Skip to content
  • §00 · QHAM
  • 5 MARKETS · US CA IN CN JP
  • EST. ONTARIO

Disciplined investing, by design.

Quantitative · Hedged · Adaptive · Multifactor.Built in Ontario, investing across the US, Canada, India, China and Japan. International risks

Talk to usHow the strategy works

// OPERATING PRINCIPLES

Six rules behind every position.

  1. P/01

    Evidence before conviction

    An idea earns capital only after it survives out-of-sample testing.

    QHAM: Q
  2. P/02

    Risk is decided first

    We set how much we are prepared to lose before we decide what to own.

    QHAM: H
  3. P/03

    Many drivers, no single bet

    Returns come from a blend of independent factors, never one call.

    QHAM: M
  4. P/04

    Adapt, don’t predict

    We respond to measured shifts in market regime instead of forecasting headlines.

    QHAM: A
  5. P/05

    Protection is permanent

    The hedge overlay is always on — not something we reach for in a panic.

    QHAM: H
  6. P/06

    People own the model

    Machine learning helps us find and test ideas; people set every assumption, review every change and can halt anything live instantly.

    QHAM: ALL
SHIPRA://QHAM — 4 MODULESSIMULATED · ILLUSTRATIVE 1

Four parts, one process.

Each module below runs a real, stated model on simulated data so you can see the mechanism, not a marketing animation.

MODULE 1/4 · SHIPRA://Q/SIGNAL

Quantitative

Every position starts as a measurable hypothesis, tested on data before it is trusted.

We turn raw market and company data into standardised signals: each characteristic is scaled across the investable universe, extreme values are capped, and the result is ranked. A signal is only kept if its relationship with future returns holds on data it was not built on.

Machine learning is used in our research and model development, alongside classical statistics. People choose the inputs, set the assumptions and review every change before anything reaches a live portfolio.

SHIPRA://Q/SIGNALSIMULATED · ILLUSTRATIVE
DISTRIBUTION OF Z-SCORES (WINSORISED ±3σ)
−3σ0+3σ
MEAN FORWARD RETURN BY SIGNAL QUINTILE
  • Q1−0.74%
  • Q2−0.47%
  • Q3−0.20%
  • Q4+0.75%
  • Q5+0.73%
n
500
IC
0.180
Q5−Q1
+1.48%
z = (x − μ)/σ on log(x) · winsorised ±3σ · n=500 · fwd = 0.004·z + 0.03·ε

MODULE 2/4 · SHIPRA://M/EXPOSURE

Multifactor

Returns come from several factors that do not move together, so no single bet dominates.

Value, momentum, quality, low volatility and size each reward investors for different reasons and tend to perform at different times. Combining factors with low or negative correlation lowers the volatility of the whole portfolio below that of its parts.

The panel computes portfolio volatility directly as the square root of wᵀΣw — the weights applied to the covariance of the factors.

SHIPRA://M/EXPOSURESIMULATED · ILLUSTRATIVE
Factor correlation matrix used for the portfolio
VALMOMQUALVOLSIZE
VAL1−.35.10.15.20
MOM−.351.15.05−.10
QUA.10.151.35−.15
LVOL.15.05.351−.25
SIZE.20−.10−.15−.251
WEIGHTS w (INVERSE VOLATILITY)
  • VAL19.0%
  • MOM14.2%
  • QUA28.4%
  • LVOL21.3%
  • SIZE17.1%
σ_p = √(wᵀΣw)
3.93%
Σwᵢσᵢ
8.53%
DIV. RATIO
2.17×
σ_p = √(wᵀΣw) · Σ = D·R·D · w ∝ 1/σᵢ (inverse-volatility) · σ = [9 12 6 8 10]% ann.

MODULE 3/4 · SHIPRA://A/REGIME

Adaptive

The portfolio responds to measured changes in market regime rather than forecasts.

Markets alternate between calmer and more stressed periods with very different behaviour. A regime model estimates, day by day, the probability that markets are in a stressed state, using only information available at the time.

When that probability crosses a threshold, factor weights and risk settings step to their stressed-regime values. Nothing depends on predicting the news.

SHIPRA://A/REGIMESIMULATED · ILLUSTRATIVE
Simulated price path (index = 100)
Filtered P(risk-off) · threshold 0.5
DAYS
750
RISK-OFF DAYS
263
LAST P
0.01
model: 2-state Gaussian HMM · μ₁=+0.08 σ₁=0.12 · μ₂=−0.25 σ₂=0.30 (ann.) · P=[[.985 .015][.05 .95]] · forward filter · seed 0x5A1F

But there’s one more thing…

MODULE 4/4 · SHIPRA://H/DRAWDOWN

Hedged

A permanent overlay reduces exposure as estimated risk rises, aiming to limit drawdowns.

