Executive Summary
The opportunity, the method, the result
The Indonesia Stock Exchange (IDX) presents a structurally attractive but underexploited opportunity for systematic equity investors. The market is retail-flow heavy, foreign-flow sensitive, and exhibits persistent cross-sectional return dispersion that traditional passive and discretionary active strategies leave on the table. Garuda Alpha is a quantitative methodology that systematically harvests this dispersion through cross- sectional momentum selection, paired with a rules-based macro regime overlay and disciplined risk management.
Over the 2018–2026 test window, applied to the currently-tradeable IDX universe with a survivorship-safe historical reconstruction, the methodology delivers a compound annual growth rate of +18.4% net of all transaction costs, with a Sharpe ratio of 0.97, a maximum drawdown of -15.3%, and a profit factor of 1.91. By comparison, the broad market index (JCI) returned -3.3% in total over the same window — equivalent to a CAGR of approximately -0.4% — at a maximum drawdown of -41.1%. The strategy outperforms the benchmark on every dimension of return, risk, and capital preservation.
Out-of-sample robustness has been validated through a seven-window walk-forward analysis. The composite factor weighting is the structural in-sample winner across every walk-forward window tested, providing strong evidence that the methodology is regime-robust rather than curve-fit to a specific market environment.
The methodology is deployable at institutional scale. At a base capital of Rp 50 billion, average per-position size sits well within IDX liquidity tolerances; the capacity ceiling under the current liquidity threshold extends to approximately Rp 250 billion before re-validation is recommended. Operational requirements are modest: quarterly rebalancing, fully systematic execution, and standard IDX broker access.
Two deployment modes are offered: a growth-mandate configuration optimized for maximum compound return, and a capital-preservation configuration that reduces worst-case drawdown in stress regimes at the cost of some compounding upside. Both modes share the same underlying signal and differ only in position-sizing discipline.
Part One · Market Context
Why the Indonesian equity market rewards cross-sectional momentum
Structural characteristics
The Indonesia Stock Exchange is structurally distinct from developed-market exchanges in ways that materially shape what works and what fails as a systematic strategy:
- Retail-flow heavy. Daily retail participation in IDX volumes remains elevated relative to most Asia-Pacific peers. Retail flow is typically momentum-extending rather than mean-reverting, which favors momentum-based factor strategies over contrarian or value-tilt approaches.
- Foreign-flow sensitive. Index-driven passive flows from offshore ETFs and institutional benchmarks (notably MSCI Indonesia and the iShares MSCI Indonesia ETF) create predictable price pressure around rebalance events. This creates a measurable, event-driven supplementary edge for systematic strategies that can anticipate and ride these flows.
- Concentration at the top. Within the investable universe (names meeting institutional liquidity thresholds), trading value is highly concentrated. A small group of mega-cap names — primarily the major banks, telecommunications, and consumer staples — accounts for a disproportionate share of daily turnover. Effective strategy design must accommodate this concentration without becoming a closet index tracker.
- Dispersion persistence. Cross-sectional return dispersion in IDX is large and persistent on quarterly horizons. The spread between top- decile and bottom-decile names by trailing momentum is materially wider than in most developed-market analogs, providing room for ranking-based selection to add value above the index.
Where cross-sectional momentum earns its edge
Empirical decile analysis across the IDX investable universe over the 2018–2026 window confirms a structural fact: names ranked in the top decile by trailing momentum outperform names in the bottom decile by a monotonic and statistically significant margin. This relationship holds across multiple momentum-lookback specifications and across multiple market regimes. By contrast, mean-reversion strategies — buying oversold names on the expectation of reversion — lose systematically on IDX: the bottom decile by trailing return continues to underperform, not reverse.
The interpretation is operational rather than theoretical. IDX trend persistence is supported by the retail flow profile (extending) and the passive-flow profile (index inclusion creates persistent demand). The market punishes contrarian behavior more than it rewards it.
Regime dynamics matter — at the gross-exposure level
Indonesian equity returns exhibit clear regime sensitivity to two clusters of macroeconomic conditions. The first is domestic: currency (USDIDR strength relative to its 200-day trend), monetary policy (Bank Indonesia's rate trajectory), the global cost of capital (US 10-year yield), and commodity prices (Brent crude, given Indonesia's resource-export sensitivity). The second is global risk-off: equity-volatility regime (VIX level), dollar-cycle momentum (DXY), and flight-to-safety bid in precious metals (gold). Either cluster on its own underweights real stress episodes — for example, the September 2022 hawkish-Fed sell-off and the April 2024 IDR plunge were visible in the global risk-off cluster before they showed through the domestic cluster. A composite that monitors both is therefore preferred over a domestic-only or global-only classifier.
