GuluStock and AutoServa · GuluStock and AutoServa: User Guide
The Strategies page in GuluStock lists every trading strategy the platform has researched for Bursa Malaysia, shows which ones are admitted to the live shelf and allowed to produce Buy signals for paper trading, and records what each experiment taught. It is used by clients who want to see why GuluStock is, or is not, publishing Buy signals.
Select Strategies in the left sidebar. The top of the page shows three summary tiles: Trading Now, In Research (not killed, not admitted) and Killed (latest run graded kill), then two tabs: Register and Research.
Trading Now lists the strategies the live scan is allowed to fire in the paper book. Each admitted strategy shows a card with figures such as OOS PF, CPCV p10, Avg Hold, Paper Open, Paper Closed and Realised (RM). If no strategy is admitted, the page says so: the scan still classifies the market, but no Buy can be booked until a strategy clears the full validation ladder. This is the honest state, not a fault.
The ladder a strategy must clear includes cross-validation across many splits (CPCV), reproducibility across market phases, an in-sample floor, a point-in-time stock universe, a locked recent holdout and stress tests. An admitted strategy is sized at full or half size depending on its CPCV grade; a strategy that fails the ladder is not admitted.
The Register tab has one row per strategy family, newest research first.
| Column | What It Means |
|---|---|
| Strategy | The strategy name and version |
| Archetype | The kind of setup it trades |
| TF | The bar size the strategy trades on |
| Runs | How many backtest runs the family has accumulated; a high count means iteration, not quality |
| OOS PF | Out-of-sample profit factor: gross winners divided by gross losers |
| Sharpe | Out-of-sample Sharpe ratio (annualised) |
| OOS Ret | Out-of-sample return as a percentage of starting equity |
| Trades | Number of out-of-sample trades |
| Hold (d) | Average holding time in calendar days |
| Last Run | When the family's most recent concluded run finished |
| Verdict | The engine's walk-forward grade: SHIP, REVIEW or KILL |
| CPCV | That run's cross-validation grade, one rung of the ladder |
| Live Shelf | What the live engine does with the strategy now; this is the authoritative answer. LADDER NONE means the ladder refused it |
The Research tab explains the reasoning: Campaigns (with their status, goal and next action), Dead Ends, ideas that were tested and refuted, with the mechanism and evidence, and Lessons from the newest experiments that concluded something.
No strategy currently clears every rung of the validation ladder. GuluStock publishes no Buy rather than trade an unproven edge.
SHIP passed the backtest gates, REVIEW is parked for more evidence (the normal state of active research), and KILL failed. None of them is admission on its own: only the Live Shelf column says what the scan uses.
The live engine governs on the strictest current grade across the whole family, so a newer run can override an older row.