The Flat Handle Era: What the First Weeks of the NFL Season Mean for CPA vs. RevShare Deals

Legal U.S. sportsbooks are entering the first NFL season with a forecast of almost no growth in betting volume, and their margins fluctuate between 1.3% and 12.1%, depending on the week and the state. Affiliate managers are already rewriting their offers. 15M used data from regulators to calculate the hold levels at which it’s profitable for affiliates to opt for RevShare, when it’s better to stick with CPA, and what to adjust in a hybrid model.

Key Takeaways

• Handle is growing unevenly, and margins are volatile. Six weeks into the season (through October 18) in New York, mobile handle is up 8.5% while GGR is down 8.5%: the hold fell from 8.9% to 7.5%. A seventh week (through October 25, 12.1% hold) reverses the picture: over seven weeks, handle is up 9.6% and GGR up 6.7%. In New Jersey, football bets earned operators 1.3% of handle in September, compared to 4.8% a year earlier. In Pennsylvania, by contrast, revenue rose by 50% despite a 6.1% decline in handle: in September 2025, the hold was 3.5%;
• For the operator, CPA becomes more expensive on its own. With a lifetime handle of $12,000 per player, a CPA of $250 equals 35% of NGR at a hold of 9% and 63% at a hold of 6.3% (15M calculation). This creates pressure: hybrid or RevShare for everyone except verified sources;
• For affiliates, a 30% RevShare outperforms a $250 CPA only when the hold is above approximately 9.9% (15M calculation). With the hold in fall 2026, CPA comes out on top, while the hybrid model protects both sides. Meanwhile, prediction markets pay affiliates on a CPA basis – Better Collective reports a 45% rise in CPA Publishing revenue, mainly due to them.

What the Forecasts Say

As early as September 4, the AGA estimated that wagers on the 2026 NFL season at legal sportsbooks totaled $29.5 billion – a 0.3% increase from last year’s $29.4 billion. This marks the weakest growth since the repeal of PASPA in 2018. AGA President Bill Miller attributed the slowdown in growth to “prediction markets,” where, according to Covers, sports contracts account for about 80% of the business of the largest platforms.

On September 9, H2 Gambling Capital issued an even more conservative forecast: $31.4 billion in handle for August–February, down 0.8% from the 2025–26 season, with underlying demand up 4% and $0.6 billion from Missouri’s first full season. Gross win: $3.5 billion, down 2%. According to H2’s estimates, approximately $2.0 billion in handle has already shifted from licensed bookmakers to prediction-market platforms, primarily Kalshi. AGA and H2 use different bases ($29.5 billion vs $31.4 billion); compare only the direction.

15M checked these forecasts in the article “NFL Month: More Bets, Less Money.” What interests us here isn’t the audience, but how flat volume and fluctuating margins change a player’s value to the operator – and, consequently, the terms for affiliates.

What Regulators Are Saying

New York’s weekly reports for mobile sports betting reveal that the state’s opening six weeks of the 2026 season (weeks ending September 13 through October 18) have seen $3.59 billion in handle, up from $3.31 billion for the opening six weeks of the 2025 season (weeks ending September 7 through October 12). However, Gross Gaming Revenue (GGR) has gone down to $268.7 million from $293.6 million: over six weeks, players bet more, but operators earn less; with the seventh week, the picture reverses. The seventh week (through October 25) delivered $656.2 million in handle and $79.4 million in GGR (12.1% hold vs 5.8% a year earlier); over seven weeks, handle is +9.6%, GGR +6.7%, hold 8.2% vs 8.4%. (calculated by 15M based on the NYSGC report).

The key feature of the data is the spread of weekly hold rates, which for 2026 have been 2.9%, 6.3%, 8.5%, 7.9%, 9.0%, and 10.3%. For 2025, hold rates were 5.8%, 10.7%, 10.2%, 9.1%, 5.1%, and 12.4% for the same period. It’s also worth noting that the first week of 2026 made $594.6 million in handle but only $17.0 million in GGR. This could mean RevShare affiliates who earned commissions for players they referred to the sites will have seen September payouts for newly signed-up players fall by almost half (GGR −44.5%) and, after promotions, to roughly zero.

In New Jersey, September handle was down 3.6% YoY to $1.086 billion. Operator revenue on those events sank 16.7% to $68.1 million as hold rates fell to 6.3% (from 7.3%). Handle on football fell 15.3% to $318.8 million, with operator winnings plunging 76.4% to $4.3 million (and hold down to 1.3% from last year’s 4.8%). Parlays accounted for $380.6 million of handle at an 11.8% hold. Online sportsbook gross revenue in New Jersey fell 18.8% in September (online casino rose 13.0%), with FanDuel down 34.4% and DraftKings down 5.9%.

