Tailfifty desk
424,069 observationssnapshot 2026-10-15Registry

Sports · 5d · asked 2026-10-02 · scored 2026-10-07

Will the New Orleans Saints beat the Atlanta Falcons at home on October 6, 2026?

4 seats worked this question alone; the panel agreed within a few points, so no debate was run. Code ran every test and computed the number. How the desk works

Committee53%PM-stated number (before 2026-10-02) · stated range 45% to 60%
Market at the time56%committee minus market −3%
ResolvedNoBrierBrier score How wrong a probability was, squared. Say 70% and it happens: (0.7 − 1)² = 0.09. Say 70% and it doesn't: 0.49. Lower is better; always saying 50% scores 0.25, so anything above 0.25 is worse than a coin flip. glossary 0.276 vs marketMarket Brier The same squared error, but for the Kalshi price at the moment the question was asked. It is the bar: beating it means knowing something the price did not. glossary 0.308 · lower is better

Judged by Kalshi KXNFLGAME-26OCT05ATLNO-NO settlement: If New Orleans wins the ATL Falcons vs NO Saints Pro Football game originally scheduled for Oct 5, 2026, then the market resolves to Yes.

0%50%100%
Sports Analyst53%
Quantitative Researcher52%
Market Positioning Analyst53%
Causal Skeptic52%
Committee53%

Hollow: before the debate. Gold: after. Blue rule: the market. Red rule: what happened. Debate skipped: the seats agreed within a few points. Sat out: Historical Analog Analyst.

The dispute that matteredAll four analysts land within 0.52-0.53 on the final probability, so there is no material numerical disagreement. The substantive disagreement is about WHY the market's 0.56 is off: A/B/D attribute the gap to a fundamentals-based mispricing (Falcons show a higher long-run win rate and better point margin than the Saints, and no significant home-field coefficient exists in this sample to justify a home favorite), while C attributes most of the gap to the market having already partially repriced toward Atlanta via an implied-probability drift (0.408 -> 0.445) and spread narrowing (-3.5 -> -2.5), leaving only a small residual edge. I resolved this by favoring the fundamentals-based view: D's own evidence (near-unit autocorrelation and bimodal prints in the probability series) suggests the 'drift' C cites could be a stale-quote artifact rather than fresh informed repricing, which weakens C's efficiency argument. The team win-rate gap (~30% Saints vs ~41% Falcons) and the non-significant, wrong-signed home coefficient are observed directly from outcome data and are harder to explain away, so they anchor the consensus toward 0.52-0.53 rather than the market's 0.56.

