Zero Pundit Priors. Zero Secret Pollster Grades. Pure Sample-Size Pooling, Empirical Variance, & Exact Convolution.
All survey schemas (pollster, date, sample_size, population, candA_pct, candB_pct) follow the public open-data standard established by open-access election archives:
pres_polls.csv, senate_polls.csv, generic_ballot_polls.csv) used for our 2024 7-Swing-State Backtest. LV) screens, and margin-of-error disclosure. LV) surveys fielded Oct 20 β Nov 3, 2024 across PA, MI, WI, GA, NC, NV, and AZ, paired with official State Secretary of State / FEC certified popular vote margins. π₯ 100% Visitor-Auditable & Replaceable: Nothing is locked. Click π₯ Paste / Export Raw CSV in the Poll Table below to export, inspect, edit, or paste any live CSV polling table from any open-source repository.
A+ vs. C- multipliers). i is weighted strictly by its disclosed sample size n_i and age Ξt_i: w_i = n_i Β· 2^(βΞt_i / Ο), normalized into a Lean 4 ProbStats.FiniteDist M distribution. Baseline non-battleground Senate seats: 45 D/Ind vs 46 R, leaving 9 competitive battlegrounds (NC, ME, MI, OH, GA, TX, AK, IA, NE). Reaching 50 seats requires winning 5 of 9; outright 51-seat majority requires 6 of 9.
Click any race below to inspect its open-source polls, bell-curve density, and step-by-step mathematical conversion.
N_pool = 4,350 voters across 4 polls β ΞΌΜ = β(w_i Β· m_i) / βw_i = +2.39% Ο_stat = 100/βN = 1.52% | s_polls = 2.22% | Ο_total (Tier 2) = 3.97% z = ΞΌΜ / Ο_total = (2.39%) / (3.97%) = +0.601Ο P(James Talarico (D) Wins) = ZTable4Dec(z=+0.60) = .7257 = 72.57% (vs Ken Paxton (R): 27.43%) Actual o = 1 (James Talarico (D) Won) β BS = (p β o)Β² = (0.7257 β 1)Β² = 0.0752 (β
Beats 0.2500 Coin Flip) Uncheck any poll to test outlier sensitivity, edit numbers inline, or add/paste raw CSV polling data below.
| On | Open-Source Pollster & Methodology | Date (Ξt) | Sample (n_i & MoE) | James Talarico (D) | Ken Paxton (R) | Margin (m_i) |
|---|---|---|---|---|---|---|
| Live Phone (Cell + Landline RDD) & Voter-File ABS | Oct 28βNov 1 (2d) | Β±2.9% | % | % | +4.0% | |
| Matched Online Panel (Propensity + Post-Stratified) | Oct 25β30 (4d) | Β±2.8% | % | % | +4.0% | |
| Mixed-Mode IVR + SMS-to-Web + Online Panel | Oct 22β27 (7d) | Β±3.2% | % | % | +1.0% | |
| UT / YouGov Statewide Matched Internet Panel | Oct 16β22 (12d) | Β±3.0% | % | % | -0.5% |
After an election, anyone can say "I knew who would win!" The Brier Score (Glenn W. Brier, 1950) grades honest probabilities using simple subtraction and squaring β zero calculus required. Graded below against the actual November 3, 2026 election results (Wikipedia / AP / state election boards; click any winner pill to test counterfactual upsets).
