Two Ways to Read the Market

Open the Dhaka Stock Exchange website on any trading day and you will find somewhere between seventy and ninety new disclosures waiting. A quarterly result here. A board meeting notice there. A credit rating update. A factory inspection report. A director buying shares. A record date announcement.

Every one of them is written in the language of auditors and lawyers.

If you hold five stocks, perhaps three of those disclosures concern you. Finding which three means reading all eighty. And even when you find yours, the hardest part is still ahead — because a disclosure almost never tells you whether the news is good.

Consider a real sentence from a real filing:

Is that good news?

You cannot answer. Not from that sentence. If the company paid 20% last year, shareholders are about to be disappointed. If it paid 5%, this is a company recovering. If it paid nothing for three years, this might be the most significant announcement of its decade. The number 10% carries no meaning on its own — and the filing will not tell you what came before.

This is the gap Assetick was built to close. We do it two different ways, using two different kinds of information, and it is worth understanding both because they answer genuinely different questions.

Reading what companies actually said

Every trading day, Assetick collects every disclosure the DSE publishes. Not a selection — all of them. We have done this going back two years, which comes to more than thirty thousand filings.

Most of them are routine. "Daily NAV." "Dividend disbursement will commence." "Trading will remain suspended on the record date." These are announcements a company is legally required to make, and they contain no information a reasonable investor would act upon. We keep them, because you may want to see the complete history of a company you own, but we do not analyse them. Roughly six in ten filings fall into this category.

The rest — earnings, dividends, audit findings, production halts, insider transactions, regulatory actions — get read properly.

When we read them, we bring the company's own history along. So when that 10% dividend announcement arrives, the analysis already knows what the company paid before, what its recent earnings look like, and which exchange category it trades in. The judgement is made against that background, not in a vacuum.

The result is a single number between −1.00 and +1.00, and one sentence explaining it.

An example worth understanding

In July 2026, MARICO Bangladesh announced an interim cash dividend of 500% — fifty taka per share on a ten-taka face value. Excellent news by any measure.

But here is where naive analysis goes wrong. MARICO's full-year dividend figure, as published, is a much larger number, because the company pays several interim dividends across a year and they add up. Compare one quarter's declaration against the full year's total and it looks like a catastrophic decline — when in fact nothing bad happened at all.

Getting this right required understanding something specific about how Bangladeshi companies report dividends. That kind of local detail is not something a general-purpose financial tool built for other markets would ever get right. It is the reason we built this ourselves rather than adapting something foreign.

The most important idea on this page

Here is the concept that will make our scores make sense — and without it, some of them will look wrong.

Picture a small textile company. Z category. It has lost money for three consecutive quarters. Its shares have been drifting downward all year. Then it announces: no dividend this year.

Our score for that announcement is close to zero.

At first that seems absurd. No dividend is bad news! But think about it as an investor. Did anyone expect a dividend from a company that has lost money three quarters running? Nobody did. The announcement confirms what the market already assumed. A rational investor reading it changes their view by approximately nothing.

The company's condition is poor. But its condition was already poor yesterday, and the day before, and that was already reflected in the share price. The disclosure adds no new information.

Now picture the opposite. A profitable, well-regarded company with a long record of paying dividends announces: no dividend this year. Same words. Completely different meaning. That score would be strongly negative, because something has genuinely changed.

This is why our neutral label reads "As expected" rather than simply "Neutral". It is not the absence of an opinion. It is a specific finding: this news did not move the needle.

What you will see

On any stock page, the News & Sentiment card shows the strongest signal from the last ninety days, with the date and the reason. Below it, you can expand the full list of disclosures — each with its own score, and routine filings clearly marked as routine.

One detail worth knowing: the DSE sometimes republishes an unchanged notice every day for weeks. A factory closure report might appear again and again. We show it once, with a note that it has been reported repeatedly and when it first appeared. Dozens of identical entries would tell you nothing except that the exchange has a scheduler.

The second system does not read news at all. It reads prices — about 1.5 million daily records covering 533 securities going back to 1999.

And it answers a question that most people find surprising when they first encounter it.

Why "will this stock go up?" is the wrong question

Ask any prediction tool whether a stock will rise, and it faces an impossible problem. Because that question secretly contains two questions:

1. Which way will the whole market move?
2. How will this stock do compared to the market?

The first one is not answerable. When the DSEX drops two percent, almost everything drops with it — the strongest technical setup in the world will not save a stock from a falling market. Anyone claiming to predict the index tomorrow is guessing.

The second question, though, is different. It is genuinely answerable, at least partially. And crucially, it is the question that actually matters when you are choosing between stocks.

So that is the question we ask.

What our model actually predicts

Over the next five trading days, will this stock beat or lag the typical DSE stock by more than two percent?

Three possible answers: it leads, it lags, or it moves in line with everything else.

If the market falls four percent next week and this stock falls one percent, it outperformed. You still lost money — but you lost less than someone holding the average stock. In a rising market, the same signal means you gain more than average.

This sounds like a technicality. It is not. It is the difference between a question that can be answered and one that cannot.

Where the model is strongest

Being straight with you: the model is noticeably better at identifying stocks likely to lag the market than stocks likely to lead it.

That is not a flaw we are apologising for — it reflects something real about the Dhaka exchange. In twenty-seven years of history, there has not been a single year in which the majority of stocks rose on the majority of days. Decline is the more common state. Patterns that precede weakness are simply more visible in the data than patterns that precede strength.

So use it accordingly. This is more valuable as a tool for avoiding capital traps than for finding rockets.

Monday morning, checking your portfolio

You hold six stocks. Over the weekend, four of them had disclosures. Instead of opening six company pages and reading legal prose, you scan six sentiment badges. Five show "As expected". One shows −0.72 with a reason: an audit qualification.

That is where your attention goes. Two minutes, not forty.

Considering a new position

A stock has caught your eye. The chart looks constructive, the price is off its highs, and a friend mentioned it.

Before committing, you check two things. The sentiment card shows nothing significant in ninety days — no bad news lurking. The model says "In line". So: nothing alarming, but nothing suggesting this will distinguish itself from the market either.

That is not a signal to buy or avoid. It is context you did not have five minutes ago, and it lets you make your decision knowing what you are and are not being told.

When something looks too cheap

A share is trading well below where it was six months ago. Tempting.

The sentiment feed shows the reason in seconds: a production halt reported six weeks ago, an unanswered exchange query, a delayed board meeting. None of that was visible on the price chart. All of it was public information you would have had to find yourself across dozens of filings.

This is the case where the tool earns its keep — not by predicting, but by surfacing what was already there and hard to find.

We would rather tell you this plainly than have you discover it later.

It would be easy to write a page promising to predict the market. Plenty of places do.

We think that is a bad trade. A tool that overclaims wins your subscription once and loses your trust permanently. A tool that tells you exactly what it knows, what it does not, and how confident it is — that one you can actually build a process around.

Everything on this page can be checked. The disclosures we analyse are public. The prices we learn from are public. The claims we make about our own accuracy are measured against a baseline anyone could compute, on data the model had never seen.

We would rather be useful than impressive.