Synthesis

BASE RATE: Generic NFL home-favorite base rate (used implicitly by the market's 0.56 price) assumes a standard home-field edge of roughly 2-3 win-probability points league-wide. This dataset's own test of that assumption (OLS win~is_home, p=0.592, wrong-signed point estimate) fails to confirm a home-field effect for these teams, which is the single most important update away from the market price. UPDATE: Starting from the market's 0.56, the committee updates down to a consensus of ~0.525 based on: (1) team-level win-rate and margin data favoring Atlanta by a modest but consistent amount, and (2) the absence of a statistically significant home-field coefficient in this sample. Analyst C's evidence that the market has already partially repriced toward Atlanta is real but is discounted because the same series shows near-unit autocorrelation and bimodal prints -- consistent with illiquid/stale quoting rather than fresh information -- so it does not fully explain away the residual gap between 0.56 and the fundamentals-implied ~0.52-0.53. TRADEABILITY: This is weakly tradeable. The edge (market 0.56 vs fair ~0.525, roughly 3.5 probability points, ~6% edge on a 0.44-priced "Falcons win" contract) is real but thin, resting on non-significant statistical tests (p=0.59, p=0.75) and a small, possibly stale market-drift signal. I would NOT size this as a high-conviction directional bet. Proposed expression: a small position buying "Falcons to win" (i.e., fading Saints) at the current ~0.44 price, sized to reflect low confidence (conf~0.35 across the panel) -- e.g., 1-2% of a sports-event risk budget, not a core position. Exit/invalidation: if the market price for Saints drops below 0.53 before kickoff on fresh volume (confirming the repricing thesis), take profit; if it holds at or above 0.56 through Friday 10/2 close with volume, reduce size further since the thesis depends on illiquidity that may not resolve in time. SYNTHESIS NOT AVERAGE: The consensus of 0.525 is not a naive mean of {0.53, 0.52, 0.53, 0.52} (which would also be ~0.525) -- it is explicitly weighted toward A and D (0.30 each) for calibration/relevance/rigor, with B down-weighted (0.25) for evidence duplication with D, and C down-weighted most (0.15) for the worst calibration record in this segment, with C's independent market-microstructure evidence retained but discounted for the stale-quote concern D raised.
Observed · in the data
  • Kalshi/market mid implied probability for a Saints home win is ~0.56 as of the as_of timestamp.
  • In the underlying dataset (46 observations), Saints (d1) empirical win rate is ~30.4% and Falcons (d4) win rate is ~41.3%, an ~11pt gap favoring Atlanta.
  • Point-margin means: Saints -4.44 (95% CI [-8.63, -0.07]), Falcons -3.33 -- Atlanta's average margin is ~1.1pts better.
  • The spread series (d7) narrowed from -3.5 to -2.5 and the ATL-implied-probability series drifted from a historical mean of 0.408 to a recent plateau of ~0.445 (bootstrap CI [0.4038, 0.412] excludes recent prints).
Modeled · a tool computed it
  • OLS HAC regression of win outcome on is_home: coefficient = -0.0606, p=0.592, 95% CI [-0.28, 0.16] -- statistically indistinguishable from zero and wrong-signed relative to a home-favorite premise.
  • Conditional-probability test of outcome given an unspecified signal: P(win|signal)=0.273 vs base rate 0.304, lift=0.896, p=0.754 -- no significant edge.
  • Committee-level weighted synthesis of four analyst posteriors (0.52-0.53) into a consensus point estimate of 0.525.
Inferred · the committee's reasoning
  • The ~11pt win-rate gap and the slightly better Falcons point margin imply Atlanta is the stronger team on raw historical form, independent of venue.
  • The observed spread/probability drift toward Atlanta implies the market has already partially adjusted away from a pure home-field prior, but the degree to which this reflects genuine new information versus stale/duplicated quoting (flagged via near-unit autocorrelation) is disputed.
Predicted · what it expects
  • Fair probability of a Saints home win on 2026-10-06 is approximately 0.525 (90% plausible range 0.45-0.60), below the market's current 0.56.
  • If the mispricing is real, the market price should drift down toward ~0.52-0.53 as kickoff approaches and any remaining informational lag resolves.
Uncertain · what it does not know
  • Whether the observed implied-probability drift (C's evidence) reflects real new information or stale/duplicated quotes (D's autocorrelation/bimodality concern) -- this is the key unresolved question.
  • Sample size is modest (46 game observations, n=2-15 analyst calibration samples), so both the team-strength gap and the home-field null result carry wide confidence intervals.
  • No injury report, weather, or line-movement data beyond the five series analyzed were available; omitted variables (QB status, rest, weather on 10/6) could shift the true probability materially.
  • Analyst A's perfect calibration score (accuracy 1.0, brier 0.112) is based on only n=2 prior calls and should not be fully trusted as a predictor of future accuracy.
How the seats were weighed · not an average
Analyst A (Sports Analyst)0.30Highest domain relevance to sports (1.0) and best reported calibration (brier 0.112, accuracy 1.0), but n=2 is too small to trust that calibration score at face value; weight capped despite nominal top score. Evidence (win-rate gap, OLS home coefficient) partially overlaps with D.
Analyst D (Causal Skeptic)0.30Solid calibration on a reasonable sample (n=15, brier 0.245, accuracy 0.53), high general relevance, and the most rigorous treatment of statistical significance (flags non-significant home effect, flags stale/autocorrelated market series). Treated as the anchor for the 'no detectable edge' view, with A's overlapping OLS evidence folded in rather than double-counted.
Analyst B (Quantitative Researcher)0.25Decent sample (n=13) but weaker calibration (brier 0.259) and lower sports-specific relevance (0.7). Evidence (conditional win-probability test) duplicates D's signal test, so down-weighted for non-independence even though thesis is directionally consistent.
Analyst C (Market Positioning Analyst)0.15Worst calibration in segment (brier 0.300, accuracy 0.47, below coin-flip historically) despite high market relevance. Provides the only genuinely independent evidence line (implied-probability drift, spread narrowing), which earns it a non-zero weight, but the autocorrelation/bimodality flagged by D as possibly stale-quote artifact undercuts confidence in reading that drift as informative.
Wrong if
  • If the Saints win the game outright on 2026-10-06, the 'fade the home favorite' thesis is falsified for this instance (though a single game cannot fully falsify the broader probability claim).
  • If the Kalshi market price for Saints moves to or through 0.60+ with volume (not just a stale repeated quote) before kickoff on 2026-10-06, that is evidence the market disagrees with the committee's downward adjustment and the consensus should be revisited.
  • If a re-run of the OLS home-field regression on an expanded sample (post 2026-10-06, more games) yields a significant positive is_home coefficient (p<0.05), the 'no home-field advantage in this dataset' evidence is undermined and the consensus probability should be revised upward.
Model uncertainty

