Will the Cleveland Browns beat the Pittsburgh Steelers at home on October 2, 2026?

CLE's season base rate (32.6% win rate) sits well below the market's 42% implied probability for this home game vs PIT. My best-tested signal for a home/recent-form bump (s6) lifts the conditional estimate to only 38.1%, and that lift is not statistically significant (p=0.53, CI spans zero). The Kalshi price itself is stable and slow-moving (autocorr 0.65), suggesting no fresh news the market hasn't absorbed. I therefore have no validated edge that beats a home-field+market-line baseline; my point estimate (0.39) sits modestly below the market (0.42), a mild fade driven mostly by the raw base rate, held with low-moderate confidence given the insignificant supporting test.

Own objection: CLE's raw season win rate (32.6%) is the most direct base rate, and the market's 42% already sits well above it, correctly pricing in home field, opponent (PIT) mediocrity, and any injury edge that d7/d8 capture. My upward adjustment to 0.39 relies on s6, a test that failed significance (p=0.53, wide CI spanning zero) — I have no validated feature that beats the home-field + market-line baseline, which is the explicit bar this role requires before trusting any signal. The intellectually honest read is that I have no edge and should sit at or below the market price, not construct a bespoke number that happens to land near consensus.

The unconditional CLE win-rate base rate (30.4%) sits meaningfully below the market-implied 0.42, but none of the tested home-field or conditional-probability effects clear conventional significance (p=0.108–0.75), so I cannot claim a demonstrated statistical edge against the market. I set my estimate at 0.40 — a small, low-confidence discount to the market price reflecting the gap between the unconditional base rate and market pricing, tempered by recent-form data and a directionally (if not significantly) positive home effect.

Own objection: The sample (n=46) mixes home and away games and possibly different seasons/opponents, so the unconditional base rate of 0.304 is not a clean prior for a single home game against a specific division rival (Steelers). The home-field regression (s3/s7) is suggestive (beta≈6 pts, p≈0.11-0.13) but underpowered (n=23 per group) and not significant — meaning I cannot statistically confirm CLE has a real home edge large enough to justify moving materially off market. If the 'signal' tested in s2 (lift=1.14, p=0.75) is the actual home/division-game conditioning, it is essentially indistinguishable from noise. My discount of the market to 0.40 vs 0.42 is a small, low-confidence nudge, not a demonstrated edge — a critic could argue I should simply defer to the market price of 0.42 since none of my tests clear conventional significance thresholds.

The Kalshi market price for CLE win (0.42) is close to an efficient equilibrium built from 313 stable observations; the only signal suggesting mispricing (a noisy, possibly unrelated series spiking to 0.585) lacks statistical support once corrected for autocorrelation, so there is limited exploitable edge and my estimate sits only marginally above the market price.

Own objection: The divergence in d2 (jumping to 0.585, well above d1's 0.425) could represent genuine new information — e.g., a Steelers injury or lineup change reported close to the as_of time — that the thinner/slower Kalshi market (d1) has not yet fully incorporated. If d2 is a sharper or faster-moving proxy (such as a sharp book's line or a model incorporating late news), then the 0.42 market price understates CLE's true win probability, and the correct trade is to buy CLE ahead of a likely upward repricing. My weighting of d2 as 'noise' rests on an unverified assumption about what d2 actually measures; if d2 is in fact a more informed source, my probability estimate is too low and the market offers a real long-CLE edge rather than being efficient.

The available evidence does not establish a robust, non-circular causal signal that would justify diverging materially from the market's implied 0.42 probability that CLE beats PIT at home on 2026-10-02. The only regression offered is statistically insignificant, the correlation check is weak, requested regime/placebo tests failed at the tool level, and the market-price series itself is being partly used as its own evidence (circular). I therefore anchor close to market, adjusted slightly upward toward the generic NFL home-field base rate.

Own objection: If d1 truly is CLE's win indicator and its last three games are 0,1,1 (a real recent uptick against real opponents), that momentum plus true home-field advantage could push true win probability above 0.42, meaning the market is underpricing CLE and my flat/skeptical stance is too conservative. I cannot rule this in or out because the data package gives no semantic labels or opponent-quality controls, and I have no injury/roster/QB-status data to adjudicate it.

debate skipped: panel unanimous within 0.05

debate skipped: panel unanimous within 0.05

debate skipped: panel unanimous within 0.05

debate skipped: panel unanimous within 0.05

BASE RATE: CLE's unconditional season win rate is 0.30-0.33; this is the anchor before any adjustment. All four analysts converge tightly (0.39-0.44) around the market-implied price (0.42-0.44), and every attempt to find a statistically significant deviation from that market price failed (p-values 0.108-0.75, CIs spanning zero across regression, conditional-probability, and bootstrap tests). Collapsing duplicated evidence (the market-price-stability series cited independently by A, C, and D is one evidence line; the point-differential HAC regression cited by B and D with beta≈6.04 is another single line), there are really only three quasi-independent evidence threads: (1) the low unconditional base rate, (2) the stable/converged market price, and (3) an insignificant positive home-field/recent-form tilt. None of these threads is individually strong enough to move confidently away from market, and they partially offset each other. CONSENSUS: 0.41, weighted most heavily toward Analyst A (best calibration in this narrow sports segment, most sport-specific relevance, and a distinct conditional test) and B (best raw accuracy, genuinely independent regression and bootstrap work), with C and D given less weight due to weaker calibration and heavy evidential overlap with each other and with A. TRADEABILITY: With consensus (0.41) essentially indistinguishable from the market price (~0.42) and every underlying test statistically insignificant, there is no demonstrated, non-circular edge here. Expected return from a directional position is ~-0.01, well within model uncertainty and almost certainly inside any realistic bid-ask/fee spread on a sports-contract market. RECOMMENDATION: No trade. This is a case where the correct action is to defer to the market price rather than express a view — the disagreement among analysts is a symptom of small-sample noise (n=2 to n=9 calibration histories, wide bootstrap CIs) rather than a genuine informational edge. If forced to express a view, a small, low-conviction fade of the market (short CLE-win / long PIT-win) at a max size consistent with a ~1-2 point edge would be the only defensible expression, and only if transaction costs are near zero.

Consensus 41% · interval 33% to 48%