Skip to main content
LEARNAIMARKETS · AI-POWERED TRADING

Educational guide · Competitors

Competitors and baselines: how we compare the Core12 model

The Core12 OOS model is compared daily against 6 classic strategies (S&P 500, Equal-Weight, HRP, Inverse Volatility, Mean-Variance, 60/40 EU) over the same walk-forward period. This guide explains what each baseline does, why it matters, and how to interpret the metrics table and equity curve.
See live comparison What is walk-forward

The 7 strategies in detail

Each baseline is computed on the same Core12 prices (12 UCITS ETFs), with monthly buy-and-hold rebalancing, base €10,000, gross (no transaction costs).

  • Core12 OOS (walk-forward)

    The AI model evaluated only on dates posterior to its training.

    Why it's included

    Single source of truth for realizable returns — no look-ahead.

    Pros

    Alpha vs S&P 500 computed over the exact same OOS period.

    Cons

    Limited OOS period (~12.3 years); modest skill compared against bullish benchmarks.

  • S&P 500 (SXR8.DE UCITS)

    100% capital in iShares Core S&P 500 UCITS ETF.

    Why it's included

    Reference market benchmark for any active US equity strategy.

    Pros

    Liquid, cheap (TER ~0.07%), well-known to every retail investor.

    Cons

    100% USA equity concentration — high volatility and severe drawdowns in bear markets.

  • Equal-Weight (1/N)

    Trivial allocation: 1/12 to each ETF in the Core12 universe.

    Why it's included

    Control with no assumptions. If an active strategy doesn't beat this, it likely adds no value.

    Pros

    Diversified by construction; monthly rebalance keeps balanced exposure.

    Cons

    Ignores correlations, volatilities and expected returns; overweights small assets.

  • Inverse Volatility (simple Risk Parity)

    Weights inversely proportional to each asset's volatility.

    Why it's included

    Classic roboadvisor strategy. Reduces concentrated risk in the most volatile assets.

    Pros

    Stable, low aggregate volatility, good performance in sideways regimes.

    Cons

    Overweights bonds during negative-rate periods (low return); lags in bull markets.

  • Hierarchical Risk Parity (HRP)

    Hierarchical clusters by correlation + inverse-variance allocation.

    Why it's included

    Modern improvement on Risk Parity; avoids ill-conditioned matrix inversion (Lopez de Prado, 2016).

    Pros

    Robust to estimation errors; smooth and diversified weights.

    Cons

    Penalizes absolute return strongly in periods of stagnant bonds.

  • Mean-Variance (Markowitz Max Sharpe)

    Classic optimization maximizing Sharpe over a 252-day rolling window.

    Why it's included

    Markowitz's Modern Portfolio Theory (1952). Academic foundation of quantitative allocation.

    Pros

    Theoretical foundation; can exploit positive expected return opportunities.

    Cons

    Very sensitive to estimates (expected return and covariance); unstable with few observations.

  • 60/40 EU

    60% diversified equity (SXR8 30%, IMEU 15%, XDEV 15%) + 40% EUR bonds (EUNH 25%, IEAC 15%).

    Why it's included

    Classic benchmark for European conservative investor with 5-10 year horizon.

    Pros

    Simple, UCITS-regulated, balanced regions + asset exposure.

    Cons

    Poor simultaneous equity + bonds behavior in 2022 (worst recent year).

How to read the metrics table

  • CAGR

    Compound annualized return. A 10% CAGR means capital grows by 10% on average each year.

  • Sharpe

    Return per unit of risk. ≥1 is good, ≥2 is exceptional. Compare strategies with different volatility.

  • Max Drawdown

    Largest peak-to-trough loss. Indicates the worst transient loss scenario you'd have lived through investing at the worst moment.

  • Volatility

    Annualized standard deviation of daily returns. Higher volatility = more erratic path (not necessarily worse long-term).

Frequently asked questions

Why compare the AI model to deterministic baselines?

To avoid the common trap of evaluating a model only against its own backtest. If a model doesn't beat Equal-Weight (1/N), the added complexity adds no value. Baselines are the honest minimum bar.

What are the key metrics in the table?

CAGR (annualized return), Sharpe (risk-adjusted return), Max Drawdown (largest peak-to-trough loss), Volatility (annualized risk), and Win Rate (% of positive return days). All computed over the same OOS period for fair comparison.

Why doesn't Core12 OOS include more years of history?

The OOS period is limited by walk-forward: the model only predicts dates posterior to each training fold. Currently covers 2013-12 → 2026-04 (~12.3 years, 17 folds). The period extends automatically with each pipeline run.

Do the figures include transaction costs?

Main figures are gross (without costs). Costs vary by broker (DEGIRO, Trade Republic, ING, IBKR) and it's up to each user to apply theirs. The OOS card includes an optional reference with traditional IBKR costs (€2.50 fixed + 5bps slippage per trade).

What does it mean for the model to beat the baseline?

If Core12 OOS shows higher CAGR and higher Sharpe with comparable or lower MaxDD, the model adds value over the baseline. If it only beats CAGR but with higher volatility and drawdowns, alpha disappears when risk-adjusted.

Reference: MSCI index methodology

This guide is educational content. The information does not constitute financial, investment, legal or tax advice. Core12 model signals and metrics are indicative only. Investing involves risk of capital loss. Consult a financial advisor before making investment decisions.

Keep learning

Related guides to deepen your understanding of signals, metrics and validation.