Pennsylvania’s September handles slipped 6.1% to $798.7 million, but operators saw revenues surge 50.5% YoY to $44.7 million (up from $29.7 million in September 2025). The revenue jump came not from an expanding market but from a higher hold (5.6% against 3.5%; 15M calculation from PGCB data. PGCB reports taxable revenue (after promotional deductions), so this hold is not strictly comparable with GGR hold in New York and New Jersey). Illinois had not published its September report as of November 4; as of August 2026: handle of approximately $1.01 billion, up 0.5% year-over-year (according to an IGB-based summary); in June, handle rose 20%, while for the first half of the year it fell 2%.

Table 1. Handle, Revenue, and Hold in Three States, September–October 2026 vs. 2025. Sources: 1 (New York, 15M calculation based on weekly data; 2025 – weeks ending September 7–October 19), 2 (New Jersey, statistics for completed events, unaudited), 3 (Pennsylvania; hold – calculated by 15M).

What the Operators Say

Quarterly reports paint the same picture. DraftKings reported $13.1 billion in wagers and contracts for the second quarter, up 15%, with revenue of $1.443 billion, down 5%. Revenue from sports betting fell from $997.9 million to $891.9 million, and sports net revenue margin fell from 8.7% to 6.8%. The number of unique paying players increased by 9% to 3.6 million, but average revenue per player decreased by 13% to $132. Adjusted EBITDA was $114.6 million, compared to $300.6 million.

“Predictions is already growing faster than we anticipated,” DraftKings CEO Jason Robins explained. The company put the revenue shortfall down to “customer-friendly sport outcomes and increased promotional reinvestment associated with new customer acquisition”.

FanDuel’s quarterly handle rose 2% to $11.96 billion, sportsbook active players rose 8% (9% across all U.S. products), and sports revenue fell 15% to $1.039 billion. Flutter expects U.S. third-quarter adjusted EBITDA of about break-even, and cut its 2026 U.S. revenue outlook by $460 million to $7.4 billion, mostly for planned investment in product and promotions ($385 million of revenue) and the NFL start shift ($75 million). EBITDA forecasts were trimmed by $210 million to $760 million. Net revenue margin has slipped 170 basis points to 8.7%, 70 bps (compared to last year) from match outcomes and 140 bps from higher promotions, partly offset by +40 bps of structural margin (more same-game parlays). All while marketing spend ballooned 61% to $353 million, and promotions climbed 140 basis points to 5.4% of the handle, or the total amount bet.

“The momentum we are seeing gives us confidence to increase investment in the second half,” wrote Peter Jackson, then CEO of Flutter, to shareholders.

The first third-quarter report came from BetMGM on October 13, and it ended with $710 million in net revenue, up 7% YoY. Online sports generated $216 million in revenue, up 7% YoY, on $3.24 billion in handle, up 3% YoY. Average monthly actives across all products, including iGaming, were 878,000, down 3% from 902,000; BetMGM attributes this to its “refined player acquisition and management” strategy, meaning it is being more judicious in who it brings on board. However, net revenue per active sports bettor was up 18% YoY, and handle per player was up 13% YoY. Gross gaming revenue margin was 10.7%, up 55 bps YoY.

“BetMGM delivered a strong Q3 as we continue to execute with discipline and generate profitable growth,” said BetMGM CEO Adam Greenblatt.

DraftKings and FanDuel have been spending heavily on player acquisition promos and, as such, have been seeing their margins compressed. BetMGM, meanwhile, has been focusing on extracting more value per player. For affiliates, DraftKings and FanDuel share one denominator: the acquired player now brings less than CPAs were priced at a year ago. BetMGM is the exception, with revenue per player up and a 10.7% hold.

Prediction Markets: Where Is the Volume Going, and Who Pays Affiliates?

Prediction market volume on the weekend of Oct. 17–18 reached $11 billion (according to Aldrin Research, as reported by Covers), making it the highest recorded weekend in the space. Kalshi had $8.51 billion (according to 15M’s calculations, approximately 72% of the $11.8 billion total volume over the two days). According to Covers’ share of its base, it is 58.4% ($8.51 / 0.584 ≈ $14.6 billion).