(1/10) β (p_i β o_i)Β² β’ β
Beats 0.2500 Coin Flip! BSS = 1 β (BS / 0.2500) β’ Lower BS = Higher Skill 1/βN upset blowups. BS_i < 0.2500 Suppose a forecaster gives Candidate A a probability p (between 0.00 and 1.00) of winning. After the votes are counted, the actual outcome o is either 1 (Candidate A won) or 0 (Candidate A lost). The Brier Score is simply the squared distance between the forecast and reality:
Brier Score = (p β o)Β² | Aggregate BS = (1/R) β (p_i β o_i)Β²0.0000 (Crystal Ball β Perfect): You said 100% (p = 1.00) and they won (o = 1): (1.00 β 1)Β² = 0.0000. 0.0400 (Strong 80% Forecast Wins): You said 80% (p = 0.80) and they won (o = 1): (0.80 β 1)Β² = (β0.20)Β² = 0.0400. 0.2500 (Blind 50β50 Coin Flip): If you guess 50% (p = 0.50) on every race, your score is always(0.50 β 1)Β² = (0.50 β 0)Β² = 0.2500. Any model below 0.2500 beats flipping a quarter! 1.0000 (100% Confidently Wrong): You screamed "100% Lock!" (p = 1.00) and they lost (o = 0): (1.00 β 0)Β² = 1.0000. Squaring the error punishes fake certainty! A good model shouldnβt call every race 99% or 50%. Calibration means the probabilities match real-world win rates: when a weather forecaster says "70% chance of rain," it should rain on about 7 out of 10 such days!
| Forecast Bucket | Races | Avg Forecast (pΜ) | Actual Win Rate (Ε) |
|---|---|---|---|
| Underdog / Lean B (P < 50%) | 1 | 47.2% | 0.0% (0/1) |
| Toss-Up / Lean A (50%β75%) | 4 | 67.0% | 100.0% (4/4) |
| Likely / Safe A (P β₯ 75%) | 5 | 93.8% | 100.0% (5/5) |
brierScore_complement_symm):(p β o)Β² or Candidate B ((1βp) β (1βo))Β² gives the exact same Brier Score! Why do pundits who only look at sample size 1/βN (Tier 0) get destroyed by the Brier Score? Because a +2.0% poll lead on N = 4,000 voters looks like a 94% lock under raw 1/βN. When a normal 3% statewide polling miss flips that toss-up, Tier 0 takes a huge (0.94 β 0)Β² = 0.8836 penalty!
| Uncertainty Conversion Tier | Aggregate Mean BS | Skill vs. Coin Flip |
|---|---|---|
Tier 0: Raw CLT (1/βN Only) | 0.0397 | 84.1% |
Tier 1: Poll Dispersion (s_polls) | 0.0491 | 80.3% |
Tier 2: + Systemic Error (Ο_sys) | 0.0712 | +71.5% |
| Race | Pooled Poll Margin (ΞΌΜ Β± Ο) | Pre-Election Forecast P(Win) | Post-Election Winner o_i (Click to Toggle What-If) | Tier 0 BS (1/βN) | Tier 1 BS (s_polls) | Tier 2 BS (+Ο_sys) | Active Brier Score (p_i β o_i)Β² | Grade vs. Coin Flip (0.2500) |
|---|---|---|---|---|---|---|---|---|
| US-HOUSEU.S. House Generic Ballot | D +8.73% (Β±3.71%) | D: 99.1% / R: 0.9%p_i = 0.9906 | 0.0000 | 0.0000 | 0.0001 | β BEATS COIN FLIP | ||
