What is the difference between technical, fundamental and quantitative analysis?

Fundamental analysis values the business from its financials and decides whether it deserves capital. Technical analysis reads price and volume to decide when to commit. Quantitative analysis measures the distribution of possible outcomes — how wide the range is and how likely each part of it is — which is what actually determines position size.

What is a probability band?

Instead of one price target, the model publishes quantiles: P10, P25, median, P75 and P90. An 80% band means the model expects the price to finish inside it four times out of five. Wide bands mean high uncertainty and demand smaller positions.

How is the consensus forecast built?

Four independent models — technical trend, quantitative distribution, fundamental factors and market regime — are pooled in log-odds space, the same aggregation Metaculus uses for its community prediction, and then priced like Polymarket binary contracts in cents.

Is the track record published?

Yes. Every forecast horizon and every market is graded against the official PSX close and scored with mean absolute percentage error, band hit-rate, direction hit-rate and the Brier score.

The PSX trading guide

Four lenses on every Pakistan Stock Exchange listing — technical, fundamental, quantitative probability and consensus. Run the scanner · Forecast board

Market regime: Risk-off 99 up / 350 down

Graded forecasts0resolved against PSX closes

Direction hit-rate—coin flip is 50%

80% band hit-rate—target is 80%

Brier score—lower is better; 0.25 = uninformed

Lens 1 — Technical: when

Price is the only variable that pays you, so technical work is about timing and risk, not prophecy. The engine reads four things on every symbol:

Trend — a blended 20/60/180-session regression, with the R² published so you know how much of the move is actually trend and how much is noise.
Momentum — RSI(14) for stretch, MACD(12/26/9) for the direction of the change in momentum.
Structure — the 20/50/200-session averages, 40-session support and resistance, and the 52-week range.
Volatility — realised daily volatility, which sets stop distance and every band width on the site.

Rule of thumb: a setup without a defined invalidation level is not a setup, it is a hope.

Lens 2 — Fundamental: whether

Fundamentals decide whether a company deserves capital at all. Every listing with filed financials is scored on four axes, each 0–100:

Value — earnings multiple against the PSX median, weighted with dividend yield.
Growth — revenue and profit growth, with negative growth punished harder than flat growth.
Quality — return on equity against balance-sheet leverage.
Risk — debt/equity, multiple expansion risk and dividend coverage.

A cheap company in a broken sector is not value, it is a value trap — which is why the regime lens exists.

Lens 3 — Quantitative: how wide

A single price target is close to useless because it has no error bar. The v4.1 engine publishes a full terminal distribution at 1 day, 1 week, 1 month and 3 months:

Log-normal core — prices cannot go negative, so returns are modelled in log space.
Volatility term structure — 30-session volatility scaled against 90-session volatility, so a calm stock that just woke up is not modelled as if it were still calm.
Fat tails — realised excess kurtosis widens the cone, because PSX moves are not Gaussian.
Quantiles — P10, P25, median, P75, P90 plus explicit probabilities of clearing ±2%, ±5% and ±10%.

Position size comes from band width. Everything else is opinion.

Lens 4 — Consensus: how confident

The four models rarely agree, and their disagreement is information. They are combined with log-odds pooling — Metaculus's community-prediction method — and then priced like Polymarket binary contracts:

Probability, in cents — a 63¢ YES is a 63% modelled chance, nothing more.
Spread from dispersion — when the models fight, the spread widens; a tight spread is a clean read.
Explicit resolution — every market states the exact PSX closing price and date that settles it.
Brier scoring — resolved markets are graded, so calibration is measurable instead of claimed.

Forecasting is a scored sport here. A 70% call that fails is not "wrong" — a 70% call that only lands half the time is.

Multi-lens board

Symbol	Price	Technical	Fundamental	Quant	Composite	P(up 1M)	Signal
NRL National Refinery Limited	546.90	91	28	78	64	65%	Buy
ATRL Attock Refinery Limited	1,131.02	70	28	75	56	55%	Hold
PRL Pakistan Refinery Limited	92.14	65	28	74	54	51%	Hold
PPL Pakistan Petroleum Limited	231.45	81	28	50	52	54%	Hold
CNERGY Cnergyico PK Limited	13.47	55	28	73	50	47%	Hold
OGDC Oil & Gas Development Comp	316.23	32	28	49	35	22%	Sell
GAL Ghandhara Automobiles Limi	545.92	25	28	51	33	21%	Sell
MCB MCB Bank Limited	387.99	28	28	47	33	20%	Sell
HBL Habib Bank Limited	304.12	13	28	55	31	16%	Sell
PSO Pakistan State Oil Company	351.07	26	28	40	30	25%	Sell
KEL K-Electric Limited	6.05	14	28	30	23	19%	Strong Sell
NBP National Bank of Pakistan	165.14	13	28	21	20	16%	Strong Sell

Risk rules that survive contact with the market

Fixed fractional risk. Risk a constant small percentage of equity per idea, computed from the stop distance the model publishes — not from how good the story sounds.
Size from the band, not the target. Two stocks with the same target and different band widths are completely different trades.
Liquidity before edge. PSX traded value below the scanner's floor means your exit is theoretical.
One invalidation per position. Written before entry, honoured without debate.
Score yourself. Log every call with the probability you assigned, then compute your own Brier score. Calibration beats conviction over any real sample.

Educational material and model output only. QW Market Intelligence AI is not a broker, adviser or portfolio manager, and nothing on this page is investment advice.