DraftKings Predictions saw $567.8 million of volume, while Polymarket in the U.S. saw roughly $1 billion. Total NFL volume on Sunday, October 18, was more than $870 million, with Kalshi seeing $497.2 million, down from $575 million the week prior. It’s worth noting that this is nominal volume and includes parlays, which, that Sunday alone, were $3.7 billion across all sports and platforms, and is not the same thing as handle.

Affiliates are more interested in something else: prediction sites are already buying traffic on a Cost Per Action (CPA) basis. In its second-quarter report, Better Collective states that CPA revenue in North America rose 50% to €5 million, driven primarily by the prediction markets (group: +11% to €19 million), “mainly reflecting strong customer acquisition demand from prediction market partners in North America”; elsewhere the report says partner agreements with them “are structured on a CPA basis”.

The public terms and conditions on the platforms themselves currently specify referral programs rather than affiliate programs. Kalshi states in its help section that commission amounts “vary by offer,” apply only to users in the U.S., and that the code can be linked within 72 hours of registration. Since May 28, 2026, Polymarket pays 10% of the net trading commissions it makes from direct referrals and 5% from indirect referrals, but affiliates will only receive commissions after they generate $10,000 of turnover in total, and commissions are paid only for the first 30 days after a referral registers (or until they reach the Platinum tier). In effect, this is a short-term revenue share based on commissions, not on the player’s losses: if there are no trading fees, there is nothing to pay the affiliate from.

Three Models With a Flat Handle

In a growing handle market, the operator will pay a CPA because a new player will pay for itself. Conversely, an affiliate will accept CPA because money now is worth more than money later. When handle isn’t growing and hold fluctuates, the situation is thornier.

Let’s run the numbers for a single player (15M calculation). Assumptions: the player’s lifetime handle is $12,000 over three years; promotions and bonuses amount to 3% of the handle (an assumption; in Q2 FanDuel’s promotions were 5.4% of handle, per Flutter’s report. At 5.4% promotions and a 7.5% hold, NGR per player is $252, so a $250 CPA is almost 100% of NGR and the RevShare threshold rises to ≈12.3%); RevShare of 30% of NGR, where NGR is GGR minus promotions (for simplicity, we do not deduct taxes and payment processing fees, although New York operators pay a 51% tax on GGR, which many contracts deduct from the RevShare base), CPA of $250 (above Track360’s $120–220 range), and a hybrid model – $125 + 15% NGR (per Track360, hybrids run $80–120 + 15%). We take the hold figures from Table 1: 8.9% (≈9.0%) – New York, 2025 season; 7.5% – New York, six weeks of the 2026 season; 8.2% – New York, seven weeks of the 2026 season; 6.3% – New Jersey, September 2026.

Table 2. Affiliate payout per player over three years at different hold levels. Calculation: 15M; assumptions are explained in the text. Hold – sources 1 and 2.

Under the conditions mentioned above, with a fall 2026 hold rate, a 30% RevShare would pay affiliates 25–52% less than the CPA, and the break-even point is around 9.9% – above the average hold for any of the three markets during the period (5.6–8.2%). From an operator’s perspective, the CPA is looking less and less attractive too: at $250 it represents 35% of Net Gaming Revenue with a hold of 9% and 63% with a hold of 6.3%. So it’s not surprising affiliate managers are cutting CPAs – not out of greed: at the previous rate, at a 6.3% hold, they were paying affiliates more than half of the player’s three-year NGR. This is why a hybrid agreement is attractive: it reduces risk for the operator and affiliate. The cost for the operator is not that great compared to a CPA agreement when the hold rate is high, but it does mean affiliates get only 74–82% of what they’d get under CPA with a low hold.

Added to this is the week-to-week variation. With a hold of 2.9% in the first week of the season, the NGR according to our model is negative: the 3% promotional handle exceeds the GGR. If the contract allows for the carryover of a negative balance, a RevShare affiliate starts the season slightly in the red. If the balance is reset to zero every month, the risk lies entirely with the operator, who factors it into the RevShare rate.

Pressure From the Operator and the Affiliate’s Response

Better Collective, one of the industry’s biggest affiliate groups, is already mostly on RevShare. Figures compiled by Better Collective show that in the second quarter, 70% of its 373,000 new depositing customers (NDC) were recruited on a RevShare basis. RevShare contracts account for 49% of Better Collective’s income while CPA contracts account for 22%. An elevated sports margin (“above the normalized level”) added about €3 million to RevShare income, and falling hold would take it away. Track360 compiled research from a sample of over 40 affiliate programmes (the sample is biased towards Europe and Latin America programmes, and is not truly random) to discover that in 2026 33% of new contracts pay affiliates via RevShare, 26% via CPA and 41% with hybrid contracts.

