01
At Closing
Your builder seeds a $25,000 account in our AI-powered quant hedge fund — a tax-neutral purchase-price adjustment, not a discount that disappears the moment you sign.
The Quant Club · Elevate Your Living. Amplify Your Wealth.
Your Home.
Your Alpha.
At closing, builders now give you a $25,000 seed account in an AI-powered quant hedge fund — instead of discounts that disappear the moment you sign. Welcome to the Quant Club: where luxury living meets intelligent, compounding wealth.
The Quant Club
A new asset class — built into the home.
A members-only AI trading lounge and hedge fund ecosystem inside master-planned communities. Residents access live AI agents, strategy sessions, and community alpha discussions. Dedicated garage infrastructure powers the fund 24/7.
This is more than an amenity — it's a new asset class.
How it works
From closing day to compounding.
Four steps. The first happens before you move in. The last continues for as long as you own the home.
02
The Lounge
Access your community's members-only trading floor: live AI strategy sessions, market intelligence, and community alpha discussions in a Starbucks-meets-Wall-Street setting.
03
AI Infrastructure
Garage-based compute clusters run our agent swarms 24/7 — institutional-grade execution, residentially delivered. Each community is its own micro-fund operations hub.
04
Compound Freedom
Watch your wealth grow while you live elevated. Real AI alpha that can help cover your mortgage — turning a home from a holding into a performance asset.
For Builders
Stop competing on price. Start competing on performance.
Deep discounts hurt your margin and reset the comp. A $25,000 seed account hands buyers a story that doesn't depreciate — and gives you recurring revenue from the fund itself.
- Replace deep discounts with a high-value $25k seed account
- Faster inventory velocity + higher closing rates
- New recurring revenue from the 2/20 hedge fund share — zero upfront cost
- Powerful differentiation in a buyer-saturated market
- Turnkey amenity: lounge build-out + AI infrastructure managed end-to-end
For Residents
Your home should compound wealth, not just hold value.
A discount evaporates the moment you sign. A $25,000 seed account in a working AI hedge fund keeps working — for the rest of the time you own the home.
- $25,000 performance seed account funded by your builder at closing
- On-site luxury trading lounge — community alpha, in person
- Real AI agents that can help pay your mortgage, not theory
- A community of like-minded homeowners building wealth together
- Future RWA tokenization for liquidity when you want it
The AI Edge
Institutional-grade AI. Residential delivery.
The math that runs sovereign funds, packaged for the community garage. Discipline first, returns second.
01
Live agent swarm
Multi-agent system executing in real markets — not paper trading. Continuous learning, continuous deployment.
02
Sub-10% drawdowns
Risk-first design. Strong risk-adjusted returns measured against decades of regime data — not last quarter.
03
EG Capital incubation
Backed by $500M+ of OCIO experience. Institutional discipline applied to a residential-scale product.
04
Garage-grade infra
Dedicated on-prem compute per community. Sovereign, low-latency, scalable across every Quant Club site.
Research transparency · snapshot 2026-05-20
The numbers behind the discipline.
A live look at the Quant Club research mill. Most candidates don't make the cut — that's the point. Our 4-gate Phase 5 filter is the strictest bar we know of in residential-scale quant finance.
01
0
Agents trained
Across multiple instruments, seeds, reward functions, and architectures — our research mill never sleeps.
02
0
Through the 4-gate filter
Each candidate is screened with walk-forward · CPCV · bar permutation · bootstrap CI — institutional rigor.
03
0
At 2 / 4 gates
Best post-eval cohort. Strong signal, not yet deployment-grade. Under further review.
04
0
Deployed without 4 / 4
Discipline is the product. We don't ship until every gate is passed.
View A · Pre-evaluation
Top 15 by in-sample training return.
