Does ARK Size Positions by Business Quality? Not by Ours.
We ranked every scored holding in ARK's three published funds by weight and by Blitz Index score. In all three funds the two rankings are unrelated. That is not a criticism of either ruler — it is evidence that they measure different things.
The question
A concentrated manager decides two things: what to own, and how much. The Blitz Index speaks only to the first — whether a business is built to win its category. So a natural question for the ARK Lens is whether ARK's sizing lines up with our grading. Are the biggest positions the highest-scored businesses?
This is a descriptive check on a single day's holdings. It was not pre-registered, and we report it as a description, not a test of a hypothesis.
Method
For each published fund we took every holding with a Blitz Index score — 26 in ARKK, 24 in ARKW, 19 in ARKF — and computed the Spearman rank correlation between portfolio weight and score. Gate failures, unscored names, private positions and fund positions are excluded, as they are from every Lens aggregate.
The result
| Fund | Scored holdings | Spearman ρ (weight vs score) | Approx. 5% significance threshold |
|---|---|---|---|
| ARKK | 26 | −0.081 | ±0.39 |
| ARKW | 24 | −0.049 | ±0.41 |
| ARKF | 19 | +0.083 | ±0.46 |
All three coefficients are close to zero and far inside the band that samples of this size produce by chance. Inside each fund, knowing a holding's score tells you nothing about its weight, and vice versa.
Two cross-checks point the same way:
| Fund | Weighted avg. score, top-10 scored holdings | Weighted avg. score, the rest | Equal-weighted vs weight-averaged score |
|---|---|---|---|
| ARKK | 60.2 (8 names) | 58.7 (18 names) | 59.8 vs 59.5 |
| ARKW | 67.9 (6 names) | 63.0 (18 names) | 65.2 vs 65.3 |
| ARKF | 64.2 (7 names) | 67.9 (12 names) | 65.4 vs 65.7 |
In ARKW the top positions score higher than the rest; in ARKF they score lower; in ARKK the difference is a point and a half. No consistent direction. And in every fund, weighting by position size moves the average by less than half a point compared with weighting every holding equally.
What it looks like in practice
The pattern is visible in the largest positions themselves:
- ARKK's top five scored holdings are Tesla (62), Tempus AI (55), Circle (52), CRISPR Therapeutics (48) and Coinbase (61). Its highest-scored holding, NVIDIA (87), is its ninth-largest position at 3.03%.
- ARKF's largest holding is Block (8.64%), a G2 gate failure that carries no score. Its highest-scored holding, NVIDIA, sits at 3.10%.
- ARKW comes closest to aligning: Meta (78) and NVIDIA (87) are its second- and fifth-largest positions. But Circle (52) sits fourth, ahead of NVIDIA.
One week later: the same check on 6 November holdings
The result above is a single day’s snapshot, so the natural follow-up is to rerun it. On holdings of 11/06/2026, against the current published score record, the same Spearman calculation gives a different-looking table — and the three new funds can be included for the first time.
| Fund | Scored holdings | Spearman ρ (weight vs score) |
|---|---|---|
| ARKK | 29 | +0.105 |
| ARKW | 28 | +0.020 |
| ARKF | 23 | +0.403 |
| ARKQ | 21 | +0.535 |
| ARKX | 17 | +0.107 |
| ARKG | 20 | +0.324 |
Two things changed. First, the counts grew: ARKK from 26 scored holdings to 29, ARKW from 24 to 28, ARKF from 19 to 23, as new scores entered the published record — SpaceX (60) among them — and as holdings moved. Second, every coefficient is now at or above zero, where the 11/04 table had two negatives and one small positive. The largest move is ARKF, from +0.083 to +0.403: Shopify (71) is now ARKF’s largest scored holding at 8.47%, and Meta (78) sits at 4.83%, so the fund’s biggest positions happen, on this day, to be its better-scored ones. ARKQ’s +0.535 is the only coefficient outside the rough 5% significance band for its sample size — but it rests on 21 names and a single day’s weights.
Does that overturn the conclusion? Not on this evidence. ARKW is still +0.020 — effectively zero — and ARKK and ARKX sit near +0.11. Three of six funds show no relationship at all, and the three that show a positive one are the three smallest samples in the table. A coefficient that moves by 0.32 in a week, in a fund of 23 names, is measuring the week as much as the manager. The honest update is narrower: the near-zero of 11/04 has drifted mildly positive by 11/06, driven by a few large positions, and the drift is inconsistent across funds and across days.