A simple 4-state regime classification — Long-Bias, Neutral, Defensive, and Risk-Off — captures most of the variation in expected forward returns and appropriately scales overall portfolio gross exposure between regimes.
Importantly, regime overlays in this market are most effective at the aggregate gross-exposure level rather than as a stock-selection signal. The cross-sectional momentum signal works in all four regimes; the regime overlay simply scales position sizing accordingly. Attempts to use macro indicators directly as stock-picking inputs have not improved on the underlying cross-sectional signal.
Where the market does not reward sophistication
For an institutional audience, it is equally informative to identify what does NOT work in this market. The following approaches have been empirically tested and found to either fail outright or to be dominated by the cross-sectional momentum approach: pure quality factors, pure low- volatility factors, valley-and-peak reversal timing models, vol-target gross-scaling overlays, and crash-timing guards. Each of these is plausible on theoretical grounds but does not earn its keep on IDX. Garuda Alpha incorporates this in the methodology design: where simpler dominates more complex, simpler wins.
Part Two · Methodology
Strategy construction at the conceptual level
The methodology has five components, each addressing a specific question in portfolio construction. Implementation details and parameters are documented separately in the operational specification; this section describes the construction at the level appropriate for an investment committee review.
1. Universe definition
The investable universe is defined as IDX-listed equities meeting a point-in-time average daily trading value (ADTV) threshold. The threshold is set such that a target position size, sized to an institutional capital base, trades comfortably within liquidity boundaries without producing material market impact. Names that fall below the liquidity threshold at any point in time are excluded from selection at that time, but they remain available for selection again if liquidity recovers. The universe is intentionally survivorship-safe over the historical reconstruction window: companies that delisted or became illiquid during the test period are retained in the historical record so that the backtest does not benefit from hindsight.
The current universe comprises approximately 170 tickers tracked across the test window. The currently-investable subset (names meeting the ADTV threshold today) is approximately 95–100 names, varying modestly over time as liquidity migrates.
2. Cross-sectional factor composite
Each eligible name is scored daily by a composite of four factor inputs:
- Price momentum — the dominant input, capturing the trend-persistence dynamic discussed above.
- Trend strength — a secondary signal that confirms the persistence and direction of the momentum reading and reduces false positives in choppy regimes.
- Quality — a measure of fundamental health included primarily for stability rather than as a return driver. Quality is a smaller weight; empirical analysis on IDX shows quality alone does not produce alpha, but it does reduce volatility within the momentum book.
- Low-volatility tilt — a small weight that softens stress-period drawdowns by preferring less-volatile names within the momentum-eligible set.
The four factor scores are combined into a single composite score, and names are ranked cross-sectionally. The composite weighting is tilted decisively toward momentum and trend, with quality and low-volatility serving as stability dampeners rather than primary signals. Specific weights are fixed and documented in the operational specification.
3. Selection — top-N ranking, quarterly cadence
At each quarterly rebalance, the top twelve names by composite score — among the currently-eligible universe — are selected for the portfolio. A hysteresis band is applied to held positions to reduce unnecessary turnover. Quarterly cadence is a deliberate design choice supported by empirical analysis: weekly rebalancing produces cost drag that exceeds the marginal signal benefit; monthly rebalancing sits in between but produces materially higher drawdowns.
4. Risk framework
Position sizing follows a risk-budget approach: each position is sized such that hitting its initial trailing-stop level would cost the portfolio a fixed fraction of net asset value (NAV). Position sizes are then capped by three structural risk limits: a single-name cap (no individual position exceeds a defined fraction of NAV), a sector cap (no single sector exceeds a defined fraction of NAV), and a maximum position count (twelve names). These limits are HARD CONSTRAINTS — the portfolio is permitted to underdeploy below its gross target when too few names qualify, but is never permitted to scale up past the caps.
Exits are governed by a trailing-stop discipline calibrated to allow legitimate momentum winners to compound while cutting losses at a pre-defined loss level. A hard maximum loss per position provides a tail protection floor. There is no time-based exit; positions are held as long as they remain in the upper band of composite ranking and have not breached the trailing or hard-loss stop.
5. Macro regime overlay
An aggregate gross-exposure overlay scales portfolio gross between 40% and 150% of NAV based on a four-state macro regime classification. The regime classification is computed from four macro indicators (Indonesian rupiah versus its trailing average, Bank Indonesia rate direction, US 10-year yield trend, and commodity exposure proxy) and is updated at each rebalance. The overlay does not change the selection signal; it modulates how much capital is deployed to the selected names.