| NCNorth Carolina Senate | D +8.12% (Β±4.03%) | D: 97.8% / R: 2.2%p_i = 0.9783 | 0.0000 | 0.0000 | 0.0005 | β BEATS COIN FLIP | ||
| MIMichigan Senate | D +6.05% (Β±4.06%) | D: 93.2% / R: 6.8%p_i = 0.9319 | 0.0000 | 0.0003 | 0.0046 | β BEATS COIN FLIP | ||
| MEMaine Senate | D +4.18% (Β±4.14%) | D: 84.4% / R: 15.6%p_i = 0.8438 | 0.0001 | 0.0081 | 0.0244 | β BEATS COIN FLIP | ||
| OHOhio Senate Special | D +2.66% (Β±3.97%) | D: 74.9% / R: 25.1%p_i = 0.7486 | 0.0027 | 0.0202 | 0.0632 | β BEATS COIN FLIP | ||
| GAGeorgia Senate | D +6.45% (Β±4.06%) | D: 94.4% / R: 5.6%p_i = 0.9441 | 0.0000 | 0.0000 | 0.0031 | β BEATS COIN FLIP | ||
| TXTexas Senate | D +2.39% (Β±3.97%) | D: 72.6% / R: 27.4%p_i = 0.7257 | 0.0034 | 0.0349 | 0.0752 | β BEATS COIN FLIP | ||
| AKAlaska Senate | D +1.40% (Β±4.17%) | D: 63.3% / R: 36.7%p_i = 0.6331 | 0.0665 | 0.0808 | 0.1346 | β BEATS COIN FLIP | ||
| IAIowa Senate | D +0.76% (Β±4.13%) | D: 57.1% / R: 42.9%p_i = 0.5714 | 0.1239 | 0.1460 | 0.1837 | β BEATS COIN FLIP | ||
| NENebraska Senate | R +0.27% (Β±4.07%) | I: 47.2% / R: 52.8%p_i = 0.4721 | 0.2010 | 0.2010 | 0.2229 | β BEATS COIN FLIP | ||
π AGGREGATE MEAN BRIER SCORE Across All 10 Races:meanBrierScore = (1/10) β (p_i β o_i)Β² | 0.0397 | 0.0491 | 0.0712 | 0.0712 Skill BSS: +71.5% vs Coin Toss | π AGGREGATE CALIBRATED (< 0.25) | |||
How we teach students to go from finite coin/dice sample spaces to the Bell Curve and look up exact 4-decimal probabilities (.XXXX) using only addition, subtraction, and the Complement Rule.
Every probability experiment begins with a finite Sample Space S and events E β S (formalized in Lean 4 as ProbStats.eventProb P E = β x β E, P.pmf x):
0 β€ P(E) β€ 1. (0 = Impossible, 1 = Certain, P(E) β€ 0.05 = Unusual event). β P(x) = 1 across all outcomes in S. A and B cannot happen together (P(A and B) = 0), then P(A or B) = P(A) + P(B). kolmogorov_complement):P(E') = 1 β P(E) βΊ P(E) + P(E') = 1. This single rule lets us find right-tail areas without ever integrating! Known to Jia Xian (~1050), Yang Hui (1261), Zhu Shijiei (1303), and Blaise Pascal (1654), the n-th row of Pascalβs triangle counts combinations nCx = n! / (x!(nβx)!) and sums to 2^n. Graphing Row 10 (sum 2^10 = 1024) reveals a Bell Curve from pure counting:
BINS Binomial: ΞΌ = nΒ·p, ΟΒ² = nΒ·pΒ·q, Ο = β(nΒ·pΒ·q) When can we replace discrete counting with the Standard Normal z-table? In Introductory Statistics, we check the CLT Gatekeeper:
nΒ·p = 20(1/6) = 3.33 < 5 β Fails Gatekeeper! Do not use the normal distribution. NΒ·pΜ_A = 2,146 β₯ 5 & NΒ·qΜ_A = 2,204 β₯ 5 β Passes CLT Gatekeeper by 435Γ!Standardize without Calculus:z = (x β ΞΌ) / Ο, x = ΞΌ + zΒ·Ο, Ο_xΜ = Ο / βn, Ο_pΜ = β(pΜΒ·qΜ / n).
