Apex Alpha

Markets move.
We follow the evidence.

Apex Alpha is a research-driven systematic investment firm. Our lab studies what drives returns, how those relationships evolve, and what the evidence supports as markets change.

How research becomes a position.

Data

Features

Signal

Ranking

AI

Portfolio

Every trading day, the chain runs again

Workflow

Research-backed, fully automated workflow.

  1. Data acquisition

    Prices, filings, news and positioning are acquired, cleaned and validated for correctness, and stamped with the moment each became public.

  2. Models score

    Features are built from the validated data, and every model in the ensemble scores the full universe of US equities.

  3. Book is built

    Scores combine into one ranking. Positions are sized to liquidity and held inside fixed risk limits.

  4. Desks review

    Each proposed trade passes review at every desk: research, risk, portfolio, compliance and trading.

  5. Orders execute

    Approved orders are executed. Fills, costs and exposure are recorded and fed back into research.

  1. 01 · Clean data

    Survivorship-free, point-in-time, no leakage.

    Every figure is stamped with the day it became public. Companies that went bankrupt or were delisted stay in the history, bad prints are repaired, and nothing a model sees could not have been known at the time.

    Survivorship-freePoint-in-timeNo look-ahead

  2. 02 · Feature tuning

    Plateaus, not peaks.

    Thousands of candidate features are built from the data. Each one is tuned for broad plateaus, settings where performance holds when the inputs shift, rather than narrow peaks that only fit the past.

    Thousands of candidatesPlateau searchStable across years

  3. 03 · Models

    Diverse models, trained walking forward.

    Different families of models learn from the features. Each is retrained as markets evolve and judged only on data it has never seen, with a gap between training and testing so nothing bleeds through.

    Model familiesWalk-forwardPurged gaps

  4. 04 · Regimes

    Markov models read the market’s state.

    Regime-state Markov models estimate whether the market is calm, shifting or under stress, and how likely it is to change. Risk is set by the state the market is in, not by yesterday’s return.

    Markov modelsRegime probabilitiesRisk scaling

  5. 05 · Portfolio

    Long the best, short the weakest, hedged.

    The strongest names are held long and the weakest short, each sized to what it can trade, with market hedges and limits on positions and exposure fixed in advance.

    Long and shortHedgingFixed limits

  6. 06 · Backtest

    Every cost, every day, out of sample.

    Results are simulated as if traded: orders filled at the next open, then spreads, market impact, slippage, commissions, borrow and fund fees, all deducted. Trades are capped by each stock’s liquidity, and only unseen days count.

    Next-open fillsSlippage and impactFees and borrow

  7. 07 · AI review

    A final check before anything trades.

    Every position passes through AI review at each desk in turn: research, risk, portfolio, compliance and trading. Each desk validates and verifies it against the filings, the news and our limits before an order is executed.

    Every deskValidationBefore execution

Our mission

Our goal is to outperform the market with less risk, year after year, by holding every decision to an institutional standard of evidence.

  • Durable over impressive

    A strategy that holds up across many markets beats one that shines in a single backtest.

  • Evidence over stories

    Every idea earns its place on data it has never seen, or it is discarded. No exceptions for favourites.

  • AI at every desk

    Every position is validated by AI at each desk, from research to risk to trading, before an order is executed.

  • Never finished

    Alpha decays, so the research never stops. Our models adapt as conditions shift, and the portfolio is rebuilt and executed every trading day.