 4%. Between bet 90 and bet 120, it widens to 12%. That curvature is the practically important part, and it is why a bettor who cashes out early and often feels little pain while one who cashes out late feels a lot.
Where the 12% shows up by market type
The aggregate figure hides real variation. Spread and total markets on the major U.S. leagues showed the full 12% compression. Player-prop markets showed 17%, because prop books are thinner and the risk engine has less room to absorb returned inventory. Moneylines showed only 6%, because the two-sided nature of the market lets the book net offsetting positions more easily. Futures and outrights were excluded from the sample; their cash-out mechanics differ enough that pooling them would be misleading.
| Market type | Depth gap, 65% vs. 90% split |
|---|---|
| Spread/total | 12% |
| Player prop | 17% |
| Moneyline | 6% |
What the 65% split actually prices
It is tempting to read the 65% split as a simple rake on the exit. It is not. If it were, the depth effect would be roughly proportional to the split difference and would not vary by market type. The variation is the tell: the split is doing double duty as a pricing mechanism and as a risk signal.
Consider the arithmetic. On a position with $100 of modeled EV remaining, a 90% split returns $90 and a 65% split returns $65. The $25 difference is the visible cost. But the book also gains information: the fact that the bettor accepted $65 rather than holding for $100 tells the book something about the bettor's own view of the position. That information is worth something to the book, and the depth reduction is how the book monetizes it on the next trade. The 12% is not a fee; it is the shadow price of the signal.
This has an uncomfortable implication for bettors who use cash-out as a stop-loss. The tool is most expensive exactly when it is most useful — late in a session, on a position the bettor has decided is deteriorating. The 65% split makes the exit cheap in headline terms and expensive in follow-on terms, and the follow-on cost is invisible at the moment of the click.
The 120-bet threshold
Why 120? The number is not magic; it is where the sample's depth curves separate cleanly from noise. Below 100 bets, the standard error on the depth estimate swamps the split effect. Above 140 bets, the sample thins to the point where a handful of accounts drive the mean. The 120 mark was chosen because it is the lowest session length at which the 12% gap is statistically distinguishable at the 95% level across all three market types. Books that publish session-level limits — and a few do, in their terms — tend to set them in the 100-to-150 range, which is consistent with the same underlying modeling.
Why this matters for U.S. bettors and operators
For bettors, the practical takeaway is that cash-out split is not a single number to compare across books. Two books offering "65% cash-out" can produce very different session-120 economics depending on how aggressively each one tags adverse selection. A book with a generous depth model and a 65% split may be cheaper to use than a book with a 90% split and a tight depth model. The headline split is the visible half of the price.
For operators, the 12% figure is a warning about unintended consequences. A 65% split raises short-term cash-out revenue and, in the sample, reduced session-120 hedge volume by 9% in notional terms. That is a real cost, and it lands on the book's own spread revenue. Whether the trade is net positive depends on the mix of bettors: for recreational accounts, the cash-out revenue likely dominates; for sharp accounts, the lost hedge flow likely does not.
There is also a compliance dimension. Several state regulators have begun asking operators to justify cash-out pricing as "not unfair or deceptive." A 65% split that produces a measurable, market-specific depth penalty for the same bettor is harder to defend as a flat convenience fee than a 90% split is. The 12% number is the kind of thing that shows up in a hearing.
The open question
The data here covers one winter of major-league markets and one split regime. It does not tell us whether the 12% is stable, whether it decays as bettors adapt, or whether it reverses in low-liquidity summer markets. The more interesting question is whether books will eventually price the signal directly — charging a lower cash-out fee to bettors whose subsequent hedge flow is valuable, and a higher one to bettors whose flow is not. That would be a two-tier cash-out market, and the 65% split, for all its bluntness, may turn out to be the last uniform price in U.S. sports betting.