All four posteriors cluster tightly (0.52-0.53), but this reflects shared reliance on two underlying evidence pools (team win-rate/margin data, and a single OLS/conditional-probability test) that are duplicated across A, B, and D rather than four truly independent confirmations. Calibration scores for the panel are mediocre-to-poor in this segment (brier 0.11-0.30, with only Analyst A near the floor on a tiny n=2 sample), so the tight clustering should not be read as high precision. The CI is set wider than the raw point-estimate spread (0.52-0.53) to reflect this true epistemic uncertainty.

The seats

Each analyst's own memo, evidence and tests. Open a seat to read it.

Sports Analyst53%confidence 35%

Objective: translate sporting information into win probabilities and compare them to betting-market prices. Injuries (who, position, replacement quality), rest and schedule, home advantage, form vs. noise, and how fast lines move after news. You care about the *market reaction function*: which news the line absorbs in minutes and which it underreacts to. A 60% team is only interesting if the market says 52%. Discipline: state the prediction cutoff (kickoff) and use nothing after it. Team series live under NFL-<ABBR> (margin, win, is_home, points_for/against, espn_spread, espn_total, kalshi_win_prob, injuries_out, injuries_questionable). Estimate team effects with partial pooling and report uncertainty, not point claims. Model dispersion as well as the mean. Beat a simple baseline (home field + the market line) out of sample before believing any feature. Injury counts are a proxy for who is missing; say so.

Thesis. The market's 56% for New Orleans at home is slightly too high relative to the team-strength signals in the available data (Atlanta shows a better average win rate and point margin), while the home-field effect itself is statistically undetectable in this dataset. I adjust down modestly to ~53%, treating the market as close to fair but with a small overpricing of New Orleans' home edge.
Base rate

55% Generic NFL home team win rate (roughly 53-57% across recent seasons, before accounting for team quality)

The best case against

The d1/d2/d3/d4 series include August (preseason) dates, meaning a meaningful fraction of the 46 observations reflect backup-heavy, low-stakes games that are poor proxies for Week-5/6 regular-season team strength; the market, informed by real-time depth-chart and injury-return information (d7 shows injuries_out dropping from 5 to 2 right before the game), may be correctly pricing a healthier NO roster that this noisy historical sample cannot capture. If the 56% price is driven mainly by that injury recovery rather than a mispriced home-field assumption, my downward adjustment is wrong.

Strongest evidence
  • Win-rate base rates in the underlying series: ~30.4% (d3) vs ~41.3% (d4) imply Atlanta has the better recent/longer-run record, a ~11pt gap not explained by home field.
  • Point-margin means: NO -4.44 (CI95 [-8.63,-0.07], barely excludes 0) vs ATL -3.33 — ATL about 1.1 pts better on average margin, opposite direction of a 56% home favorite.
  • OLS win~is_home coefficients are small, negative, and non-significant (p=0.59, p=0.45) — no detectable home-field advantage in this sample, undermining the standard home+market baseline the market price likely assumes.
Wrong if

If New Orleans wins the game on 2026-10-06 (margin >= 1), this specific forecast of 'market slightly overprices NO' is falsified in direction (though a single game is weak evidence for a probabilistic claim); more directly, if post-game review shows the ATL-favoring signals (win rate, margin) were artifacts of preseason-contaminated samples rather than true current team-strength differences, the thesis is undermined.