Where the weight actually sits now
The largest scored holding in each fund on 11/06/2026 makes the point more directly than any coefficient:
| Fund | Largest scored holding | Weight | Score |
|---|---|---|---|
| ARKK | Tesla | 10.74% | 62 |
| ARKW | Tesla | 8.30% | 62 |
| ARKF | Shopify | 8.47% | 71 |
| ARKQ | Tesla | 12.44% | 62 |
| ARKX | SpaceX | 10.45% | 60 |
| ARKG | Tempus AI | 9.77% | 55 |
Tesla, at 62, is the largest scored holding in three of the six funds — including ARKQ, where it is 12.44% of the fund. SpaceX (60) leads ARKX at 10.45%, and Tempus AI (55) leads ARKG at 9.77%. None of the six funds is led by a 65+ name, and none is led by its highest-scored holding. NVIDIA (87) sits at 3.70% in ARKK, 4.21% in ARKW, 2.79% in ARKF and 5.73% in ARKQ — a meaningful position in every fund, and nowhere near the top of any of them. If sizing tracked our scores, the table above would be led by NVIDIA in four funds. It is led by mid-table B and C names instead, which is the same pattern the 11/04 article found, stated in weights rather than in ranks.
Why a single day can move ρ by 0.4
Three mechanics explain the week’s movement, and all three are reasons to distrust any single reading. First, the samples are small: with 17 to 29 names, one position changing rank moves the coefficient by several hundredths, and a handful of moves add up to the 0.32 shift ARKF shows. Second, weights move daily with prices even when the manager does nothing — a rank correlation on holdings is partly a rank correlation on the week’s returns. Third, the scored set itself moves: new scores enter the record (SpaceX’s 60 was published after the 11/04 snapshot), and each entrant reshuffles the ranks of everything below it. The 11/04 conclusion was never that the true correlation is exactly zero; it was that, on the evidence of one day, sizing and grading looked unrelated. A week later the evidence is noisier and mildly positive, which is consistent with a relationship that is weak, unstable, or absent — and not consistent with a manager that sizes by this ruler.
Why near-zero is the expected answer
Our own research program already explains why a manager's sizing and our score should not line up. The Blitz Index is built to ignore price. In three pre-registered tests — crossing the 65 line, the profitability turn, and score momentum — no score-derived signal predicted returns, and the common thread was entry multiple and market cycle. Our ARK backtest found the same at the 2021 peak: PayPal, our highest score, fell 78% because the business call was right and the price was not.
A manager who sizes positions on expected return should, rationally, hold a lower-scored business at a larger weight when the price looks attractive, and a higher-scored business at a smaller weight when the price looks full. If ARK's weights did track our scores closely, that would be the surprising result — it would suggest either that ARK ignores price or that our score secretly encodes it.
So the zero is informative in both directions. It says the Lens is not a mirror of ARK's convictions, and it says ARK's convictions are not a proxy for business quality as we define it. Readers who want both views need both instruments.
Sensitivity
Alphabet's Class C shares (GOOG) are now mapped to Alphabet's score of 83 on our company index, and the figures above already include that mapping. Before the mapping, ρ was +0.015 in ARKK and −0.102 in ARKW. The conclusion does not move.
The boundary
One day, three funds, 69 positions. Weights move daily with prices and trades, and a cross-section cannot say anything about how ARK changes sizing over time. The correlation is a rank statistic and ignores the size of gaps between scores. Nothing here evaluates whether any position is sized well — that question depends on price, which the Blitz Index deliberately does not measure.
BlitzLab scores businesses against a scholar's standard — not investment advice. No price targets, no buy/sell recommendations. Holdings come from ARK Invest's daily public disclosures; BlitzLab is not affiliated with ARK Invest and this research is not endorsed by them.
Sources
- BlitzLab, ARK Lens fund pages — https://blitzlab.app/ark/arkk/ · https://blitzlab.app/ark/arkw/ · https://blitzlab.app/ark/arkf/ — holdings as of 11/04/2026 — checked 2026-11-05
- BlitzLab, "What the Blitz Index Measures — and What It Doesn't" — https://blitzlab.app/research/synthesis — 2026-10-15 — checked 2026-11-05
- BlitzLab, "Can Our Model Tell ARK's Winners from Its Blowups?" — https://blitzlab.app/research/ark-backtest — 2026-10-15 — checked 2026-11-05
- BlitzLab, "Alphabet — Blitz Index 83 (A)" — https://blitzlab.app/companies/googl/ — checked 2026-11-05
- Calculation file: ark_lens_holdings_2026-11-04.csv, build.py (BlitzLab internal, attached to this draft)