Jacob E. ChandranRésumé PDF ↗︎
← All insights

Research distributions

The result is a distribution

Compare the middle, spread, benchmark and drawdown across technology, energy and healthcare research cohorts.

Snapshot Oct 16, 2026

Performance has a spread

All valid full-window tests in each selected cohort. Boxes show quartiles, whiskers the full range, and gold markers the recorded same-period SPY return.

Return distribution across retrospective research tests; boxes are quartiles, whiskers full ranges; gold marks matched SPYTechnologytechnology: 78 retrospective full-window tests; return min, Q1, median, Q3, max: 29.28%, 67.05%, 90.92%, 118.71%, 158.95%Matched SPY +12.0%Energyenergy: 12 retrospective full-window tests; return min, Q1, median, Q3, max: 1.77%, 25.16%, 29.30%, 43.78%, 49.51%Matched SPY +12.0%Healthcarehealthcare: 10 retrospective full-window tests; return min, Q1, median, Q3, max: -19.18%, -14.34%, -12.62%, -9.53%, -6.68%Matched SPY +12.0%Utilitiesutilities: 12 retrospective full-window tests; return min, Q1, median, Q3, max: -12.42%, -8.09%, -3.21%, 0.67%, 10.45%Matched SPY +12.0%−19%0%+159%
SectorTestsMin25thMedian75thMaxSPYAbove SPY
Technology78+29.3%+67.1%+90.9%+118.7%+158.9%+12.0%78/78
Energy12+1.8%+25.2%+29.3%+43.8%+49.5%+12.0%10/12
Healthcare10−19.2%−14.3%−12.6%−9.5%−6.7%+12.0%0/10
Utilities12−12.4%−8.1%−3.2%+0.7%+10.4%+12.0%0/12

Stock selection is retrospective. These model configurations reuse a holdout; they are not independent forward replications. The median is not the return of a combined portfolio.

The cost of getting there

Maximum drawdown is the worst peak-to-trough decline in each simulation. The boxes show its distribution across tests.

Drawdown distribution across retrospective research tests; boxes are quartiles, whiskers full rangesTechnologytechnology: 78 retrospective full-window tests; drawdown min, Q1, median, Q3, max: -24.46%, -20.64%, -18.12%, -15.72%, -9.91%Energyenergy: 12 retrospective full-window tests; drawdown min, Q1, median, Q3, max: -12.40%, -8.82%, -7.24%, -6.54%, -0.23%Healthcarehealthcare: 10 retrospective full-window tests; drawdown min, Q1, median, Q3, max: -24.78%, -23.25%, -21.84%, -18.95%, -14.09%Utilitiesutilities: 12 retrospective full-window tests; drawdown min, Q1, median, Q3, max: -16.48%, -15.36%, -13.26%, -10.91%, -7.68%−25%0%

Technology’s returns and drawdowns tell different parts of the story. Healthcare’s failed cohort remains visible. The simulations’ fees and execution assumptions are included in their saved metrics; these are not audited live-account outcomes.

The work behind the snapshot

491completed experiments with saved results
196completed across the seven dated days below
DayCompleted
2026-10-090
2026-10-100
2026-10-110
2026-10-120
2026-10-1336
2026-10-1472
2026-10-1588

Counts include saved completed research beyond the three cohorts shown above. A completed experiment is evidence of work, not necessarily a favorable investment result.

Open portfolio figures, sectors and the full selection method →

What this view establishes

  • Boxes are the 25th and 75th percentiles; the line is the median. Whiskers include the full range. Gold markers show matched-period SPY.
  • These tests share evaluation dates and benchmark within each cohort. Different model configurations on a repeated holdout are not independent replications.
  • Stock groups were selected retrospectively, after the evaluation period. Selection and repeated testing can inflate apparent performance.
  • Results include losing tests. They are offline simulations, separate from the owner-reported live portfolio estimate.
Public research aggregates & timestamps ↗︎Research selection and comparison method ↗︎
Read the original study or full series entry →