Assumes
  • d1 and d2 correspond to NO and ATL point-margin series respectively (unconfirmed by explicit labels)
  • d3/d4 correspond to win-rate indicators for NO and ATL respectively (unconfirmed)
  • d7 (injuries_out) reflects a recent reduction in missing players for one of the two teams, consistent with the market's upward move from a ~0.41 historical mean (d9/s7) to the current 0.56 quote
  • Preseason (August) games are included in the 46-game samples and add noise rather than signal for current team strength
Memo to the committee
Market prices NO at 56% to beat ATL at home on 10/6/2026. Available series show Atlanta with a better long-run win rate (41% vs 30%) and point margin (-3.33 vs -4.44), both unfavorable to NO, while home-field itself shows no significant effect in regression (p=0.59, p=0.45). However, these series include preseason games (noisy, backup-heavy) and the recent injuries_out data (dropping 5->2 near as_of) likely explains the market's move up from a ~41% historical base rate, which I can't fully verify given data lags behind the quoted 56%. Net: I see a small, low-confidence case that the market is slightly too bullish on NO, landing at 53% vs the market's 56%. Confidence is limited by small samples and ambiguous series labeling.
Rebuttal
debate skipped: panel unanimous within 0.01

Would change my mind: If New Orleans wins the game on 2026-10-06 (margin >= 1), this specific forecast of 'market slightly overprices NO' is falsified in direction (though a single game is weak evidence for a probabilistic claim); more directly, if post-game review shows the ATL-favoring signals (win rate, margin) were artifacts of preseason-contaminated samples rather than true current team-strength differences, the thesis is undermined.

After the debate · 53% → 53%

debate skipped: panel unanimous within 0.01

Data and tests · 16, as of
Quantitative Researcher52%confidence 35%

Loading…

Market Positioning Analyst53%confidence 35%

Loading…

Causal Skeptic52%confidence 35%

Loading…

Claims and evidence

Observation · seen in the data · 9
Model · a tool computed it · 3
Inference · reasoned from the above · 5
Forecast · what a seat expects · 2
Uncertainty · what a seat does not know · 3

The debate

Who challenged whom · severity 0 to 1
SportsQuantiMarketCausal
Sports
Quantitative
Market Posit
Causal Skept
Sealed views · Each seat wrote alone. Nobody had seen another view.
Sports Analyst
53% · confidence 35%

The market's 56% for New Orleans at home is slightly too high relative to the team-strength signals in the available data (Atlanta shows a better average win rate and point margin), while the home-field effect itself is statistically undetectable in this dataset. I adjust down modestly to ~53%, treating the market as close to fair but with a small overpricing of New Orleans' home edge.

Own objection: The d1/d2/d3/d4 series include August (preseason) dates, meaning a meaningful fraction of the 46 observations reflect backup-heavy, low-stakes games that are poor proxies for Week-5/6 regular-season team strength; the market, informed by real-time depth-chart and injury-return information (d7 shows injuries_out dropping from 5 to 2 right before the game), may be correctly pricing a healthier NO roster that this noisy historical sample cannot capture. If the 56% price is driven mainly by that injury recovery rather than a mispriced home-field assumption, my downward adjustment is wrong.

Quantitative Researcher
52% · confidence 35%

Market price of 0.56 for a Saints home win roughly matches a generic home-field-advantage prior, but the team-level base rates in the data (Saints ~30% win rate, Falcons ~41-48% win rate over the last 46 observations) suggest the Saints are a weaker team than a typical home favorite, and none of the tested signals (conditional win probability given various 'on' conditions, scoring-margin differences) show a statistically significant edge in either direction. I therefore see the market as only mildly overpriced, not clearly wrong.

Own objection: NFL home-field advantage is a well-documented, large-sample effect (historically team-agnostic home win rate ~55-58%), and it is not represented in d1-d4 since those appear to measure team form/scoring margin, not home/away splits specific to this matchup. If d1's 0.304 reflects road-heavy or mixed-opponent games rather than Saints-at-home specifically, the true conditional home-win probability could be meaningfully higher than the raw base rate, which would argue the market's 0.56 is actually fair or even conservative, not overpriced.

