Portfolio Research · NQ Day Trade

Why More Systems Isn’t Always Better

What correlations and drawdowns reveal about Capstone’s 2, 3, 7, 18, and 25 System NQ portfolios

David Bean · Capstone Trading · July 2026

When traders look at Capstone’s NQ day-trade portfolios, a natural question comes up:

If the 25 System Portfolio has more systems — and more total profit — why does the 18 System Portfolio often look like the stronger book?

This post walks through that research from the start: what we compared, how we measured diversification, what the numbers actually say, and what it means when you’re choosing a portfolio tier.

Short answer: The 18 System Portfolio wins on risk-adjusted performance — not because its average correlation is magically lower, but because it avoids stacking too many near-duplicate systems that lose on the same days.

Net / Max Drawdown by portfolio — Figure 1. Full-sample net ÷ max drawdown. 18 System leads at 40.6 — well ahead of 7 (33.3) and 25 (23.8).

The portfolios we compared

Capstone’s NQ product ladder is a set of equal-weight day-trade baskets. Each included system trades Nasdaq futures (or micros) and is flat by the end of the day. Results use closed trades with a consistent cost model.

Portfolio	Systems	Role
2 System Portfolio NQ	2	Focused entry tier
3 System Portfolio NQ	3	Focused + Open Range
7 System Portfolio NQ	7	Core NQ basket
18 System Portfolio NQ (Stock Index Portfolio 18)	18	Advanced index day-trade basket
25 System Portfolio NQ	25	Largest NQ bundle

Study window: Jan 1, 2020 – Jul 23, 2026 · Costs: $10 slippage + $2.50 commission per side · Weights: equal

Period	Dates	Purpose
Full sample	Jan 2020 – Jul 2026	Overall quality
In-sample (IS)	Jan 2020 – May 31, 2025	Longer “known” history
Out-of-sample (OOS)	Jun 1, 2025 – Jul 23, 2026	Recent holdout

What “better” should mean

Raw net profit favors larger baskets. That isn’t surprising — more systems take more trades.

For portfolio design, we care more about:

Net / Max Drawdown (net/DD) — profit earned per dollar of worst peak-to-trough pain
Average trade — quality per trade after costs
Correlation structure — whether systems truly diversify, or just look different on a product sheet

If Portfolio A makes more money but doubles the drawdown, it is not automatically the better portfolio.

Step 1 — Score the portfolios on performance

Net profit vs max drawdown — Figure 2. 25 System posts the highest net P/L (~$3.41M) but more than doubles 18 System’s max drawdown (~$143k vs ~$61k).

Portfolio	Net P/L	Avg trade	Max DD	Net / Max DD
2 System	$365,850	$171	$15,945	22.9
3 System	$688,665	$222	$28,110	24.5
7 System	$1,223,785	$166	$36,720	33.3
18 System	$2,492,030	$140	$61,345	40.6
25 System	$3,409,315	$161	$143,275	23.8

What jumps out

25 System makes the most money (~$3.41M).
18 System has the best net/DD (40.6) — clearly ahead of 7 (33.3) and far ahead of 25 (23.8).
25 System’s drawdown is the problem. Roughly $143k max DD vs about $61k for 18. That more-than-doubling of pain collapses risk-adjusted quality below even the 3 System tier.

The puzzle is not “why does 18 make more than 25?” It doesn’t. The puzzle is: why does adding systems from 18 → 25 hurt risk-adjusted performance so badly?

Step 2 — Measure diversification the right way

For each portfolio we build each system’s daily P/L series, compute the Pearson correlation between every pair, then summarize with:

Average pairwise correlation — the typical relationship
Max pairwise correlation — the worst near-duplicate pair
% of pairs above 0.30 and 0.50 — how thick the “highly related” tail is

A low average correlation is good. But average alone can hide a dangerous cluster of near-clones.

Correlation structure by portfolio — Figure 3. Average correlation is similar for 7/18/25 — and slightly lower for 25 than 18. The discriminating signal is the high-correlation tail.

Portfolio	Avg corr	Max corr	% pairs >0.30	% pairs >0.50	Net/DD
2 System	−0.062	−0.062	0%	0%	22.9
3 System	−0.007	0.11	0%	0%	24.5
7 System	0.072	0.53	4.8%	4.8%	33.3
18 System	0.076	0.84	3.3%	0.7%	40.6
25 System	0.067	0.91	9.0%	3.7%	23.8

Conclusion #1: Average correlation is useful — but it is not why 18 beats 25. The discriminating signal is the tail: max correlation and how many pairs sit in the “too similar” zone.

