Infinity Engine — Side-by-Side Observation on Five Documented Scientific Forecasting Events

Quiet observation ledger. Infinity's locked harmonic regression is applied to the same data that several published forecasting methods used at their prediction time, on five documented events with now-known truth values. Each method's number is placed next to truth in chronological order by publication year. No rankings, no winners, no scores. The numbers speak for themselves. Per `feedback_no_competition_framing.md`: Infinity is not in competition with NASA, NOAA, ECMWF, JMA, or any published method. Each method has its own assumptions and natural domain. The point is not to declare a winner — it is to honestly report what each method produced when given the same input, alongside what later proved true. The five events: SC23 — Solar Cycle 23 peak SSN. Truth: 180.3 (April 2000, SILSO V2.0 smoothed). Published: Schatten & Pesnell 1993 (170, -5.7%), Hathaway 1994 (160, -11.3%), NOAA SEC Panel 1996 (160 ± 30, -11.3%). Infinity (today): 155.4 [125.3, 185.8] (-13.8%). SC24 — Solar Cycle 24 peak SSN. Truth: 116.4 (April 2014). Published: Dikpati 2006 (155-180, +43.5%), NOAA Panel 2007 HIGH (140 ± 20, +20.3%), NOAA Panel 2007 LOW (75 ± 8, -35.6%), Choudhuri 2007 (80, -31.3%), Hathaway 2010 (70, -39.9%). Infinity (today): 192.9 [185.6, 200.3] (+65.7%). Cycle 24 was anomalously weak; harmonic regression has no information about cycle-to-cycle amplitude variation. SC25 — Solar Cycle 25 peak SSN. Truth: 159.0 (October 2024). Published: Upton & Hathaway 2018 (95-130, -29.6%), Bhowmik & Nandy 2018 (118, -25.8%), NOAA SWPC Panel 2019 (115 ± 10, -27.7%), McIntosh 2020 (210-260, +46.5%), Pesnell metareview 2020 (median 115, -27.7%). Infinity (today): 191.8 [182.9, 200.4] (+20.6%). ENSO-2014 false alarm — ONI peak NDJ 2014. Truth: +0.74. Published: IRI/CPC ENSO Plume 2014 (~+1.7, +129.7%), NOAA CPC official 2014 (~+1.25, +68.9%), JMA ENSO Outlook 2014 (~+1.3, +75.7%). Infinity (today): 0.02 [-0.1, 0.2] (-97.3%). Harmonic regression naturally tends toward zero on irregular signals because no periodic component dominates the fit. ENSO-2015 super event — ONI peak NDJ 2015-16. Truth: +2.64. Published: IRI/CPC ENSO Plume 2015 (~+1.5, -43.2%), ECMWF SEAS5 2015 (~+1.7, -35.6%), NOAA CPC official 2015 (~+1.75, -33.7%). Infinity (today): 0.12 [-0.0, 0.3] (-95.5%). What these observations map. Solar cycles are quasi-periodic but not strictly periodic. Each cycle has its own amplitude. When the next cycle's amplitude happens to be near the average of past cycles, harmonic regression naturally lands close to truth (SC23, SC25). When the next cycle is anomalously weak (SC24), harmonic regression lands far from truth — because the method has no information about cycle-to-cycle amplitude variation. ENSO is irregular and coupled. It does not have a stable fundamental period. Harmonic regression applied to ONI naturally tends toward zero for any horizon longer than a few months. These are not failures of Infinity. They are the natural behavior of a particular method on particular kinds of data. The point of running the observation is to document that natural behavior, not to score it. The honest framing for any reader: 'Infinity applied its locked harmonic-regression configuration to the same data several published methods had used. For each event, here is the number it produced and how far it landed from truth. Other methods used other information.' Method class. Infinity's harmonic regression is in the same broad class as other purely statistical methods. It is NOT in the class of physics-informed methods (Dikpati's flux-transport dynamo, McIntosh's terminator method, ECMWF's coupled climate model). Different classes use different information. Source code: infinity/failed_predictions_evaluation.py (~480 lines). Field names use neutral language throughout: observation_table not ranking_table, distance_from_truth not error, summary not record. No 'win', 'lose', 'beat', 'rank #1', 'leaderboard', 'best', or 'worst' anywhere in code, dashboard, or documentation. Per `feedback_no_competition_framing.md`.