Quantitative Finance · IDX · 2026

SahamQuant

A quantitative trading system for Indonesian equities (IDX/BEI).

◆ SahamQuant Strategi & HasilSinyalLive (simulasi) IHSG −6.47% SIM

Core + Satellite · equity curve (OOS)+20.76%

OOS Sharpe 1.349

StrategyIHSG benchmarkwalk-forward · long-only · T+2

Alpha+12.96%

Sharpe0.289

60/40

Core · IDX4 factor 60%
Satellite · mean-rev 40%

StrategyRet

IDX4 Factor+18.2%

Mean Reversion+9.4%

Foreign-flow tilt+6.1%

EOD signals · IDX80 universe · reviewed before order5 / 80 actionable

TickerSignalScoreFactor

BBCA BUY 0.91 value + foreign-flow

TLKM BUY 0.84 quality + reversal

ANTM BUY 0.78 momentum (satellite)

ASII HOLD 0.52 size neutral

UNVR SELL 0.23 profitability decay

Unrealized P&L · paper+Rp 4.82 jt

Day+1.4%

Positions4

execution stays manual · no retail API

TickerLotEntryLastP&L

BBCA20 lot9 6909 875 +1.9%

TLKM50 lot3 0453 120 +2.5%

BBRI30 lot4 4604 410 −1.1%

ANTM40 lot1 5201 565 +3.0%

BBCA 9 875 +1.8%TLKM 3 120 +2.4%ASII 4 980 −0.6%BBRI 4 410 +1.1%BMRI 6 050 +0.9%UNVR 2 230 −1.2%ANTM 1 565 +3.1%GOTO 62 −1.6%BBCA 9 875 +1.8%TLKM 3 120 +2.4%ASII 4 980 −0.6%BBRI 4 410 +1.1%BMRI 6 050 +0.9%UNVR 2 230 −1.2%ANTM 1 565 +3.1%GOTO 62 −1.6%

Illustrative product mockup · representative data, no sensitive content

+20.76% Annual return (IHSG −6.47%)

+12.96% Alpha

1.349 Walk-forward OOS Sharpe

8 Interactive UI screens

Backtested / walk-forward out-of-sample result · not investment advice, not a live track record.

The problem

Indonesian equities lack clean retail tooling: fragile free data, no sanctioned order API, and most factors (momentum, low-vol) are NOT robust at the index level. What actually has alpha?

The approach

◆ Core+Satellite strategy: 60% IDX4 factor model (value/size/quality/profitability) + 40% mean reversion.
◆ Backtested in vectorbt / zipline with IDX microstructure baked in — lots, tick tiers, ARA/ARB, T+2, taxes.
◆ Walk-forward out-of-sample validation to guard against overfitting; foreign-flow ("bandarmology") tilt.
◆ FastAPI dashboard + EOD ingest pipeline generate signals reviewed manually before any order.

What it proves

✓ Robust walk-forward result — not a single-split overfit
✓ Market mechanics fact-checked (post-2025 ARB / circuit-breaker values)
✓ Honest scope: research & signals in Python, execution stays semi-manual by law

Next project

BC 2.5 Generator

→