The hedge overlay is always present. In calmer regimes a small portion of market exposure is offset; when the regime model signals stress, the hedge ratio increases. Hedging has a cost, and it reduces upside as well as downside.

The simulation shows the same return path with and without the overlay, and the drawdowns of each.

SHIPRA://H/DRAWDOWNSIMULATED · ILLUSTRATIVE
  • Unhedged
  • Hedged overlay
Underwater (drawdown from peak)
MDD UNHEDGED
−19.3%
MDD HEDGED
−17.6%
Δ
1.8 pts
ROBUSTNESS
MDD lower in 196/200 paths
h = 0.15 → 0.6 when P(risk-off)ₜ₋₁ > 0.5 · r_h = (1 − h)·r − 0.4bp/day · same path as A · seed 0x5A1F

SPEC SHEET

FACTORS IN THE MODEL
5
MARKETS COVERED
5
REGIMES MODELLED
2
HEDGE OVERLAY
Always on

All panels use simulated data from the stated models and are not actual or hypothetical performance. Read the disclosure.

// THE PIPELINE

Data → Signal → Risk & Hedge → Execution

Every dot is an idea. Most are rejected along the way — that is the discipline working.

SHIPRA://PIPELINESIMULATED · ILLUSTRATIVEideas tested 115 · in production 61
model: each idea passes gate k with probability pₖ (Bernoulli) · n=115 · p=[.42 .31 .55 .80]

Of 115 ideas tested, 6 reached production (illustrative figures).

↺ Execution → Data · measured, reviewed, improved — continuously

Custodian, auditor and administrator details are provided to qualified investors on request. Request them.

// FOLLOW THE SUN

24 hours of QHAM

Five markets, two continents. A day in the life of the system — not a person.

00061218
  • 1TSETokyo · 09:00–11:30, 12:30–15:30 local
  • 2SSEShanghai · 09:30–11:30, 13:00–15:00 local
  • 3NSEMumbai · 09:15–15:30 local
  • 4TSXToronto · 09:30–16:00 local
  • 5NYSENew York · 09:30–16:00 local
  • Rings outer → inner · times on the dial in ET (Toronto)

Draft schedule, pending confirmation. International market risks

  1. 18:00ET · Toronto

    Overnight data ingest

    Market and company data arrive, are checked for quality and versioned point-in-time.

    Detail

    Failed checks block downstream jobs; nothing trades on unvalidated data. Risk limits for the Asian session are refreshed.

  2. 09:00JST · Tokyo

    Tokyo open

    Signals for Japanese equities go live for the session.

    Detail

    Orders are scheduled around the lunch break (11:30–12:30 JST) with expected costs measured before submission.

  3. 09:15IST · Mumbai

    Mumbai open

    The Indian session begins.

    Detail

    Position sizes respect liquidity limits specific to the market and its access rules.

  4. 09:30CST · Shanghai

    Shanghai open

    The China session begins.

    Detail

    Trading follows the morning and afternoon sessions, with the midday break observed.

  5. 15:30IST · Mumbai

    Asia close

    Asian positions are reconciled and the hedge overlay is re-checked.

    Detail

    Fills are compared with expected costs; any break triggers a review before the North American session.

  6. 08:00ET · Toronto

    Regime update

    The regime model is updated and the day’s rebalance plan is prepared.

    Detail

    If the estimated probability of a stressed regime crosses its threshold, the hedge ratio and factor weights step to their stressed settings.

  7. 09:30ET · Toronto · New York

    North American open

    Execution in Canadian and US markets.

    Detail

    Orders are worked through the session to limit market impact.

  8. 12:00ET · Toronto

    Live risk monitoring

    Exposures, limits and the hedge are monitored continuously.

    Detail

    Kill-switch procedures are rehearsed: any live process can be halted immediately.

  9. 16:00ET · Toronto · New York

    Close and cost review

    Transaction costs are measured against expectations.

    Detail

    Persistent slippage feeds back into the execution and capacity models.

  10. 21:00ET · Toronto

    Research runs

    Backtests, and machine-learning model training and validation.

    Detail

    Research results never change live models automatically; every change is reviewed by people first.

§07 · PERFORMANCE

Performance, with its context.

Results are presented alongside market indexes for the same period, together with the risk taken. Past performance is no guarantee of future results.