| data (TX Polls) | fence (2-Tailed Tie Test Hβ: ΞΌ = 0) |
|---|---|
xΜ (margin) = +2.39% | xΜ_c = Β±7.79% |
z = +0.60 | z_c = Β±1.96 |
P (1-tail loss) = .2743 | Ξ±/2 = .0250 (Ξ± = 0.05) |
P(Z < +0.60) = .7257(72.57%)P(Z > +0.60) = 1.0000 β .7257 = .2743(27.43%)zTable_complement_rule), or look up βz = -0.60 β .2743. P(-0.60 < Z < +0.60) = .7257 β .2743 = .4514 (45.14%)zTable_three_region_partition). P(Z < z) β Positive z (0.00 to 3.49 β’ 4 Decimals Only)| z | .00 | .01 | .02 | .03 | .04 | .05 | .06 | .07 | .08 | .09 |
|---|---|---|---|---|---|---|---|---|---|---|
| 0.0 | .5000 | .5040 | .5080 | .5120 | .5160 | .5199 | .5239 | .5279 | .5319 | .5359 |
| 0.1 | .5398 | .5438 | .5478 | .5517 | .5557 | .5596 | .5636 | .5675 | .5714 | .5753 |
| 0.2 | .5793 | .5832 | .5871 | .5910 | .5948 | .5987 | .6026 | .6064 | .6103 | .6141 |
| 0.3 | .6179 | .6217 | .6255 | .6293 | .6331 | .6368 | .6406 | .6443 | .6480 | .6517 |
| 0.4 | .6554 | .6591 | .6628 | .6664 | .6700 | .6736 | .6772 | .6808 | .6844 | .6879 |
| 0.5 | .6915 | .6950 | .6985 | .7019 | .7054 | .7088 | .7123 | .7157 | .7190 | .7224 |
| 0.6 | .7257 | .7291 | .7324 | .7357 | .7389 | .7422 | .7454 | .7486 | .7517 | .7549 |
| 0.7 | .7580 | .7611 | .7642 | .7673 | .7704 | .7734 | .7764 | .7794 | .7823 | .7852 |
| 0.8 | .7881 | .7910 | .7939 | .7967 | .7995 | .8023 | .8051 | .8078 | .8106 | .8133 |
| 0.9 | .8159 | .8186 | .8212 | .8238 | .8264 | .8289 | .8315 | .8340 | .8365 | .8389 |
| 1.0 | .8413 | .8438 | .8461 | .8485 | .8508 | .8531 | .8554 | .8577 | .8599 | .8621 |
| 1.1 | .8643 | .8665 | .8686 | .8708 | .8729 | .8749 | .8770 | .8790 | .8810 | .8830 |
| 1.2 | .8849 | .8869 | .8888 | .8907 | .8925 | .8944 | .8962 | .8980 | .8997 | .9015 |
| 1.3 | .9032 | .9049 | .9066 | .9082 | .9099 | .9115 | .9131 | .9147 | .9162 | .9177 |
| 1.4 | .9192 | .9207 | .9222 | .9236 | .9251 | .9265 | .9279 | .9292 | .9306 | .9319 |
| 1.5 | .9332 | .9345 | .9357 | .9370 | .9382 | .9394 | .9406 | .9418 | .9429 | .9441 |
| 1.6 | .9452 | .9463 | .9474 | .9484 | .9495 | .9505 | .9515 | .9525 | .9535 | .9545 |
| 1.7 | .9554 | .9564 | .9573 | .9582 | .9591 | .9599 | .9608 | .9616 | .9625 | .9633 |
| 1.8 | .9641 | .9649 | .9656 | .9664 | .9671 | .9678 | .9686 | .9693 | .9699 | .9706 |
| 1.9 | .9713 | .9719 | .9726 | .9732 | .9738 | .9744 | .9750 | .9756 | .9761 | .9767 |
| 2.0 | .9772 | .9778 | .9783 | .9788 | .9793 | .9798 | .9803 | .9808 | .9812 | .9817 |
| 2.1 | .9821 | .9826 | .9830 | .9834 | .9838 | .9842 | .9846 | .9850 | .9854 | .9857 |
| 2.2 | .9861 | .9864 | .9868 | .9871 | .9875 | .9878 | .9881 | .9884 | .9887 | .9890 |
| 2.3 | .9893 | .9896 | .9898 | .9901 | .9904 | .9906 | .9909 | .9911 | .9913 | .9916 |
| 2.4 | .9918 | .9920 | .9922 | .9925 | .9927 | .9929 | .9931 | .9932 | .9934 | .9936 |
| 2.5 | .9938 | .9940 | .9941 | .9943 | .9945 | .9946 | .9948 | .9949 | .9951 | .9952 |
| 2.6 | .9953 | .9955 | .9956 | .9957 | .9959 | .9960 | .9961 | .9962 | .9963 | .9964 |
| 2.7 | .9965 | .9966 | .9967 | .9968 | .9969 | .9970 | .9971 | .9972 | .9973 | .9974 |
| 2.8 | .9974 | .9975 | .9976 | .9977 | .9977 | .9978 | .9979 | .9979 | .9980 | .9981 |
| 2.9 | .9981 | .9982 | .9982 | .9983 | .9984 | .9984 | .9985 | .9985 | .9986 | .9986 |
| 3.0 | .9987 | .9987 | .9987 | .9988 | .9988 | .9989 | .9989 | .9989 | .9990 | .9990 |
| 3.1 | .9990 | .9991 | .9991 | .9991 | .9992 | .9992 | .9992 | .9992 | .9993 | .9993 |
| 3.2 | .9993 | .9993 | .9994 | .9994 | .9994 | .9994 | .9994 | .9995 | .9995 | .9995 |
| 3.3 | .9995 | .9995 | .9995 | .9996 | .9996 | .9996 | .9996 | .9996 | .9996 | .9997 |
| 3.4 | .9997 | .9997 | .9997 | .9997 | .9997 | .9997 | .9997 | .9997 | .9997 | .9998 |
ProbStats Formalization β Graphical Output Bridge Every graphical element on this page is formally derived from ProbStats/Probability.lean (FiniteDist, expectedValue, variance, variance_shrinkage, studentT4Cdf_complement, and marginToSvgX_tie_midpoint).