Market Positioning Analyst
53% · confidence 35%

The Kalshi mid of 0.56 for Saints is close to fair value, not a clean mispricing. The implied-probability series (d1/d8, read as ATL win probability) shows a persistent upward drift from a historical mean of 0.408 to a current plateau of 0.445, implying the market has already repriced Saints down from roughly 0.59 to ~0.555 over recent sessions, absorbing poor recent form (d4: 2 of last 6 won) and volatile score margins (d2/d3). The spread (d7) also narrowed from -3.5 to -2.5, consistent with information flowing toward Atlanta, not a stale line. Partially offsetting, d5/d6 (likely injury counts) show an improving trend into game week that a sticky market (autocorr 0.83, repeated 0.445 prints) may underweight. Net: a slight, low-confidence tilt toward fading the market, not a strong disagreement.

Own objection: The series semantics are inferred, not confirmed. If d1/d8 and d7 do not map to this exact ATL-NO matchup, the entire 'market already moved' narrative collapses and there is no basis to deviate from the stated 0.56.

Causal Skeptic
52% · confidence 35%

The market's 0.56 implied probability for a Saints home win is not well-supported by the statistical tests provided. Every causal test of a 'home effect' or 'signal effect' on outcome/margin in this dataset is statistically indistinguishable from zero, several point in the wrong direction, and the underlying market-probability series show suspicious near-unit autocorrelation and bimodality consistent with stale/duplicated quotes rather than fresh information. I default close to the generic NFL home-win base rate rather than the market price, with wide uncertainty.

Own objection: The regressions and conditional-probability tests use a small, likely generic/pooled sample (n=46, n=22 subsets) that may not isolate Saints-Falcons-specific dynamics (injuries, recent coaching changes, divisional familiarity) — a null result on a noisy pooled home-effect regression does not prove the market is wrong about THIS specific matchup; the market may be pricing idiosyncratic information (QB status, recent form) that these generic statistical tests cannot see or were never designed to capture. In that case 0.56 could be a well-calibrated specific-game price even though the generic home-field test fails.

Revisions · Every change of mind pinned to one piece of evidence.
Sports Analyst
53% → 53%

debate skipped: panel unanimous within 0.01

Quantitative Researcher
52% → 52%

debate skipped: panel unanimous within 0.01

Market Positioning Analyst
53% → 53%

debate skipped: panel unanimous within 0.01

Causal Skeptic
52% → 52%

debate skipped: panel unanimous within 0.01

Synthesis · The portfolio manager weighs the seats. Not an average.
Portfolio Manager

BASE RATE: Generic NFL home-favorite base rate (used implicitly by the market's 0.56 price) assumes a standard home-field edge of roughly 2-3 win-probability points league-wide. This dataset's own test of that assumption (OLS win~is_home, p=0.592, wrong-signed point estimate) fails to confirm a home-field effect for these teams, which is the single most important update away from the market price. UPDATE: Starting from the market's 0.56, the committee updates down to a consensus of ~0.525 based on: (1) team-level win-rate and margin data favoring Atlanta by a modest but consistent amount, and (2) the absence of a statistically significant home-field coefficient in this sample. Analyst C's evidence that the market has already partially repriced toward Atlanta is real but is discounted because the same series shows near-unit autocorrelation and bimodal prints -- consistent with illiquid/stale quoting rather than fresh information -- so it does not fully explain away the residual gap between 0.56 and the fundamentals-implied ~0.52-0.53. TRADEABILITY: This is weakly tradeable. The edge (market 0.56 vs fair ~0.525, roughly 3.5 probability points, ~6% edge on a 0.44-priced "Falcons win" contract) is real but thin, resting on non-significant statistical tests (p=0.59, p=0.75) and a small, possibly stale market-drift signal. I would NOT size this as a high-conviction directional bet. Proposed expression: a small position buying "Falcons to win" (i.e., fading Saints) at the current ~0.44 price, sized to reflect low confidence (conf~0.35 across the panel) -- e.g., 1-2% of a sports-event risk budget, not a core position. Exit/invalidation: if the market price for Saints drops below 0.53 before kickoff on fresh volume (confirming the repricing thesis), take profit; if it holds at or above 0.56 through Friday 10/2 close with volume, reduce size further since the thesis depends on illiquidity that may not resolve in time. SYNTHESIS NOT AVERAGE: The consensus of 0.525 is not a naive mean of {0.53, 0.52, 0.53, 0.52} (which would also be ~0.525) -- it is explicitly weighted toward A and D (0.30 each) for calibration/relevance/rigor, with B down-weighted (0.25) for evidence duplication with D, and C down-weighted most (0.15) for the worst calibration record in this segment, with C's independent market-microstructure evidence retained but discounted for the stale-quote concern D raised.

