Joseph Noko (pronounced “nookoo”) is an equity research analyst. He finds the losses the numbers don’t show: in the accounts, in the price, and in risk models that treat a loss like a gain. He publishes independent research for institutional investors worldwide, often on companies that carry no analyst coverage at all: Philippine gold, Hungarian security printing, Estonian construction, Danish regional banking, Nigerian palm oil.

The rule behind the work is simple. Lose 50% and you need 100% to recover. He learnt it before he wrote about it. For ten years he ran a gold mine in Zimbabwe through hyperinflation, and he later managed third-party capital in physical metals trading. His master’s thesis, The Nature of Risk, sets out the theory. Drawing on Knight, Keynes, Bernoulli, Shannon and Kelly, it argues that risk is subjective and that not all risks can be quantified. It also argues that because wealth compounds multiplicatively, loss aversion is rationally obligatory rather than a behavioural bias.

That argument shapes the method. Each thesis starts from the primary filings, not vendor data. Reported earnings are rebuilt into NOPAT, invested capital and ROIC, line by line, under US GAAP and IFRS. A reverse DCF then solves for the growth and returns the current price already assumes, rather than forecasting a future and defending it. Cost of equity rests on downside and expected-shortfall beta, not variance, because variance treats a gain and a loss as the same event. Institutional clients receive the thesis and the model workbook, long and short. Every closed position is published with its realised return, including the losses.

He is ranked in the top 20 (LTM) globally and top 40 all-time on SumZero, among a professional membership in excess of 16,000. Since February 2026, New Constructshas commissioned Long Idea research from him for its institutional clients. New Constructs’ forensic accounting data powers the Bloomberg New Constructs Core Earnings Leaders Index and reaches institutional investors through FactSet, S&P Capital IQ and LSEG’s Refinitiv platform. He was a top-10 forecaster in the Good Judgment Project’s COVID-19 tournament, a contest scored on probabilistic accuracy against realised outcomes. His newsletter, The Mirandolan, and his SumZero research are read by more than 570 analysts and portfolio managers worldwide.

He also builds the systems that do this work. His engine applies an economic-value-added methodology from primary filings to banks, REITs and non-financial companies, under US GAAP and IFRS. Deterministic logic sits in versioned files rather than prompts. Two independent derivations of NOPAT must reconcile exactly, and language models are confined to governed advisory roles. In testing, 99.4% of filing-sourced figures have been machine-checked as present on the exact printed page cited. Running an autonomous agent on production research taught him that the failures that matter are not obvious hallucinations. They are plausible values in the right shape: internally consistent and externally wrong. He serves as a finance expert and AI trainer at micro1, an AI laboratory.

He holds a Master’s in Law and Finance from the Université d’Angers, France, graduating magna cum laude. He is preparing doctoral research on regime-conditional security selection. His 2011 Cato Journal paper, “Dollarization: The Case of Zimbabwe”, has been cited 133 times.