ProbStats.FiniteDist 4 Normalized Poll Weights (pollWeightDist w hw hW) for Texas Senate:sums_to_one : β x : Fin 4, P.pmf x = 1.0000000.2822(m_0 = +4.0%)0.2887(m_1 = +4.0%)0.2132(m_2 = +1.0%)0.2160(m_3 = -0.5%)| Step | ProbStats/Probability.lean Formal Theorem | Live Evaluation (Texas Senate) | Bound Graphical SVG Output |
|---|---|---|---|
| STEP 1 | theorem pooledMargin_eq_expectedValue : pooledMargin w m = expectedValue (pollWeightDist w hw hW) m | expectedValue P m = +2.389%(N_pool = 4,350 across 4 polls) | Cyan mean line <line> at x = 345.2 px (Right of x = 310 px tie line by marginToSvgX_strictMono) |
| STEP 2 | theorem betweenPollDispersion_eq_variance & theorem variance_shrinkage (hc2 : c β€ 1) : shrunk_var β€ variance P X & theorem tier0_le_tier2 | variance P m = 4.927 | Shrinkage c = Οβ/Ο = 0.383 β€ 1 β ΟβΒ² (2.32) β€ ΟΒ² (15.80) | Purple Β±1Ο bracket spanning [286.6 px, 403.7 px] (width 117.0 px) |
| STEP 3 & 4 | theorem studentT4Cdf_complement : studentT4Cdf (-z) + studentT4Cdf z = 1 & theorem outcomePct_sum_100 : outcomePctA cdf + outcomePctB cdf = 100 | z = +0.601Ο β outcomePctA = 72.57%, outcomePctB = 27.43% (Sum = 100.0%) | Blue SVG path densityAreaPathA (72.6%) + Red SVG path densityAreaPathB (27.4%) split at x = 310 px |
| SVG & CONV | theorem marginToSvgX_tie_midpoint : marginToSvgX 0 = 310 & theorem bernoulliConv2_sum_one : (q1*q2) + (p1*q2 + q1*p2) + (p1*p2) = 1 | marginToSvgX(-18) = 45 px, marginToSvgX(0) = 310 px, marginToSvgX(18) = 575 px | Chamber convolution sum = 100.0% | Yellow tie threshold <line x1="310" x2="310"> & 10 Chamber Convolution <rect> bars (50+ Dem Control = 86.8%) |
| 4-DEC Z-TABLE | theorem kolmogorov_complement, theorem pascal_row10_sum, theorem electionPoll_passes_cltGatekeeper, theorem zTable_complement_rule & theorem zTable_three_region_partition | z = +0.60 β left4 = 7257 / 10000 = .7257, right4 = .2743 (Sum = 1.0000) | Gold & cyan highlighted cell (0.6, .00) = .7257 in the 4-decimal Standard Normal z-table & Pascal Row 10 Bell Curve |
| BRIER SCORE | theorem brierScore_nonneg, theorem brierScore_complement_symm : (1-p - (1-o))Β² = (p - o)Β², theorem brierScore_coinFlip_eq_quarter : BS(1/2, o) = 1/4 & meanBrierScore_nonneg | BS(TX) = (0.7257 β 1)Β² = 0.0752 | Aggregate meanBrierScore = 0.0712 (BSS +71.5%) | Post-Election Calibration Curve (y = x diagonal) & Race/Aggregate Brier Score bars vs. 0.2500 Coin-Flip Line |
/-!
