2.1 Battlefield Evolution Since the June Report

The largest structural changes are in Texas and Ohio, which have crossed from the June Republican-leaning firewall to narrow Democratic advantages in the current forecast. Alaska and Texas are effectively centered at the toss-up boundary, Iowa remains narrowly Republican, and the Democratic side is stronger in Michigan, North Carolina and Georgia than it was in the June narrative.

2.2 Geographic Map of the 35 Senate Contests

2.3 State-Level Projection Table for the 35 Senate Contests

For Nebraska only, the Democratic-aligned opposition slot can correspond to an Independent candidacy under the seat-control convention; when that slot is projected to win, the public label is shown as IND rather than DEM.

2.4 Matrix of the 35 Contests — Projected Competitiveness

2.5 Where Markets and the Model Diverge Most

This table is included because disagreement can be informative. It shows where public betting-market expectations differ most from BUNKER’s current probability estimate. The market number is not a second projection and does not replace the official forecast shown above.

3.1 State Map of District-Level Projections

3.2 Territorial Matrix of All 435 Districts

3.3 House Competitiveness — Official Projection by Tier

These are BUNKER presentation bands applied to the official expected Democratic vote; they are not external race ratings.

4.1 Chamber Seat Distributions

4.2 Interdependence and Balance of Power

The joint distribution preserves a common national shock across both chambers. It is intended to show how control outcomes move together, not to create a separate race-level forecast.

4.3 Empirical Core: Intervals and Risk Boundaries

5.1 Historical National Midterm Turnout and 2026 Forecast

BUNKER’s current 50-state VEP-weighted turnout forecast is 47.3%, with an empirical 80% predictive bracket of 43.6–50.4%. The weighting is national in the sense relevant for the state model: each state’s projected turnout is weighted by its share of the voting-eligible population rather than giving small and large states equal influence. Because 2026 VEP is not yet observable and the operational VEP_STATE_2026 vector was not persisted numerically in the recovered turnout artifacts, this public aggregation uses 2024 state VEP — the latest observed Election Lab state weights — as the transparent weighting vintage rather than silently substituting VAP or CVAP.

Higher turnout does not mechanically imply a better result for either party. Its value here is to show whether the 2026 forecast is being generated in a historically low-, normal-, or high-participation midterm environment before examining where participation is concentrated.

5.2 State-Level 2026 Turnout Map

The state map shows why a national average is only the starting point: the modeled participation environment varies materially across the country. Those differences matter most where control is being decided at the margin.

5.3 Turnout in Competitive Senate States

This comparison focuses on the states carrying the greatest leverage in the current Senate control distribution. It is intended to show the participation environment around the battleground, not to imply that turnout by itself determines the winner.

5.4 Turnout Across Competitive House Districts

The district chart shows all 12 Pure Toss-Up districts and deliberately uses a relative turnout index rather than presenting a fabricated district VEP percentage. A value above 100 means the district is projected to participate more strongly than its state’s CVAP-weighted district average; below 100 means the opposite. This lets the public see where turnout intensity may reinforce or weaken the most competitive House terrain without mixing incompatible denominators.

6.1 Historical Backtesting Validation Report

|Metric                       |  Result   |
|:----------------------------|:---------:|
|Historical races tested      |   2,722   |
|Correct final projections    |   2,594   |
|Final-forecast call accuracy |   95.3%   |
|Validation span              | 2012–2024 |

|Metric                       |  Result   |
|:----------------------------|:---------:|
|Historical races tested      |    241    |
|Correct final projections    |    223    |
|Final-forecast call accuracy |   92.5%   |
|Validation span              | 2012–2024 |

6.2 Monte Carlo Stability

6.3 Stability Curve

6.4 Uniform District-Center Vote-Shift Stress Test

This stress test asks a simple counterfactual question: how far would the current expected-vote centers have to move uniformly toward either party before the central seat count changes control? For the House, the exercise applies the same vote shift to all 435 district expected-vote centers and then recounts which side is above 50% in each district. For the Senate, the same uniform shift is applied to the current state-race centers and the central seat count is recomputed. The horizontal axis therefore represents a uniform movement in race-level vote centers; it is not a generic-ballot estimate and it is not a probability forecast.

In the House, the first Republican-side control boundary appears at roughly a 2.24-point uniform shift. At that point, enough Democratic-projected districts cross below 50% for the central seat count to fall below the 218-seat majority threshold. This is broader than a sweep of the 12 Pure Toss-Up districts: if Republicans won all 12, Democrats would still hold 221 seats to 214 Republican seats, because only 5 of those 12 are currently Democratic calls. Republicans would still need at least 4 additional Democratic-projected districts outside the Pure Toss-Up band to take control. The uniform-shift exercise reaches those additional districts as well.

For the Senate, the current central majority is lost after roughly a 0.62-point Republican shift, while about a 0.42-point Democratic shift moves the central projection to at least 52 Democratic/aligned seats. These thresholds measure sensitivity around the current forecast, not the probability that such a national shift will occur.

6.5 Current Forecast Coverage