Sources

Every signal comes from a source, and each source is measured on its own. A change to how a source chooses trades starts a new version with a fresh record.

Validating — published and measured; it has not met the pay criteria yet.
Active — has met them.
In testing — never published. It runs live with frozen rules to check, out of sample, what its historical test showed; its result is published, not its holdings.
Paused — no longer published. It keeps running in shadow, so its record is still measured.
Retired — replaced by a newer version; it produces nothing, and its record stays.

Each signal source is evaluated at 50, 100 and 200 resolved signals. It is paused when the upper end of the 95% interval of its average result is below zero — evidence that following it loses. Not beating chance keeps it validating, never active. For now, pauses are applied by hand, with the reason published here and in the changelog. Numbers as of Nov 7, 2026.

Model portfolios hold a basket instead of publishing signals, so these milestones don't apply to them: each one is measured by its own live record, as its card says.

A source's picks are only presented as signals once its evidence passes explicit gates. Certification is not enforced yet, so every pick is still published; sources that do not pass will run in shadow, measured and never presented as actionable.

How sources are certified

Out of sample: a walk-forward span of at least 5 years with at least 200 trades (picks held days to weeks) or 36 rebalances (portfolio picks), from the parity backtest. The average after costs has a 95% bootstrap lower bound above zero and beats its baseline (random timing, or the benchmark) at a Holm-adjusted p below 0.05 across the variants tried. Worst drawdown at most 25R, or at most the benchmark's. Then, live on the frozen version: provisional, and certified at 30 live outcomes whose average is not below the backtest's 95% range. A live average below that range, or a live drawdown beyond the backtest's 95th percentile, suspends the source.

AI watchlist v1

Validating

Validating since Sep 23, 2026.

Not yet certified Not certified yet. Its picks are still published; they will run in shadow once certification is enforced.

No backtest evidence has been recorded for this version.

Trading: no resolved signals yet · next evaluation at 50

Portfolio: 2 priced picks · in profit now 50% (95% CI 9–91%) · avg −0.3% (95% CI −5.8% to +5.1%) · next evaluation at 50

vs SPY over the same periods: +5.1% vs −1.4% · excess +6.5% · p 0.50 · 1 pick · provisional

History (1)

Sep 23, 2026 Validating — Published and measured; not yet meeting the pay criteria.

Earnings window v1

Validating

Validating since Sep 23, 2026.

Not yet certified Not certified yet. Its picks are still published; they will run in shadow once certification is enforced.

No backtest evidence has been recorded for this version.

Trading: no resolved signals yet · next evaluation at 50

Portfolio: 25 priced picks · in profit now 44% (95% CI 27–63%) · avg +1.2% (95% CI −2.7% to +12.6%) · next evaluation at 50

vs SPY over the same periods: +19.4% vs +1.4% · excess +18.0% (95% CI +5.9% to +43.0%) · p 0.055 · 5 picks · provisional

History (1)

Sep 23, 2026 Validating — Published and measured; not yet meeting the pay criteria.

Rotation v2

Validating

risk-adjusted momentum, inverse-vol, 25 instruments with 9 commodity ETFs, pack costs

Validating since Oct 3, 2026. Futures replaced by their ETFs, costs per venue in the backtest and the live record, closed sessions only. Its live NAV starts at its first signed rebalance.

A model portfolio, not a feed of picks: its live record, its backtest and every signed rebalance are on Rotation.

History (1)

Oct 3, 2026 Validating — Futures replaced by their ETFs, costs per venue in the backtest and the live record, closed sessions only. Its live NAV starts at its first signed rebalance.

Value screen (CL) v1

Validating

Validating since Sep 23, 2026.

Not yet certified Not certified yet. Its picks are still published; they will run in shadow once certification is enforced.

No backtest evidence has been recorded for this version.

Portfolio: 13 priced picks · in profit now 23% (95% CI 8–50%) · avg −0.7% (95% CI −1.7% to +0.9%) · next evaluation at 50

vs ECH over the same periods: +6.4% vs −0.1% · excess +6.4% · p 0.50 · 1 pick · provisional

History (1)

Sep 23, 2026 Validating — Published and measured; not yet meeting the pay criteria.

Value screen (US) v1

Validating

Validating since Sep 23, 2026.

Not yet certified Not certified yet. Its picks are still published; they will run in shadow once certification is enforced.

No backtest evidence has been recorded for this version.

Portfolio: 49 priced picks · in profit now 71% (95% CI 58–82%) · avg +2.6% (95% CI +0.4% to +5.2%) · next evaluation at 50

vs SPY over the same periods: +8.8% vs −0.2% · excess +8.9% (95% CI +7.3% to +10.6%) · p 0.005 · 16 picks · provisional

History (1)

Sep 23, 2026 Validating — Published and measured; not yet meeting the pay criteria.