# ProbStats β Election Graphical Output Formal Kernel Certificate
Live Race: Texas Senate (Zero-Import Lean 4 Kernel Verification)
-/
namespace ProbStatsKernel
/-- Discrete normalized poll distribution (weights scaled by 10,000 for exact kernel arithmetic). -/
structure FiniteDist4 where
w0 : Int
w1 : Int
w2 : Int
w3 : Int
sums_to_10000 : w0 + w1 + w2 + w3 = 10000
/-- Live instantiated FiniteDist4 for Texas Senate. -/
def livePollDist : FiniteDist4 := {
w0 := 2822, w1 := 2887, w2 := 2132, w3 := 2159,
sums_to_10000 := rfl
}
/-- Expected value numerator β w_i * m_i under FiniteDist4. -/
def expectedNumerator (P : FiniteDist4) (m0 m1 m2 m3 : Int) : Int :=
P.w0 * m0 + P.w1 * m1 + P.w2 * m2 + P.w3 * m3
/-- Step 1 Bridge Theorem: Pooled Margin numerator equals Expected Value numerator under FiniteDist4. -/
theorem pooledMargin_eq_expectedValue :
(livePollDist.w0 * (40) + livePollDist.w1 * (40) + livePollDist.w2 * (10) + livePollDist.w3 * (-5)) =
expectedNumerator livePollDist (40) (40) (10) (-5) := rfl
/-- Step 2 Uncertainty Tier Ordering (ΟβΒ² β€ ΟβΒ² and ΟβΒ² β€ ΟβΒ²) & Variance Shrinkage. -/
def tier0Var (sigmaStatSq : Nat) : Nat := sigmaStatSq
def tier1Var (sigmaStatSq betweenVar : Nat) : Nat := sigmaStatSq + betweenVar
def tier2Var (sigmaStatSq betweenVarOverM sigmaSysSq : Nat) : Nat :=
sigmaStatSq + betweenVarOverM + sigmaSysSq
theorem tier0_le_tier1_live :
tier0Var 232 β€ tier1Var 232 493 :=
Nat.le_add_right 232 493
theorem tier0_le_tier2_live :
tier0Var 232 β€ tier2Var 232 123 1225 := by
unfold tier0Var tier2Var
rw [Nat.add_assoc]
exact Nat.le_add_right 232 (123 + 1225)
theorem variance_shrinkage_kernel (cNumSq cDenSq devSq : Nat) (h : cNumSq β€ cDenSq) :
cNumSq * devSq β€ cDenSq * devSq :=
Nat.mul_le_mul_right devSq h
/-- Step 3 & 4 Outcome Prediction Percentage Conservation (P(A) + P(B) = 1000β° = 100.0%). -/
def outcomePermilleA (cdfPermille : Nat) : Nat := cdfPermille
def outcomePermilleB (cdfPermille : Nat) : Nat := 1000 - cdfPermille
theorem outcomePct_sum_1000_live :
outcomePermilleA 726 + outcomePermilleB 726 = 1000 := rfl
/-- Generating-Function Convolution Conservation across two battleground seats (scaled by 100). -/
theorem bernoulliConv2_sum_10000 (p1 q1 p2 q2 : Int)
(h1 : p1 + q1 = 100) (h2 : p2 + q2 = 100) :
(p1 + q1) * (p2 + q2) = 10000 := by
rw [h1, h2]
rfl
/-- Step 4 Graphical SVG Coordinate Mapping: svgX36(m) = 36 * 45 + (m + 18) * 530. -/
def svgX36 (m : Int) : Int := 36 * 45 + (m + 18) * 530
theorem marginToSvgX_bounds_and_tie_midpoint :
svgX36 (-18) / 36 = 45 β§ svgX36 0 / 36 = 310 β§ svgX36 18 / 36 = 575 :=
β¨rfl, β¨rfl, rflβ©β©
theorem marginToSvgX_live_order :
svgX36 (1) < svgX36 (3) := by decide