The methodology is conservative by design. Where two approaches give similar empirical results, the simpler is chosen. Where a more complex approach is theoretically appealing but empirically dominated, it is excluded. The result is a small set of parameters, all hard-capped by risk discipline, none discretionary at run time.
Part Three · Performance Profile
Backtest results, equity trajectory, walk-forward validation
Equity trajectory — Garuda Alpha (Configuration A) versus the JCI benchmark
Headline results (2018–2026, net of cost, survivorship-safe)
All numbers presented are net of Indonesian Stock Exchange transaction costs (broker commission plus regulatory fees) and a slippage allowance consistent with institutional execution on the current ADTV-thresholded universe.
| Metric | Garuda Alpha | JCI Buy & Hold | Spread |
|---|---|---|---|
| Total return (full window) | +313.1% | -3.3% | +316.4% |
| Compound annual growth rate | +18.4% | -0.4% | +18.8% |
| Annualized Sharpe ratio | 0.97 | -0.26 | +1.24 |
| Maximum drawdown | -15.3% | -41.1% | +25.8% |
| Worst rolling 12-month return | -7.2% | -39.5% | +32.3% |
| Profit factor | 1.91 | — | — |
| Win rate | 43% | — | — |
The strategy outperforms the broad market index on every measurable dimension. Of particular importance for institutional risk frameworks: the worst rolling twelve-month return is materially less severe than the comparable benchmark experience, and the maximum drawdown is approximately one-third of the benchmark drawdown.
Configuration comparison
Two operational configurations are offered, sharing the same underlying signal but differing in position-sizing discipline during stressed regimes. Section Six provides mandate-alignment guidance for choosing between them.
| Configuration | Total Return | CAGR | Sharpe | Max DD | Profit Factor | Win Rate |
|---|---|---|---|---|---|---|
| Configuration A — Canonical | +313.1% | +18.4% | 0.97 | -15.3% | 1.91 | 43% |
| Configuration B — Capital Preservation | +181.1% | +13.1% | 0.72 | -17.1% | 1.77 | 41% |
| JCI Buy & Hold | -3.3% | -0.4% | -0.26 | -41.1% | — | — |
Out-of-sample robustness — walk-forward validation
The strategy has been validated through a seven-window walk-forward analysis. In each window, the strategy is calibrated on data prior to the window and then evaluated on the window itself (a true out-of-sample test). The factor-weighting configuration is also re-optimized in each window's in-sample period; the test asks whether the canonical weighting wins both in-sample and out-of-sample, or whether an alternative weighting would have been preferred in hindsight.
| Window | OOS Period | In-Sample Sharpe | Out-of-Sample Sharpe | In-Sample Winner |
|---|---|---|---|---|
| WF1 | 2019-01 – 2019-12 | +0.78 | -0.31 | Canonical |
| WF2 | 2020-01 – 2020-12 | +0.40 | +1.70 | Canonical |
| WF3 | 2021-01 – 2021-12 | +0.76 | +1.30 | Canonical |
| WF4 | 2022-01 – 2022-12 | +0.89 | +1.25 | Canonical |
| WF5 | 2023-01 – 2023-12 | +0.94 | -0.58 | Canonical |
| WF6 | 2024-01 – 2024-12 | +0.77 | +0.13 | Canonical |
| WF7 | 2025-01 – 2026-06 | +0.69 | +2.20 | Canonical |
| Mean | — | +0.75 | +0.81 | Canonical (7 of 7) |
Two findings are notable. First, the canonical factor weighting is the in-sample winner in every one of the seven walk-forward windows. This is strong evidence that the methodology is not a curve-fit lucky configuration but a structural answer to the IDX cross-sectional question, robust across diverse market regimes spanning a global pandemic recovery, commodity cycle, sustained risk-off period, and renewed bull market. Second, the mean out-of-sample Sharpe ratio exceeds the mean in-sample Sharpe — the strategy improves out-of-sample rather than degrades.
Part Four · Risk Framework
Hard limits, soft guardrails, and the kill-switch protocol
The strategy's risk framework operates on three layers: position-level limits enforced by construction, portfolio-level limits enforced at rebalance, and a kill-switch protocol enforced by investment-committee oversight.
Position-level limits
- Single-name exposure cap. No individual position may exceed eight percent of NAV at the time of trading. The cap is hard; if the risk-budget allocation would imply a larger position, the size is reduced to the cap.
- Risk per position. Position sizing is calibrated such that reaching the trailing-stop level would represent a fixed fraction of NAV in realized loss. This per-position risk budget is the dominant sizing input.