What an operator typically asks for when facing a flat handle: to switch the deal to a hybrid model, lower the CPA in states with high taxes, add player qualification requirements (minimum deposit and number of bets before the CPA is credited), and implement a negative balance carryover. In response, it makes sense for the affiliate to ask for:

• RevShare basis. Specify in the contract what is deducted from GGR: promotions, taxes, payment processing fees, and “admin fee.” In New York, with a 51% tax rate, the difference between RevShare calculated based on GGR and on NGR after tax is several-fold: $900 versus $81 per player at a 7.5% hold and 3% promotions;
• Clear the negative balance on a monthly or at least quarterly basis. A week with a 2.9% hold should not wipe out the month’s payment;
• The CPA window in a hybrid model. If an operator lowers the CPA, ask for payment upon the first deposit rather than after the 30-day qualification period – with a 5.4% handle promotion, a player who doesn’t qualify still costs the operator money;
• The right to adjust the rate when the hold is below the threshold. Linking RevShare to the actual quarterly hold is rare, but it is precisely this that shifts the debate from “who is to blame” to a matter of arithmetic;
• A separate CPA offer for prediction products, if the operator has one. DraftKings Predictions and FanDuel Predicts are part of the same apps, but they operate on a commission-based model, and the RevShare from the bookmaker’s NGR doesn’t quite reach that level.

What This Means

To affiliates. In the fall of 2026, pure RevShare in the U.S. rests on a hold that was below 8% in three weeks out of seven (2.9%, 6.3%, and 7.9%) and reached 12.1% in the seventh. If an operator wants to negotiate, look for a hybrid model and a CPA that is no less than half what you were getting before, RevShare based on NGR and a clearly defined list of deductions, and diversify by product. Some traffic to prediction sites is CPA-based, and Better Collective says prediction-market partners buy traffic on CPA; their rates are not public, so compare them against your own RevShare numbers.

To operators. It makes sense to put pressure on affiliates’ CPAs, but there is a limit: a CPA above $200 takes more than half of NGR ($396 at a 6.3% hold); Track360 puts U.S. sportsbook CPAs at $120–220 and the regulated-market average at $145. A better deal would be a hybrid model, including player qualification requirements, monthly balance resets, and a fair NGR basis. It’s cheaper than losing a traffic source (at Better Collective, 70% of new deposits in Q2 came from RevShare agreements).

To networks and aggregators. An affiliate who is privy to the ins and outs of its cohort’s handle, gross gaming revenue, and promotions weekly will be far more comfortable jumping on board the RevShare wagon than one who simply receives a single lump sum at the end of the month. Flutter (November 4), DraftKings (November 5), and Better Collective (November 18) report on Q3, which covers only about three weeks of the NFL season; October data comes with Q4 results.

Methodology

Data as of November 4, 2026. The New York data comes from the NYSGC’s weekly mobile betting report, and the file contains seven weeks of the current season (through October 25); Tables 1–2 use the first seven weeks ending September 13 to October 25; the data covers all sports. New Jersey data comes from the DGE and is unaudited data in the table “Sports Wagering Completed Events Statistics” from the DGE September press release; New Jersey’s unaudited data is different than its cash-basis gross revenue figure of $77.0 million. Pennsylvania data comes from the PGCB press release, and the hold there is estimated at 15M. The forecast market volumes are nominal, from Aldrin Research via Covers, and not all platforms are done reporting yet. Table 2’s estimates are based on a model with certain assumptions, not the industry averages, and NGR there does not include taxes or payment fees. Track360’s numbers are vendor estimates. Quotes are given in the original.

Sources

1. New York State Gaming Commission, “Mobile Sports Wagering Statewide Data” (weekly report), the version valid through October 25, 2026. https://gaming.ny.gov/statewide-sports-wagering-weekly-report-pdf
2. NJ Division of Gaming Enforcement, “DGE Announces September 2026 Gaming Revenue Results”, October 19, 2026 — $77.0 million, −14.3%; online −18.8%; the table of completed events. https://www.nj.gov/oag/ge/docs/Financials/PressRelease2026/September2026.pdf
3. Pennsylvania Gaming Control Board, “PA Gaming Control Board Reports Revenue for September”, October 19, 2026 — $798.7 million, −6.09%; $44.7 million, +50.45%. https://gamingcontrolboard.pa.gov/news-and-transparency/press-release/pa-gaming-control-board-reports-revenue-september