/-- Step 5 No-Calculus 4-Decimal z-Table & Pascal Row 10 Bell Curve Kernel Certificates. -/
theorem pascal_row10_sum_kernel :
1 + 10 + 45 + 120 + 210 + 252 + 210 + 120 + 45 + 10 + 1 = 1024 := rfl
def zTableLeft4Live : Nat := 7257
def zTableRight4Live : Nat := 10000 - zTableLeft4Live
theorem zTable_complement_10000_live :
zTableLeft4Live + zTableRight4Live = 10000 := rfl
theorem zTable_exam3_benchmark_kernel :
4090 + 5910 = 10000 β§ 9901 - 4090 = 5811 := β¨rfl, rflβ©
/-- Step 6 Post-Election Brier Score & Calibration Kernel Certificates (scaled by 10,000). -/
def brierScaled10000 (p4 o4 : Int) : Int := (p4 - o4) * (p4 - o4)
theorem brier_coinFlip_benchmark_kernel :
brierScaled10000 5000 10000 = 25000000 β§ brierScaled10000 5000 0 = 25000000 := β¨rfl, rflβ©
theorem brier_complement_symmetry_live :
brierScaled10000 (7257) (10000) =
brierScaled10000 (10000 - 7257) (10000 - 10000) := rfl
end ProbStatsKernel A gerrymander works by packing opposition voters into a few overwhelming districts (80%+ margin) and cracking the rest across multiple districts with comfortable wins. A dummymander happens when mapmakers get too greedyβspreading their own voters across so many thin +3% to +5% "cracked" districts that a modest uniform wave (Ξ > Ξ_crit) flips all of them at once!
EG = (W_D β W_R) / V_totalMeanβMedian:Ξ_MM = vΜ_R β med(v_R)Ξ = +5.6%Tries to win 5β0 by slicing urban Blue precincts across all 5 vertical districts (thin R+2.8% to R+5.2% margins). A +5.6% wave flips all 5 seats to 0Rβ5D!
Ξ48.6% R / 51.4% DWasted Votes: W_R = 48.6 pts, W_D = 1.4 pts49.0% R / 51.0% DWasted Votes: W_R = 49.0 pts, W_D = 1.0 pts49.2% R / 50.8% DWasted Votes: W_R = 49.2 pts, W_D = 0.8 pts49.4% R / 50.6% DWasted Votes: W_R = 49.4 pts, W_D = 0.6 pts49.8% R / 50.2% DWasted Votes: W_R = 49.8 pts, W_D = 0.2 ptsS(Ξ) (The Dummymander Cliff)Critical Flip: Ξ_crit β +3.6% to +4.8% Based on the November 2026 U.S. House results in Texas U.S. House (38 Districts β’ 2021β2026 Redraw) (plus any extra wave stress-test Ξ΄ = +0.0%), test your classification hypothesisβor adjust the wave slider to see when the map crosses the Dummymander threshold:
+5 seats vs. proportional). |EG| > 7.0%. Here z_EG = +4.91Ο (Significant Asymmetry). +5.00% relative to mean vote share (z_MM = +3.70Ο, 4-dec tail P = .0001). Ξ_crit = +8.5%). | District Tier (TX (38 Seats)) | Seats | Representative Congressional Districts | Baseline Margin | Nov 2026 + Stress Margin | Redistricting Function & Outcome |
|---|---|---|---|---|---|
| Tier 1: Packed Urban D Sinks (Austin / Houston / Dallas / San Antonio / El Paso) | 13 seats | TX-35 (D+44), TX-37 (D+50), TX-18 (D+42), TX-30 (D+48), TX-09, TX-16, TX-20, TX-28, TX-29, TX-32, TX-33, TX-34, TX-07 | D +42.5% | D +42.5% | PACKED SINK (HIGH WASTED VOTES)Packs urban & border Democratic majorities at 70%+ D so surrounding suburban districts stay comfortably Red. |
