Hikma
Research that keeps its own receipts
Hikma is Mizan's research voice: a weekly issue, a per-security outlook, and a librarian you can ask questions. Everything it says forward-looking is written down, fingerprinted, and scored in public when it comes due.
Weekly
Hikma Weekly
One issue, published on a fixed schedule, identical for every reader. Rankings, a compliance watch for ratios drifting toward their limits, and the status changes of the week.
Per security
Hikma Outlook
For any security: the odds its binding screen is still under its limit in a year, what the dividend is worth after purification, and a calibrated range — never a price target.
On demand
Ask Hikma
A research librarian, not an adviser. It explains methodology and what Mizan has published. It can't see your holdings, and it answers the same way for everyone.
The record
Graded in public, including the misses
Anyone can claim a good track record by remembering selectively. The only defence against that is to write the calls down before the outcome is known, and to score them automatically when the date arrives.
How a call becomes a score
Written with a deadline
Every forward-looking statement is stored with the probability Hikma assigned, the security, a baseline measured at publication, and the date it resolves — 4 weeks out.
Fingerprinted before anyone reads it
The issue is hashed with SHA-256 at the moment it's written, and the hash is published. Change a word afterwards and the hash no longer matches.
Scored without a human in the loop
On the resolve date the outcome is compared with the baseline and recorded. Nobody decides whether a call counts, so a bad one can't quietly fail to be counted.
When Hikma says 60%, does it happen 60% of the time?
That is the only question that matters about a forecast, and it's a different question from whether the calls were right. A forecaster who says 60% should be wrong 40% of the time — being right every time would mean the confidence was understated. The closer these two columns sit, the better calibrated the research is.
The record hasn't started yet
31 calls are open and none has come due. Calls are scored 4 weeks after the issue that made them, so the first results land on Oct 11, 2026.
Nothing is hidden here — there is simply nothing to score yet.
Brier score measures forecast accuracy on a scale where lower is better. Always guessing 50% scores 0.250, so anything above that is worse than refusing to have a view.
Verification
Every issue, with its fingerprint
The hash is computed from the issue's contents at publication. It proves the text you're reading is the text that went out.
“A record you have to subscribe to read isn't a record. The scorecard and the fingerprints are public whether you pay or not.”