⚠️ Directional only. These are the returns each agent saw during training — where it actively optimized to maximize. Real out-of-sample numbers are in View B.
| # | Symbol | Alias | Train Ret % | Episodes | Trades |
|---|---|---|---|---|---|
| 1 | NDX | phase_a · NDX · cvar · s7 · saf | +115.73% | 800 | 62,523 |
| 2 | COMPX | phase_a · COMPX (recover7) · s17 · saf | +103.39% | 800 | 63,789 |
| 3 | VIX | phase_a · VIX (recover5) · s7 · saf | +92.95% | 800 | 64,000 |
| 4 | — | saf_full · v_non_guided · s7 | +78.87% | 800 | 63,436 |
| 5 | NI225 | phase_a · NI225 (recover7) · s42 · saf | +76.29% | 800 | 61,651 |
| 6 | SPX | phase_a · SPX · s7 · saf | +68.13% | 800 | 61,064 |
| 7 | NI225 | phase_a · NI225 (recover5) · s99 · saf | +63.66% | 800 | 63,998 |
| 8 | — | saf_full · v_non_guided · s42 | +62.80% | 800 | 63,017 |
| 9 | NI225 | phase_a · NI225 (recover7) · s123 · saf | +60.24% | 800 | 61,046 |
| 10 | SPX | phase_a · SPX · s17 · saf | +58.99% | 800 | 64,000 |
| 11 | SPX | phase_a · SPX (smoke) · s7 · saf | +58.43% | 800 | 63,800 |
| 12 | SPX | phase_a · SPX · s99 · saf | +53.33% | 800 | 55,365 |
| 13 | SPX | phase_a · SPX (smoke) · s42 · saf | +46.39% | 800 | 62,948 |
| 14 | NDX | phase_a · NDX · s17 · saf | +43.45% | 800 | 63,897 |
| 15 | COMPX | phase_a · COMPX (recover7) · s123 · saf | +40.19% | 800 | 63,751 |
View B · Out-of-sample evaluation
All 28 evaluated agents.
The Gates column is the only deployment signal. Sorted by eval return %. None deployed until 4 / 4.
| # | Symbol | Alias | Eval Ret % | Sharpe | Gates | Note |
|---|---|---|---|---|---|---|
| 1 | VIX | VIX · s17 · saf | +99.47% | +0.890 | 1/4 | regime-dependent (vol explosion) |
| 2 | VIX | VIX · s7 · saf | +95.51% | +0.884 | 1/4 | regime-dependent |
| 3 | COMPX | COMPX · s42 · saf | +25.46% | +1.159 | 2/4 | buy-and-hold suspect |
| 4 | COMPX | COMPX · s99 · saf | +24.70% | +1.169 | 2/4 | buy-and-hold suspect |
| 5 | NDX | NDX · s17 · saf | +24.48% | +1.034 | 2/4 | buy-and-hold suspect |
| 6 | COMPX | COMPX · s17 · saf | +16.77% | +0.954 | 0/4 | all gates fail |
| 7 | — | ensemble_saf_s99 + nocond_s123 (50/50) | +12.62% | +1.663 | 2/4 | best Sharpe overall |
| 8 | — | saf_full · v_non_guided · s99 | +12.62% | +1.431 | 2/4 | strong saf-flavor |
| 9 | NI225 | NI225 · s99 · saf | +9.39% | +0.698 | 1/4 | — |
| 10 | SPX | SPX · s17 · saf | +9.23% | +0.959 | 0/4 | — |
| 11 | — | saf_option_c · v_non_guided · s99 | +8.39% | +0.843 | 0/4 | — |
| 12 | NDX | NDX · s7 · saf | +8.13% | +0.640 | 0/4 | — |
| 13 | NI225 | NI225 · s7 · saf | +6.40% | +1.405 | 2/4 | — |
| 14 | SPX | SPX · s7 · saf | +2.58% | +0.696 | 0/4 | — |
| 15 | SPX | SPX · s99 · saf | +2.44% | +1.037 | 1/4 | — |
| 16 | — | pairs_trading · w60 · entry 2.0 · exit 0.5 (rolling β) | +2.24% | — | 0/3 | first multi-symbol pilot |
| 17 | COMPX | _to_penalty_test · COMPX · s99 · timeout=0.02 | +0.00% | +0.000 | 1/4 | degenerate (never trades) |
| 18 | SPX | SPX · s42 · saf | +0.00% | +0.000 | 1/4 | degenerate |
| 19 | COMPX | COMPX · s123 · saf | +0.00% | +0.000 | 1/4 | degenerate |
| 20 | KC | KC · s17 · saf | +0.00% | +0.000 | 1/4 | degenerate |
| 21 | NI225 | NI225 · s17 · saf | -1.14% | +0.408 | 0/4 | — |
| 22 | CT | CT · s123 · saf | -2.27% | +0.052 | 0/4 | — |
| 23 | NI225 | NI225 · s42 · saf | -3.60% | +0.970 | 0/4 | — |
| 24 | SPX | SPX · s123 · saf | -4.08% | +0.262 | 0/4 | — |
| 25 | NDX | NDX · s42 · saf | -4.14% | +0.010 | 0/4 | — |
| 26 | — | variant_matrix_01 · v_non_guided · s99 | -9.26% | +0.225 | 0/4 | — |
| 27 | COMPX | COMPX · s7 · saf | -19.03% | -1.217 | 0/4 | anti-pattern |
| 28 | SB | SB · s123 · saf | -19.23% | +0.010 | 0/4 | — |
Methodology
How we screen every candidate.