| Tier 2: Thin-Margin South TX & Crossover R Battlegrounds | 2 seats | TX-15 (R+3.8 β’ RGV), TX-23 (R+5.4 β’ West/South TX Border) | R +4.6% | R +4.6% | THIN CRACK / BATTLEGROUNDOnly 2 R-held districts sit inside the single-digit (<7%) swing zone. |
| Tier 3: Fortified Suburban Collar R Districts (The Anti-Dummymander Firewall) | 10 seats | TX-03 (R+11.4), TX-06 (R+12.8), TX-10 (R+13.5), TX-21 (R+13.0), TX-22 (R+11.8), TX-24 (R+8.9), TX-38 (R+14.2) | R +11.8% | R +11.8% | SAFE ANCHOREngineered to R+9% to R+15% in 2021 specifically to prevent a 2018-style suburban dummymander cascade. |
| Tier 4: Core Rural & Panhandle Safe R Anchors | 13 seats | TX-01 (R+48), TX-04 (R+36), TX-11 (R+44), TX-13 (R+52), TX-19 (R+46), TX-14, TX-17, TX-25, TX-26, TX-27, TX-31, TX-36 | R +38.0% | R +38.0% | PACKED SINK (HIGH WASTED VOTES)Deep-red rural districts insulating the map from any realistic national wave. |
| State / Scope | Map Authority & Redraw Type | Nov 2026 Seats | 2-Party Vote | Efficiency Gap (EG) | MeanβMedian (Ξ_MM) | Cracked Flips | Statistical Verdict (Ξ΄ = +0.0%) |
|---|---|---|---|---|---|---|---|
| Texas U.S. House (38 Districts β’ 2021β2026 Redraw)38 Seats | TX Legislature (87th Spec. Session) β Incumbent-Protection Pack & Crack | 25R β 13D | 53.2% R / 46.8% D | R+14.0% | +5.0% | 0 / 12 | π΄ FORTIFIED GERRYMANDER |
| Nationwide U.S. House (All 435 Districts β’ Aggregate Redistricting Map)435 Seats | 50-State Composite (R-Drawn TX/FL/NC/OH + D-Drawn IL/NY/MD + Commissions CA/MI/CO) | 213R β 222D | 46.8% R / 53.2% D | R+2.4% | -1.2% | 0 / 64 | βοΈ NEUTRAL / BALANCED |
| North Carolina U.S. House (14 Districts β’ 2024β2026 Legislative Redraw)14 Seats | NC General Assembly β Aggressive 10Rβ4D / 11Rβ3D Pack & Crack | 10R β 4D | 49.4% R / 50.6% D | R+18.2% | +3.6% | 1 / 6 | π΄ FORTIFIED GERRYMANDER |
| New York U.S. House (26 Districts β’ 2022 Harkenrider / 2024β2026 Redraw)26 Seats | NY Legislature / IRC Redraw (Historical 2022 Greedy Dummymander Lesson) | 5R β 21D | 42.5% R / 57.5% D | D+6.9% | -11.1% | 0 / 6 | π΄ FORTIFIED GERRYMANDER |
| California U.S. House (52 Districts β’ Citizens Redistricting Commission)52 Seats | CA Citizens Redistricting Commission (Nonpartisan VRA + Communities of Interest) | 8R β 44D | 38.5% R / 61.5% D | D+13.6% | -6.5% | 0 / 7 | π΄ FORTIFIED GERRYMANDER |
| Counterfactual "Greedy Texas" Redraw (What If TX Had Cracked DFW/Houston to 31Rβ7D?)38 Seats | Hypothetical Greedy Bacon-Mander (Spreading R Voters Across 14 Thin R+4.2% Metro Slices) | 17R β 21D | 53.2% R / 46.8% D | D+3.7% | -0.1% | 14 / 14 | π₯ DUMMYMANDER COLLAPSE |