Step 3 — Look inside the strategy families

Systems are not all independent ideas. Many belong to families that respond to the same kind of market day — Gap Continue, VReversal, VSD, Cobra, Open Range/Chop, and Other (TICK, Momentum, AT, etc.).

Family mix 18 vs 25 — Figure 4. 18 System keeps Gap/VRev modest and leans on seven distinct “Other” edges. 25 System packs 48% of the book into Gap + VReversal.

V-Reversal loss stacking — Figure 5. On VReversal family-loss days, 25 System averages 3.2 members losing together — and 87% of those days have two or more losers. 18 System averages 1.3.

Hottest correlated pairs

18 System pair	Corr	25 System pair	Corr
Cobra III 2020 ↔ Cobra III 2022	0.84	VReversal 2024 ↔ VReversal ATS	0.91
Gap Cont 2024 ↔ Gap Cont 2020	0.46	VReversal I ↔ VReversal II	0.71
Open Range ↔ Pulse	0.42	Gap Cont 2024 ↔ Gap Cont 2019	0.69
VRev ATS ↔ VRev IV 2024	0.39	VRev AGR ↔ VReversal I	0.68
Gap Cont 2024 ↔ Pulse	0.36	Gap Cont 2020 ↔ Gap Cont 2020 (700SL)	0.66

Conclusion #2: 18 System’s edge is diversification quality — fewer near-duplicates, more independent edges, less same-day loss stacking. 25 System’s hottest relationships are mostly variants of the same idea.

Step 4 — Compare 18 to each peer

vs 2 System and 3 System

These are excellent focused portfolios: best pairwise diversification in the study, and the strongest out-of-sample average-trade retention (~0.83× of in-sample). Their limitation is fewer independent edges — full-sample net/DD tops out around 23–25 vs 40.6 for 18.

vs 7 System

Closest rival. OOS net/DD is almost identical (~3.0). Seven retains average trade better out of sample (0.53× vs 0.38×). Eighteen wins the full-sample net/DD race via breadth beyond the seven-system core. Seven’s main correlation wart: VReversal AGR ↔ ATS at ~0.53.

vs 25 System

This is where “more systems” becomes expensive. Twenty-five adds near-duplicates of edges 18 already has. Absolute profit rises; max drawdown jumps from ~$61k to ~$143k; full-sample net/DD falls from 40.6 → 23.8.

Annual net profit and loss by year — Figure 6. Annual net P/L ($k). In 2026 YTD through July 23, 18 System is still compounding (~+$121k) while 25 System is largely consumed by drawdown (~+$32k).

Step 5 — Check recent / out-of-sample behavior

Full-sample leadership does not mean every window looks identical.

Out-of-sample retention — Figure 7. OOS average trade and OOS÷IS retention. Smaller books held trade quality best recently; 18 still ties 7 on OOS net/DD; 25 lags at 1.90.

Portfolio	OOS avg	OOS net/DD	OOS÷IS avg
2 System	$148	4.01	0.83
3 System	$191	4.12	0.84
7 System	$95	3.02	0.53
18 System	$60	3.03	0.38
25 System	$73	1.90	0.41

The practical takeaway

What this research does not say: never trade 25; that average correlation is useless; or that 2/3 are “worse.” They are different products — focused, highly diversified books with excellent recent trade-quality retention.

What it does say:

Diversification quality beats system count.
Watch the clone tail, not just average correlation.
Same-family loss stacking is how “more systems” quietly inflate max drawdown.
Among these NQ day-trade tiers, 18 System is the risk-adjusted sweet spot.

Portfolio research summary — Figure 8. Bottom line — diversification quality beats system count.

Best full-sample net/DD: 18 Best OOS avg retention: 2 / 3 Closest rival: 7 Worst MDD / clone tax: 25

The 18 System Portfolio isn’t better because it has a lower average correlation — it’s better because it diversifies with more independent edges and fewer near-duplicates that lose on the same day.

Methodology: Capstone PortfolioMetrics closed-trade database · Pearson correlation of daily strategy P/L · equal-weight membership from live portfolio weights · family labels by strategy name · hypothetical results; past performance does not guarantee future results.

Portfolio Correlation Studies and Out of Sample Analysis