Consensus 53% · interval 45% to 60%

Assumptions

Review · 16 · a forecast resting on these resolved no
  • review
    d1 and d2 correspond to NO and ATL point-margin series respectively (unconfirmed by explicit labels)
    Sports Analyst
  • review
    d3/d4 correspond to win-rate indicators for NO and ATL respectively (unconfirmed)
    Sports Analyst
  • review
    d7 (injuries_out) reflects a recent reduction in missing players for one of the two teams, consistent with the market's upward move from a ~0.41 historical mean (d9/s7) to the current 0.56 quote
    Sports Analyst
  • review
    Preseason (August) games are included in the 46-game samples and add noise rather than signal for current team strength
    Sports Analyst
  • review
    d1 and d3 represent win/cover indicators for Saints and Falcons respectively (or vice versa) over the last ~46 games/weeks, though exact team attribution to each handle is not explicitly labeled in the provided data.
    Quantitative Researcher
  • review
    d5/d6 represent point-differential or margin series for the two teams; neither shows a significant difference (s3, s4 both p>0.25, CIs spanning zero).
    Quantitative Researcher
  • review
    d7/d8 (n=8, very recent window) are short-horizon signals (e.g., weekly injury counts or similar) with wide bootstrap CIs (s7: [2.25,3.875], s8: [0.25,3.75]) that do not resolve directly to a win/loss signal and are too noisy to act on.
    Quantitative Researcher
  • review
    No single data handle directly encodes 'Saints win rate specifically when playing at home vs. Falcons,' so all inference here is an indirect composite, not a direct test of the resolution variable.
    Quantitative Researcher
  • review
    d1/d8 represent the Atlanta win-probability series for this matchup (Saints implied = 1 minus these values).
    Market Positioning Analyst
  • review
    d7 represents the point spread for this specific Oct 6 2026 game.
    Market Positioning Analyst
  • review
    d2/d3 are recent score-margin series for the two teams over their last 6 games.
    Market Positioning Analyst
  • review
    d5/d6 represent injury or roster-absence counts, lower being better for the Saints.
    Market Positioning Analyst
  • review
    No line-moving news occurred between the last data timestamp and kickoff that is unobserved here.
    Market Positioning Analyst
  • review
    d1-d6 represent some generic sample of team-games (possibly not limited to Saints/Falcons specifically), used as a proxy test of home-field and signal effects rather than a Saints-specific model
    Causal Skeptic
  • review
    d7 and d8 are market-implied win-probability time series related to this contract or a reference set of similar contracts
    Causal Skeptic
  • review
    The 'signal' tested in s3 and s5 is some fundamental or model-based predictor relevant to the home win hypothesis, though its exact definition is not stated
    Causal Skeptic

Simulated trade

The portfolio manager proposed no paper trade. Instrument-catalog mismatch: the research question concerns a discrete sports-event outcome (Saints vs Falcons win probability), but the simulated book only contains commodity/equity/crypto tickers (HENRY_HUB, UNG, XLE, VLO, MPC, PSX, CORN, BTC, ETH, WTI). None of these have any theoretical or empirical factor loading on an NFL game result -- using any of them as a 'proxy' would be pure noise injection, not a risk-managed expression of the thesis. Declining to propose a trade is the correct action rather than forcing a spurious cross-asset hedge. If a sports-specific prediction-market contract (e.g., a Saints-moneyline or Falcons-moneyline instrument) becomes available in the catalog, the thesis above (fade Saints at ~0.56, fair value ~0.525, 1-2% of a sports-risk budget) would become tradeable at the stated thin edge, net of estimated 150-300bps transaction cost on an illiquid two-sided prediction market and crowding ~0.65. Note also that even the underlying sports thesis itself rests on non-significant statistical tests (home-field OLS p=0.59) and a possibly-stale market-drift signal, so confidence is low (~0.35) independent of the instrument problem.