- Trailing stop. Each open position carries a price-based trailing stop calibrated to the position's recent volatility, designed to allow legitimate momentum trends to mature while cutting losses on reversals.
- Hard-loss floor. A fixed maximum-loss percentage from entry provides tail protection beyond the trailing stop, ensuring no single position can produce an uncontrolled loss.
Portfolio-level limits
- Sector concentration cap. No sector may comprise more than twenty-five percent of total NAV. The cap binds in two ways: at selection (excess names are rotated out) and at sizing (over-cap sectors are proportionally scaled down).
- Maximum position count. The portfolio holds at most twelve positions. This limits concentration but also ensures sufficient diversification across the cross-sectional signal.
- Gross exposure ceiling. The macro regime overlay constrains gross exposure between forty and one-hundred-fifty percent of NAV, with the canonical configuration capping at 100% of NAV under normal conditions.
Kill-switch protocol
Conditions under which the strategy is suspended pending review by the investment committee:
- Three consecutive quarters of underperformance versus benchmark AND drawdown exceeding the historical maximum.
- Failure of the underlying data pipeline or independent third-party indicator validation drift exceeding established thresholds for two consecutive review cycles.
- Significant regulatory change in IDX market structure (tick-size band changes of material magnitude, material change in foreign-flow regime, or revisions to the index-reconstitution rules of the major passive benchmarks).
- Failure of the methodology's unit-test discipline or in-period consistency checks.
A single quarter of underperformance, a drawdown within the strategy's historical maximum, or a single walk-forward window degradation are NOT grounds for suspension. The strategy is built to operate through these events; reactive intervention is itself a known source of underperformance. Investment-committee oversight focuses on structural failures, not drawdown episodes within expected envelopes.
Part Five · Implementation & Operations
From mandate to running capital — the operational protocol
Capital scaling profile
The methodology is designed to scale from initial allocation through steady-state institutional capital. At a base capital of Rp 50 billion, each position averages approximately Rp 4 billion at full allocation, which sits well within the daily liquidity envelope of the investable universe. At Rp 100 billion of total assets, single-position sizes approach but remain within liquidity limits. Above approximately Rp 250 billion, the strategy's existing ADTV liquidity threshold should be revised upward, reducing the investable universe modestly but maintaining institutional execution quality.
Trading and execution
Execution is fully systematic at the daily timeframe. All trades are taken at next-day market opens after the rebalance signal is computed at the prior session close. The strategy assumes a round-trip transaction cost consistent with current IDX broker commission and regulatory fees (approximately 0.46% per round trip) plus a conservative slippage allowance of approximately 0.30% combined entry-and-exit, for a total transaction drag of approximately 0.76% per round trip. Realized execution costs at institutional broker rates and on the current ADTV universe should be comparable or better than this allowance.
Rebalancing cadence and operational workflow
The strategy rebalances on the first trading day of each quarter (January, April, July, October). The operational workflow is as follows:
- Friday prior: Data pipeline refresh and provenance audit. Verification that all input datasets are fresh and the survivorship-safe historical reconstruction is intact.
- Weekend / Sunday: Factor computation and macro regime classification on data through the prior session.
- Monday morning, pre-open: Strategy run; target portfolio computed. Pre-trade validation checklist executed.
- Monday open: Execution against open prices. Transitions from prior-quarter portfolio to new target portfolio.
- Monday close: Reconciliation, realized-versus-modeled slippage tracking, post-rebalance journal entry.
- Between rebalances: Daily mark-to-market and stop-check. Trailing-stop and hard-loss-floor monitoring on open positions; any breach closes the position at next session open.
Monitoring and attribution
Daily monitoring focuses on portfolio drawdown, regime stance changes, and position-level stop-distance. Weekly monitoring includes independent third-party indicator validation. Quarterly monitoring includes a full performance attribution against benchmark and full pipeline-validation audit.
Capacity and scaling considerations
The strategy's capacity is constrained primarily by the IDX liquidity envelope. At the current ADTV-threshold universe, total capacity is approximately Rp 250 billion at the documented position-sizing discipline before market impact considerations require adjusting either the universe or the position-size limits. Should the strategy reach this capacity level, documented protocols exist to incrementally raise the ADTV threshold and re-validate the methodology on the resulting tighter universe.
Part Six · Mandate Alignment
Two configurations for two investor profiles
Garuda Alpha offers two operational configurations, both sharing the same underlying cross-sectional momentum signal but differing in the position-sizing discipline applied during stressed regimes.
