How Stocker analyses every stock
No black box. This page is the honest, complete methodology behind the daily research — including the parts that are still experimental.
Same steps, same order, every trading day
| PKT | Step | What happens |
|---|---|---|
| 16:00 | Data collection | Daily OHLCV for the whole listed universe, collected after the PSX close. |
| 16:15 | Index & breadth | KSE-100 index series and market breadth (advances/declines) build the market-regime picture. |
| 16:30 | Indicators & signals | ~20 technical indicators, 18 rule-based signals, and candlestick & chart patterns computed per stock. |
| 16:35 | Composite scoring | Signals combine into the −100…+100 composite score; the daily dashboard feeds are built. |
| 16:40 | ML features | 60+ technical, fundamental and market-context features per stock are refreshed. |
| 16:45 | Regime update | Breadth + index technicals classify the market BULL / BEAR / CHOPPY. |
| 16:50 | ML ranking | The XGBoost ranker scores the universe for the 20-trading-day horizon. |
| 17:00 | Strategies & exits | Seven strategy scanners run for new plays; open plays get their exit checks (stop, target, time, sell-signal). |
Where the data comes from
Public market data only. We cross-check where we can, and we tell you plainly that third-party sources can carry errors or delays.
| Source | Provides |
|---|---|
| PSX Data Portal | Intraday ticks consolidated into daily OHLCV prices, plus company announcements collected three times a day. |
| SCS Trade | ~7 years of historical prices, quarterly fundamentals, ratios and valuations. |
| TradingView | Analyst estimates, trailing ratios, market capitalisation, sector and industry classification. |
| KSE-100 index | The index series behind the market-regime defense. |
The composite score, in the open
Eighteen rule-based signals — each with a published weight — sum into a composite score from −100 to +100. Stocks scoring +25 or higher enter the BUY feed; strongly negative scores flag the SELL side. Every card lists exactly which signals fired and what they contributed.
Candlestick and chart-pattern detection run alongside these signals and feed the same transparent breakdown. Because every component is named and weighted, you can always see — and challenge — the reasoning behind a verdict.
Gates every candidate must clear
Liquidity gate
Stocks below an average daily value-traded floor are filtered out of the dashboard and scanners. A signal you can't actually trade without moving the price is not a signal.
Market-regime gate
KSE-100 trend and market breadth classify each day BULL, BEAR or CHOPPY. New longs are blocked in BEAR regimes and sized down in CHOPPY ones — the go-to-cash discipline most retail books lack.
Reward-to-risk floor
A strategy play only opens with a reward-to-risk of at least 2:1 — target at least twice the distance to the stop. No asymmetric bets.
Portfolio risk layer
Position sizing at 1% equity risk per trade, portfolio-heat warnings when total open risk climbs, and single-sector concentration caps.
An ML ranker that has to earn promotion
Alongside the rule-based score, an XGBoost model ranks the whole liquid universe on 60+ features — price action, technicals, fundamentals, market context and earnings — aiming at one question: which stocks are likely to sit in the top quintile over the next 20 trading days?
Retrained monthly
On a rolling ~3-year window with time-series cross-validation — never trained on the future it will be judged on.
Promotion gate
A candidate model replaces the incumbent only if it wins on both top-quintile precision and rank information. Noise-level improvements never ship.
Shown separately
The ML rank is displayed next to the transparent rule-based score — so you can see when they agree and when they don't.
Seven scanners, one horizon
Each prebuilt scanner targets the weeks-to-months holding window and produces plays with an explicit entry, target, stop and holding period.
| Strategy | Looks for |
|---|---|
| Breakout (trend template) | Stocks near 52-week highs above rising long-term averages, with volume confirmation. |
| Trend-filtered mean reversion | Controlled pullbacks — oversold dips inside established uptrends. |
| Cross-sectional momentum | The market's top performers over 3–6 months, rebalanced periodically. |
| Value-momentum composite | Stocks cheap versus their sector that also carry positive signal scores. |
| Dividend capture | Upcoming cash dividends with a yield floor and signal confirmation. |
| Seasonal patterns | Calendar edges with a consistently high win rate over multiple years. |
| PEAD (earnings drift) | Continuation after earnings surprises, driven by scraped announcement dates. |
Exits are checked daily. Every open play is re-examined after each close against four barriers in order — stop-loss, target, maximum holding time, and an opposing sell signal — with levels resolved against the bar's actual intraday range, not just the close. Notifications land in-app and by email.
Every assumption gets backtested
A full backtesting engine simulates each strategy with commissions, position caps, three sizing modes and five exit reasons — reporting Sharpe ratio, maximum drawdown, profit factor, win rate and R-multiples. The same audits that keep the signals honest keep the strategies honest: we measure, we don't assume.
What Stocker is not (yet)
Honesty about limitations is a feature. Here is the current state, plainly.
The BUY ranking is under active rework. Our own audit found the composite score tilted towards recent momentum — historically the stocks that revert on PSX. The worst inversion is fixed; the ranking is still not where we want it, and we won't pretend otherwise.
End-of-day, not intraday. The pipeline runs after the close. There are no live intraday alerts; stops and targets are levels we surface, not orders we place. Your broker executes; nothing here does.
Data can be wrong or late. Prices, fundamentals and estimates come from public third-party sources. We cross-check and gate, but errors and delays happen — always verify before acting.
This is research, not advice. Stocker is not a SECP-licensed adviser and nothing on the platform is a recommendation to buy or sell anything. Every decision — and its outcome — is yours.