Forecasts

SeatStagePBaseResolvesOutcomeBrier
Committeefinal53%—2026-10-07No0.276
Sports Analystprior55%55%2026-10-07No0.303
Sports Analystindependent53%55%2026-10-07No0.281
Sports Analystrevised53%55%2026-10-07No0.281
Quantitative Researcherprior55%57%2026-10-07No0.303
Quantitative Researcherindependent52%57%2026-10-07No0.270
Quantitative Researcherrevised52%57%2026-10-07No0.270
Market Positioning Analystprior55%56%2026-10-07No0.303
Market Positioning Analystindependent53%56%2026-10-07No0.281
Market Positioning Analystrevised53%56%2026-10-07No0.281
Causal Skepticprior53%57%2026-10-07No0.281
Causal Skepticindependent52%57%2026-10-07No0.270
Causal Skepticrevised52%57%2026-10-07No0.270

Reasoning graph

Every observation, test, claim, objection and forecast, as a graph
observationmemorytestassumptionclaimcounterargumentopinionforecasttrade
testconditional_probability: P(outcome|…claim · observation · 50%Atlanta has a higher average win ra…testinfluence: corr=-0.06579516949597689counterargumentSeries include preseason games with…counterargumentSmall n=46 with likely overlapping/…opinionSports Analysttestconditional_probability: P(outcome|…claim · model · 55%No statistically significant home-f…testdescribe: mean=-3.3260869565217392counterargumentSmall sample size (n=46) gives low …observationKXNFLGAME-26OCT05ATLNO-ATL.implied_…claim · inference · 40%The market's implied win probabilit…observationKXNFLGAME-26OCT05ATLNO-ATL.implied_…counterargumentd9's last values (0.445) are still …testregression: beta[is_home]=-0.060606…claim · forecast · 30%New Orleans' win probability at hom…testregime_split: full-sample corr=-0.0…testdescribe: mean=0.40642857142857153claim · uncertainty · 60%The evidence base is weak (small n,…observationNFL-NO.win [level]claim · observation · 80%Saints' underlying win rate (d1, 0.…opinionQuantitative ResearcherobservationNFL-ATL.win [level]claim · observation · 65%Falcons' underlying win/cover rate …claim · uncertainty · 85%No tested conditional signal (s1, s…claim · uncertainty · 80%Scoring-margin difference tests (s3…claim · inference · 40%Given no significant edge from avai…claim · model · 50%The true home-field-advantage base …claim · observation · 55%The ATL-implied probability series …opinionMarket Positioning Analystclaim · observation · 40%The point spread narrowed from -3.5…counterargumentOnly 2 data points, could be noise …observationNFL-ATL.is_home [level]claim · inference · 50%Saints recent form is poor (2 wins …observationNFL-NO.is_home [level]observationNFL-NO.margin [level]claim · observation · 40%Injury or absence counts for the Sa…observationNFL-ATL.margin [level]counterargumentSeries label and directionality are…claim · model · 35%The Kalshi market shows signs of th…claim · forecast · 35%Net fair probability for Saints to …testregression: beta[is_home]=-0.060606…claim · observation · 80%The home-field effect on margin/out…counterargumentn=46 may be underpowered to detect …opinionCausal Skeptictestregime_split: full-sample corr=-0.0…claim · observation · 70%The full-sample correlation between…testinfluence: corr=-0.06579516949597689counterargumentNo sign flips in LOO suggests the (…testconditional_probability: P(outcome|…claim · observation · 75%Conditional probability tests of 's…testconditional_probability: P(outcome|…counterargumentSmall n=22 signal subset limits pow…testdescribe: mean=0.4079399141630901claim · inference · 55%The market-implied probability seri…testdescribe: mean=0.40642857142857153counterargumentHigh autocorrelation is also a comp…claim · inference · 50%Given the absence of any statistica…counterargumentNone of these point estimates are i…forecastP=0.53 Will the New Orleans Saints …
Click a node. Every conclusion traces back to observations and tests.