Every candidate that lands on the Research Leaderboard has to survive a five-stage statistical screening. Walk-forward and CPCV ask whether the edge is consistent over time. Bar permutation and action permutation ask whether the edge is real — not survivor bias, not market drift. Bootstrap CI asks whether we can statistically distinguish the result from zero. Pass all four primary gates and you're deployable. Fail any one and you stay in the lab.
01Gate
Walk-Forward
Does the agent's edge hold across time?
We chop the out-of-sample window into 4 non-overlapping segments. Each segment is backtested independently. We check both the median Sharpe across segments AND the consistency (% of segments with positive Sharpe).
Pass rule
Median Sharpe ≥ 0.5 AND consistency ≥ 60%
What failure looks like
“Big winner in 2023, big loser in 2024 → fails consistency. Edge that requires one specific regime isn't deployment-grade.”
02Gate
CPCV
Is the edge robust to data ordering?
Combinatorial Purged Cross-Validation — 10 partitions, 2-at-a-time held out, 45 folds total. Purges adjacent bars to prevent leak. Tests whether removing any combination of segments breaks the strategy.
Pass rule
Min Sharpe > -0.5 AND IQR < 1.0 across the 45 folds
What failure looks like
“Strategy works on a continuous timeline but blows up when you remove one specific 6-month chunk → catastrophic regime dependency.”
03Gate
Bar Permutation
Is the Sharpe better than random bar order?
We shuffle the BAR ORDER 200 times (block-permutation, preserves autocorrelation). For each shuffle, the agent runs again and we record its Sharpe. The actual Sharpe must be in the top tail of the distribution.
Pass rule
p-value < 0.05 (actual Sharpe at or above the 95th percentile of shuffled Sharpes)
What failure looks like
“A 'lucky' agent that scored high on the real data but matches the distribution of randomly-shuffled-data Sharpes → no real edge, just survivor bias.”
04Gate
Action Permutation
Does the agent's TIMING actually matter?
We keep the bars, but shuffle the agent's POSITION sequence 200 times. If the shuffled-actions Sharpe matches the real Sharpe, the agent isn't making good decisions — it's riding market drift (buy-and-hold in disguise).
Pass rule
p-value < 0.05 (actual Sharpe > 95th percentile of shuffled-action Sharpes)
What failure looks like
“An agent that's always long when the market is up → shuffling its actions still produces a similar Sharpe. The 'strategy' is just market beta with extra steps.”
05Confidence check
Bootstrap CI
Is the Sharpe statistically distinguishable from zero?
Block-bootstrap the trade returns 500 times to build a 95% confidence interval around the actual Sharpe. We require the LOWER bound of that CI to be strictly positive.
Pass rule
Lower 95% CI bound > 0
What failure looks like
“Sharpe of +1.2 looks great, but with a 95% CI of [-0.4, +2.8] we can't rule out that the true Sharpe is zero or negative.”
Why five stages? Each one catches a different failure mode that the others miss. A single-gate filter would let through agents that look great on one axis but fail on another. The Research Leaderboard above shows the gate count next to every candidate — that's the only number that matters for deployment readiness.
Our Story
From dream home to performance asset.
For most families, buying a home is the largest financial decision they ever make — and the moment that decision closes, the discounts evaporate. The walk-in pantry stays. The wealth curve doesn't.
Lifestyle Alpha was built to change that math. By embedding a working AI hedge fund into the home itself — funded by the builder at closing — the dream home becomes a wealth-compounding asset on day one. The lounge makes it visible; the agents make it real.
We're incubated by EG Capital — half a billion in OCIO experience — and we run live, not in backtest. This isn't a course or a community Slack. It's an amenity that ships money.
Founder note.
Get in touch
Join the Quant Club.
Pick your lane. We reply within 48 business hours.