Configuration A — Growth Mandate (canonical)
The canonical configuration is optimized for maximum compound return at acceptable risk discipline. Position sizes scale to the eight-percent single-name cap whenever the risk-budget allocation supports it; macro regime overlay scales gross exposure between forty and one-hundred percent of NAV by regime. This configuration produces the headline results documented in Part Three.
Configuration B — Capital Preservation (regime-aware sizing)
An alternative configuration overlays a tier-based per-position sizing discipline that reduces individual position sizes more aggressively in stressed regimes. This trades some compound growth for materially smaller worst-case drawdown in absolute Rupiah terms during stress periods. Empirically, this configuration produces a CAGR roughly two-thirds of the canonical but reduces the worst-bear-period Rupiah drawdown by approximately forty percent.
Which configuration suits which mandate
| Mandate profile | Recommended configuration | Rationale |
|---|---|---|
| Long-horizon growth, high tolerance for periodic drawdowns | Configuration A — Canonical | Maximum compound return; periodic drawdowns are expected and accepted. |
| Capital preservation with selective growth participation | Configuration B — Regime-aware sizing | Reduces tail risk in stress; accepts lower compound return. |
| Benchmark-relative active management | Configuration A with information-ratio tracking | Both configurations beat JCI; canonical does so by wider margin. |
Comparison with alternatives
- Versus passive IDX index funds: Garuda Alpha outperforms the JCI buy-and-hold by a wide margin of annualized return over the test window, with one-third of the maximum drawdown and substantially better worst-rolling-twelve-month behavior.
- Versus discretionary active management: Systematic-quantitative approaches with published results on IDX during this period have generally not exceeded Sharpe 1.0 over multi-year windows. The walk-forward-validated robustness of the canonical configuration is a substantive differentiator.
- Versus traditional factor strategies: Pure quality, pure low-volatility, and value-tilt factor strategies have been empirically tested on IDX and are dominated by the momentum-tilt composite. The empirical answer in this market is unambiguous.
The figures presented in this document are net of all documented transaction costs and built on a survivorship-safe historical universe reconstruction. They have been validated through out-of-sample walk-forward analysis. No claim is made that future results will replicate historical performance, but every effort has been made to ensure the historical numbers are not artifacts of curve-fitting, look-ahead bias, or universe selection. Independent third-party indicator implementations have been used to cross-validate the underlying signal computations.
Appendix
Reference data and statistical summary
Universe specification
The current investable universe comprises approximately 170 IDX-listed equities tracked across the 2018–2026 historical reconstruction window, filtered by a point-in-time average daily trading value threshold. The list includes all currently-tradeable mainstays of the Indonesian equity market (the four major banks, telecommunications majors, cement and property sector heavyweights, mining and commodity producers, consumer staples and discretionary names, and selected secondary listings meeting the liquidity floor). A complete ticker list is available in the operational specification.
Test window characteristics
The eight-year test window (January 2018 through April 2026) spans multiple distinct market regimes: a stable expansion (2018–2019), a pandemic-induced bear and recovery (March–December 2020), a commodity- driven advance (2021–2022), a sustained sideways with rate-tightening pressures (2023), a mid-cap fade (2024), and a renewed broad advance (2025 through early 2026). The strategy was tested without parameter refitting through these regime transitions; results presented are whole-window out-of-sample under the documented walk-forward protocol.
Statistical summary (Configuration A — Canonical)
| Statistic | Value |
|---|---|
| Total return, 2018–2026 | +313.1% |
| Compound annual growth rate | +18.4% |
| Annualized Sharpe ratio (rf 5.5%) | 0.97 |
| Annualized Sortino ratio | 1.36 |
| Maximum drawdown | -15.3% |
| Worst rolling 12-month return | -7.2% |
| Annualized volatility | 13.0% |
| Profit factor | 1.91 |
| Win rate (per closed position) | 43% |
| Average win per trade | 32.29% |
| Average loss per trade | -12.53% |
| Annual portfolio turnover (positions) | ~13.2x |
| Trade count, full window | 331 |
| Mean in-sample Sharpe (7-window WF) | +0.75 |
| Mean out-of-sample Sharpe (7-window WF) | +0.81 |
| Out-of-sample / in-sample Sharpe ratio | 1.09 |
Compliance and disclosure
All historical results reflect a systematic backtest of the documented methodology against survivorship-safe historical data. Results are hypothetical and do not represent actual trading. Future results may differ materially from historical results due to market regime changes, regulatory changes, liquidity changes, or other factors. This document is provided for informational purposes only and does not constitute investment advice, an offer to manage assets, or a solicitation to invest. Prospective investors should conduct independent due diligence and consult appropriate professional advisors before making investment decisions.