Momentum Chile v1

In testing

12-1 momentum, top 5 of 31 Chilean large caps, monthly

In testing since Nov 4, 2026. A live test of a strategy whose 2013-2026 backtest was positive (+9.9%/yr over random portfolios) but is unconfirmed: it is not verifiably out of sample and does not survive the correction for the variants tried. Judged on its live record from 2026-11-01. Holdings stay private until it is promoted or retired; each month's hash is public.

Historical backtest (2013–2026, 164 months, not verifiably out of sample): +9.9%/yr over random (95% CI +3.4% to +16.6%).

Live: no months measured yet.

Evaluated at 12, 24 and 36 live months: promoted when the live excess over random is positive and the interval pooled with the historical backtest is above zero; retired when the live interval is below zero, or at 36 months without a promotion. Next evaluation at 12 months.

Frozen portfolios (1)

Oct 2026a1a121e33091eb53354cb95cbb3b8c9ac74542d5cb7db3b48d52fb32243d8db4

History (2)

Nov 4, 2026 In testing — A live test of a strategy whose 2013-2026 backtest was positive (+9.9%/yr over random portfolios) but is unconfirmed: it is not verifiably out of sample and does not survive the correction for the variants tried. Judged on its live record from 2026-11-01. Holdings stay private until it is promoted or retired; each month's hash is public.
Oct 1, 2026 In testing — A live test of the strategy that passed a pre-registered out-of-sample test on 2013-2026 (+9.9%/yr over random portfolios, 95% CI +3.4 to +16.6). Its holdings stay private until it is promoted or retired; each month's hash is public.

Quant screen v3

Paused

confluence+sma200

Paused since Nov 4, 2026. Shadow: no trading rule beat random timing out of sample (research/cert_v4). Picks are measured, not published.

Not yet certified Not certified yet. Its picks are still published; they will run in shadow once certification is enforced.

No backtest evidence has been recorded for this version.

No signals in the record yet.

In shadow — trading: no resolved signals yet

History (2)

Nov 4, 2026 Paused — Shadow: no trading rule beat random timing out of sample (research/cert_v4). Picks are measured, not published.
Nov 4, 2026 Validating — Started 2026-11-03: v2 selection plus the cluster risk cap.

Quant screen v1

Retired

confluence

Retired since Sep 23, 2026. Superseded by v2 on 2026-09-23 (no trend filter).

Trading: 16 resolved · win rate 56% (95% CI 33–77%) · avg +0.27R (95% CI −0.33R to +0.72R)

vs random timing: +0.27R vs +0.23R for random entry days (95% band −0.30R to +0.79R) · p 0.43 · 16 signals · provisional

History (2)

Sep 23, 2026 Retired — Superseded by v2 (confluence filtered by a strict SMA200 trend) on 2026-09-23.
Sep 23, 2026 Validating — Published and measured; not yet meeting the pay criteria.

Quant screen v2

Retired

confluence+sma200

Retired since Nov 4, 2026. Superseded by v3 (cluster risk cap) on 2026-11-03.

Trading: 10 resolved · win rate 40% (95% CI 17–69%) · avg −0.24R (95% CI −0.91R to +0.30R)

vs random timing: −0.74R vs −0.25R for random entry days (95% band −0.93R to +0.50R) · p 0.91 · 7 signals · provisional

Portfolio: 4 priced picks · in profit now 50% (95% CI 15–85%) · avg +0.0% (95% CI −5.2% to +10.0%)

vs SPY over the same periods: +3.6% vs −1.0% · excess +4.6% · p 0.50 · 1 pick · provisional

History (2)

Nov 4, 2026 Retired — Superseded by v3 (cluster risk cap) on 2026-11-03.
Sep 23, 2026 Validating — Confluence filtered by a strict SMA200 trend. Out-of-sample 2020-2026: avg R +0.05 -> +0.13, max drawdown 87R -> 37R; modest evidence, validating live.

Rotation v1

Retired

risk-adjusted momentum, inverse-vol, 25 instruments with 9 commodity futures, flat 0.1% fee

Retired since Oct 3, 2026. Superseded by v2: commodity ETFs instead of futures, pack costs, closed sessions only. Its rebalances stay in the history, labelled v1, without a live NAV.

See the v1 rebalances on Rotation.

History (2)

Oct 3, 2026 Retired — Superseded by v2: commodity ETFs instead of futures, pack costs, closed sessions only. Its rebalances stay in the history, labelled v1, without a live NAV.
Jul 12, 2026 Validating — Published and measured; not yet meeting the pay criteria.