SHIPRA://PERFORMANCE/VS-INDEXESAS OF 2026-10-301
Shipra28.42%NASDAQ-10015.86%S&P 50010.99%NIFTY 509.48%
Shipra versus market indexes — Compound annual growth rate, 20 years: 1 Nov 2006 – 30 Oct 2026, Shipra: proprietary account, time-weighted, no fees deducted · S&P 500: total return, USD · NASDAQ-100: QQQ, dividends reinvested, after fund fees, USD · NIFTY 50: price index, dividends not included, INR
SeriesAnnual returnGrowthMax drawdownBasis
Shipra28.42%×148.510.00%Proprietary account · time-weighted · no fees deducted
NASDAQ-10015.86%×18.970.00%QQQ · dividends reinvested · after fund fees · USD
S&P 50010.99%×8.040.00%Total return · USD
NIFTY 509.48%×6.120.00%Price index · dividends not included · INR
MAX DRAWDOWN
0.0%
MAX DRAWDOWN
0.0%
MAX DRAWDOWN
0.0%
MAX DRAWDOWN
0.0%

Compound annual growth rate, 20 years: 1 Nov 2006 – 30 Oct 2026 · Shipra: proprietary account, time-weighted, no fees deducted · S&P 500: total return, USD · NASDAQ-100: QQQ, dividends reinvested, after fund fees, USD · NIFTY 50: price index, dividends not included, INR · past performance is no guarantee of future results

For illustration and discussion only. Not an offering. Past performance is no guarantee of future results. Full disclosure

§08 · TEAM

35 years of combined experience.

Three founders, one shared discipline. Each portrait is a generative signature seeded from the founder’s name.

§09 · WHO WE SERVE

Who we work with.

Who Shipra works with and how to engage
WhoDescriptionHow to engageNext step
IndividualsInvestors who want to understand a systematic, risk-first approach before committing capital.Start with an introductory conversation about the strategy and its risks.Start a conversation →
HNI / AccreditedHigh-net-worth and accredited investors eligible under applicable prospectus exemptions.Request the confidential offering memorandum and a detailed strategy discussion.Request materials →
Advisors & InstitutionsAdvisors, family offices and institutions seeking access to the QHAM strategy for their clients or portfolios.Discuss access to the same strategy, and the information you need for your own review.Discuss access →

§10 · RESEARCH

Latest research.

View all research →

Education · Nov 5, 2026 · 6 MIN

Howard Marks on the Investing Mistakes That Never Go Away

Howard Marks's 2005 memo "There They Go Again" lists investing mistakes that return every cycle. We sort them into four families, test them on Indian data from the 2008 crash and October 2026, and turn them into seven questions to ask before you buy.

Parth Patel

Market Notes · Nov 5, 2026 · 6 MIN

Nifty 50 Flat for 2.5 Years: Why Long-Term Investors Quit

The Nifty 50 returned 0.20% in the 29 months from May 2024 to October 2026, after rising about 17% and falling about 16% along the way. Here is why roughly 2.5 years without a gain is where long-term investors give up, what quitting cost, and how to plan for it.

Parth Patel

§11 · FAQ

Questions, answered directly.

What is a Quantitative Hedged Adaptive Multifactor strategy?

It is a rules-based approach that builds a portfolio from several investment factors — such as value, momentum, quality and low volatility — chosen by statistical testing (quantitative), adjusts their weights when the market regime changes (adaptive), and keeps a permanent hedge overlay to limit drawdowns (hedged). Shipra calls this QHAM.

How does Shipra hedge downside risk?

A hedge overlay offsets part of the portfolio’s market exposure at all times. When Shipra’s regime model estimates a higher probability of stressed markets, the hedge ratio increases. Hedging has a cost and reduces gains as well as losses; it aims to limit drawdowns, not eliminate them. Investments can lose money.

Which markets does Shipra invest in?

The QHAM strategy invests across five markets: the United States, Canada, India, China and Japan. Investing internationally adds currency, emerging-market, liquidity and regulatory risks, which are described in Shipra’s disclosures.

How does Shipra use machine learning?

Machine learning is used in research and model development, alongside classical statistical methods, to help find and test candidate signals. It does not make autonomous investment decisions: people set the assumptions, review every model change and can halt any live process.

Who can invest with Shipra?

Offerings, where available, are made only by delivery of a confidential offering memorandum to appropriate investors who qualify under applicable prospectus exemptions, such as accredited investors. Nothing on this website is an offer. Contact Shipra to discuss eligibility.

What is factor investing?

Factor investing selects securities based on measurable characteristics — such as cheapness (value), recent strength (momentum), profitability (quality) or price stability (low volatility) — that research has linked to differences in long-run returns. A multifactor portfolio combines several such characteristics.

What is a drawdown?

A drawdown is the decline from a portfolio’s previous peak value to a later low, usually expressed as a percentage. Maximum drawdown is the largest such decline over a period. It measures how much an investor could have lost from a high point before recovering.

How does Shipra avoid overfitting its models?

Every candidate signal is tested on data it was not built on, across different market regimes, using point-in-time data free of look-ahead and survivorship bias. Most ideas are rejected at these stages; only those that remain robust are considered for production.

§12 · CONTACT

Let’s talk.

Tell us a little about yourself and what you would like to discuss. We reply to every inquiry.

OR EMAIL hello@shipra.ca

An offering may be made only by delivery of a confidential offering memorandum to appropriate investors. Full disclosure