Articles

Recording every new token forward, pre-registering every rule, and a real-money post-mortem that blamed the safety stops, not the strategy.

Quant, Research, Crypto, Python, Next.js

An equity research note on Dangote Cement built from NGX filings: a DCF puts fair value at ₦871 against ₦1,490, a price that implies naira bond yields of 12.6%.

Equity research, Valuation, Nigeria, Python

A YouTube Nasdaq breakout strategy lost money from 2010 to 2018, then made $86,510 on sealed 2023–26 data. Neither regime filter would have told you when.

Quant, Research, Backtesting, Trading

A scoreboard of 161 pre-registered trading ideas, from chart patterns to risk premia. Lore never came close; textbook premia barely did better.

Quant, Research, Backtesting, Trading

I built a credit scorecard and a LightGBM PD model on 620,783 Lending Club loans and validated them like a bank: ranking held, but PSI missed the drift.

Credit risk, Python, Machine learning, Data

Day trading on stocks lost to costs, so I tested the best-known intraday pattern on stock indices at futures costs. It worked for a decade, then reversed.

Quant, Research, Backtesting, Trading

Ten macro regime timers passed my backtest gauntlet; buy-and-hold failed only on a technicality. Out of sample, all ten made less than staying invested.

Quant, Research, Backtesting, Python

I coded six Smart Money Concepts (ICT) rules, from liquidity sweeps to order blocks, and backtested them on FX and gold back to 2005. None passed.

Quant, Research, Backtesting, Trading

A pre-registered five-gate backtest gauntlet, a one-shot holdout, and why passing every test still wasn't enough.

Quant, Research, Backtesting, Python

Naira and Eurobond yield curves built from FMDQ, DMO and CBN data. FGN bond yields now move before MPC rate decisions; in 2010–17, in the decision week.

Fixed income, Python, Nigeria, Data