PLEX Portfolio Comparison, Ticker Inventory & Terms

Four PLEX portfolios, same engines

Every column is recomputed in the same pass through the same engines (plex_review_analyze.bucket_holdings/bucket_split, plex_review_analyze.analyze_yield (m1_yield_tally), plex_review_analyze.analyze_drawdown (m1_drawdow…). Allocations as of 2026-09-23; market metrics as of 2026-09-22 (carried). Balances are withheld — percentages and ratios only.

Why you will see more than one “Bedrock %” — all of them, reconciled

Three different methods label the same holdings, and they disagree. None of them is a typo. Every page here that prints a bucket percentage says which of these three it is using, and this table is generated from the one shared source those pages read (tools/lib_plex_schemes.py).

M1's own pie sleeves: What Joseph actually told M1 to build. Percentages are the live slice targets and the live dollar split — never the hand-typed slice names, which go stale the moment a target changes. Source: data/m1-holdings-live.json, as of 2026-09-23.

Our per-holding classifier: Re-decides every holding on its own merits and ignores which sleeve it sits in. The Academy config tiers business development companies (BDCs) as cash flow, which is why it disagrees with the pie. Source: data/plex-review/compare-all.json, as of 2026-09-23.

Fynanc Analyzer engine: A third party’s engine reading the same 13 holdings at the same weights. It has a fourth bucket, Other, and its optimizer prints a Cash-Flow target of 20% rather than the Academy’s 34% — so its “target” column is its own, not the Academy’s. Source: data/plex-flow/portfolio-analyzer-snapshot.json, as of 2026-09-22.

The one thing that is NOT a disagreement: a low number in the classifier rows is not evidence the account drifted off target. Measured against the sleeves the owner actually built, the account is within a fraction of a point of its stated Bedrock target. Which labeling scheme should govern is a methodology decision and it is the owner’s to make — this site presents all three rather than picking one silently.

Every holding, under all three labeling schemes

Weights are the live M1 pie weights. Balances are withheld — percentages, yields and verdicts only. As of: allocations 2026-09-23 · market metrics 2026-09-22 (carried) · Fynanc engine 2026-09-22.

Shaded rows are where the three schemes do not agree. “Fallback, not a rule” means our classifier found no config tier and no matching fund-name keyword, so the holding landed in the cash-flow default — an absence of evidence, not a determination.

The four buckets — and the rule that decides each

The exact test applied to every row in the inventory table below. HEDGE if correlation to the S&P is below 0. Otherwise the stability screens are volatility ≤ 10, correlation ≤ 0.4 and broker maintenance requirement ≤ 35%; the yield gate is ≥ 7%. All three stability screens plus the yield gate = BOTH; stability only = BEDROCK; yield only = CASH FLOW; neither = n/a (never guessed).

The full screened ticker inventory

Every fund in the universe was cycled through the funnel one ticker at a time. 1,998 reviewed → 1,998 (gate0_aum_ge_100m__n_reviewed) → 1,640 (gate1_debt_income__n1) → 559 (gate2_yield_expense__n2) → 340 (gate3_vol_corr__n3) → 278 (gate4_m1_marginable__final). The maintenance requirement is the broker’s own published figure per fund.

Gate drop summary — and the last date each gate’s data was pulled

Every fund in the universe is cycled through the funnel one at a time and each drop is logged against the ticker. 1,998 reviewed. The screen ran 2026-09-22, but each gate reads a different feed — so each row below carries the date its own data was actually fetched, not the date the screen ran. Convention: a gate that draws on more than one feed shows the OLDEST of them — a gate is only as fresh as its stalest input.

2 of 5 gates are flagged. A flagged gate means the data behind it is older than the screen run, or the feed came back partial, failed or was skipped. The count is still shown — it is simply not backed by data as fresh as the headline date.

Feed detail. M1 GraphQL viewer.screenFunds: 2026-09-22 17:26 — 24 page(s); M1 reports total=2342, pulled=2342 · StockAnalysis screener API: 2026-09-22 17:26 — all 8 columns returned · yfinance marketCap proxy (CEF AUM recovery): 2026-09-22 17:28, partial — AUM recovered for 237/238 non-screener funds · yfinance 3y vs SPY (vol / corr / drawdown): 2026-09-22 17:30, partial — metrics computed for 1637/1640 Gate-1 survivors (3y vs SPY) -- these DECIDE the sleeve, they no longer merely grade it

The ETF review, broken down by bucket

The 278 funds that survived all four gates, split by the tier the screen assigned. Screen as of 2026-09-22; each figure below is the median across that tier, computed from the same artifact that fills the inventory table underneath. Balances are withheld — this is the candidate universe, not anyone’s account.

This is the screen’s tier, which is a third labelling of the same kind as the ones reconciled above — it is the gate the universe sweep applied, not the Academy ticker config and not the per-row rule used in the inventory table below. Where a fund carries no tier it is shown as untiered rather than folded into a bucket.

Yield − max drawdown is calculated on this page: the trailing distribution yield minus the absolute 3-year worst peak-to-trough fall. It asks how much income a holding paid you relative to the worst hole it put you in. Ticker links use the only working per-ticker form on M1 — https://dashboard.m1.com/d/research/{kind}/details/{enc_id} — with the relay id resolved live (314 resolved this pass). M1 has no symbol-based deep link, so a ticker with no resolved id renders as plain text rather than a broken link.

Definitions — and where each term evolved

Each term links into https://www.fynancacademy.com at the session where the rule last moved, with the date and the time index inside that call. Sourced from the existing glossary generator’s data (data/fynanc-academy/glossary-evolution.json) and the session post map (data/fynanc-academy/plex-calls-post-map.json).