Learn

Week of 16 Oct to 23 Oct 2026: biggest score moves in the S&P 100 and Nifty 50 — The stocks whose scores rose and fell most from 16 Oct to 23 Oct 2026, what drove each change, the market backdrop, and how the live paper portfolios are doing. Educational, not advice.
Can machine learning predict stock returns? We tested 12 models — Twelve approaches, from simple rules to LSTM and GRU neural networks, ranked 99 large US stocks under one strict test. None showed reliable skill. Here is what we did and what it means.
Does momentum investing still work? What our data shows — Momentum, buying recent winners, is one of the best documented effects in finance. In our test on large US stocks from 2020 to 2025, a trend and momentum rule was the best of twelve models, and still not statistically reliable.
Forecasting stock market volatility - HAR, GARCH, machine learning and the VIX compared — Market direction is close to unpredictable, but turbulence is not. We compared a machine learning model with HAR, GARCH, a foundation model and the VIX. The market's own forecast and a 1986 GARCH model were hardest to beat.
Free quant tools and data for learning quant finance — A short guide to free data sources, Python libraries and learning resources for quant beginners, plus the habits that matter more than any tool.
How trading costs change backtest results — A strategy that wins before costs can lose after them. We charge 5 basis points per trade and re-test at double costs. Here is how turnover and costs ate into our simulated portfolios.
Overfitting in backtesting, explained with a real example — Overfitting is when a strategy fits the noise in past data instead of a real pattern. We tried 15 ways to weight our score; the best one looked better on paper and failed when we tested it the right way.
Quant backtesting checklist - 12 checks before you trust a result — A practical checklist for quant backtests, from point-in-time data and costs to trial counts, random controls and a sealed holdout. Each check comes with the mistake it prevents.
Quant glossary - 30 quant finance terms explained simply — Alpha, beta, Sharpe ratio, information coefficient, drawdown, walk-forward, deflated Sharpe, QLIKE and more. Thirty quant terms in plain words, with links to examples.
Quant investing in India - testing a stock score on the Nifty 50 — We applied the same quant method to India's 50 largest companies, using NSE results with exact publication times. The sealed test found no ranking skill. Here is the setup, the data and what it means for quant investing in India.
Survivorship bias in investing, with examples — Testing a strategy only on today's survivors quietly removes every company that shrank, merged or failed. Here is how survivorship bias inflates backtests, and how we flag it on this site.
The information coefficient (IC), explained simply — The information coefficient measures whether a stock ranking lined up with what happened next. Zero means no skill. Here is how it is calculated, what good values look like, and why ours are close to zero.
Walk-forward testing explained — Walk-forward testing trains a model only on the past, tests it on the next block of time, then rolls forward. It is the simplest way to stop the future leaking into a backtest.
What is a sealed holdout, and why we opened ours only once — A sealed holdout is a final stretch of data you lock away until every decision is made, then test exactly once. Ours showed our stock score ranked stocks backwards. We published it anyway.
What is look-ahead bias in backtesting? — Look-ahead bias means a backtest used information that was not available yet. It is the most common way a strategy looks brilliant on paper and fails live. Here is how it sneaks in and how to prevent it.
What is quant research? A beginner's guide with real examples — Quant research means turning an investing idea into a rule, testing it on data without fooling yourself, and checking whether it still works live. Here is how it works, step by step, with examples from our own lab.