Glossary

A system that ranks stocks on multiple independent criteria, not one. In my case four: momentum, value, quality and catalyst. The idea comes from academic research into what historically makes stocks outperform.

The stocks the system is allowed to trade. For me 554 names: roughly 500 from the S&P 500 (largest US stocks) plus 54 Dutch stocks from AEX and AMX.

Not all 554 get a score every day. Some lack sufficient data for a reliable composite. On average about 462 stocks actually get ranked.

The US data provider that feeds my financial data: daily prices, quarterly reports, analyst expectations for almost every listed company worldwide. Not free, but affordable.

The online database holding all data: daily prices, quarterly numbers, all computed scores. Like Excel, but with millions of rows and fast querying.

How well or poorly a stock has done recently. Simple idea: stocks that performed well over 12 months often continue to do well. Not always, but decades of academic research show a statistical edge.

Two components: the stock's own 12-month return and its relative strength versus the market. Together they form a momentum z-score per stock.

How much do you pay for what a company produces? Cheap is good, expensive is bad. Measured through three ratios: free cash flow yield, enterprise value over EBITDA, and earnings yield.

Important: Value is computed within the sector, not across the whole market. A chip maker with EV/EBITDA of 30 is normal for Tech. A construction firm at 30 would be ridiculously expensive. Sector-relative avoids apples-to-pears.

How healthy are the fundamentals? Good companies outperform bad ones over the long run. For me: high ROIC, low debt, healthy free cash flow growth, healthy margins.

Quality carries the heaviest weight because fundamentals are more durable than momentum or catalyst. A good company stays a good company even if the price temporarily drops.

Recent positive events that show up in the numbers but not always in the price. Earnings beats, buybacks, positive analyst revisions. Positive today means the market has not yet fully priced the good news.

The most time-sensitive factor: positive today, possibly different in a month. Lighter weight because the signal ages quickly.

A z-score tells you how far above or below the mean you are, in standard deviations. The beauty: it makes factors with wildly different scales comparable.

+1 means one standard deviation above average. +2 is firmly in the top, an outlier. +3 is rare, the extreme. I cap z-scores at +3 to prevent a single extreme stock from dominating the ranking.

The weighted sum of the four factor z-scores. Quality 30%, Value 25%, Momentum 25%, Catalyst 20%. One number that captures how interesting a stock is across all four angles.

Higher is better. Composite +1.0 means the stock is attractive on multiple dimensions. -1.0 is the opposite.

What share of stocks has a lower composite score? Percentile 95 means top 5% of the entire list.

I use percentile rather than absolute scores because it is more robust. Whether the market is high or low, percentile 75 always means: top 25%.

A backtest error where the system uses information not actually available on that historical date. For example: using quarterly numbers on Feb 1 even though they were only published Feb 19. Produces artificially high backtests that cannot be replicated live.

My system avoids this: for fundamentals I use a 60-day buffer (reports usually take a month or two to become truly public), for earnings I use the exact announcement date.

How big an individual position is in the portfolio. For me 4 to 7% per position depending on how volatile the stock is. Volatile = smaller. Calm = larger.

The idea: equal risk per position, not equal capital. A stock moving 5% per day is riskier than one moving 1%, even if both cost the same.

A measure of how much a stock moves on average per day, computed as the average daily range over the last 20 trading days. Higher ATR means more volatile.

I use ATR for two things: position sizing and stop loss calculation (how far below entry the stop sits).

Maximum 25% of the portfolio in one sector. Prevents an Energy sector drop (like 2014 or 2020) from sinking the entire portfolio. Sector spread is a form of risk management.

Maximum 70% of the portfolio in one country. Important because my system sometimes leans heavily Dutch due to universe composition. The country cap enforces spread between NL and US.

Hard exit when the position drops too far below entry. For me: max of -8% below entry or 2x ATR below entry, whichever triggers first.

For most stocks the -8% line wins. For very volatile stocks the ATR-based level wins (because 2x ATR is wider than 8%).

Activates once the position is +15% in profit. From there it follows the highest close at 10% distance. Only ratchets up, never down.

Goal: let a finally winning position run while securing the profit if it reverses. The system makes no profit until the position closes.

Close the position when the composite score drops below percentile 50. The reason to buy was: top 25% of the universe. If the stock falls to the median the thesis no longer holds.

Close the position if it underperforms the market 20 trading days in a row (S&P 500 for US stocks, AEX for Dutch). Not necessarily that the stock falls, but that it lags the broader index.

Close the position after 40 trading days if the return is between -3% and +3%. The stock does nothing yet takes a slot in the portfolio. Better free up the capital.

Close the position when valuation (EV/EBITDA) reaches the sector median from below. The stock was cheap within its sector, now it is market-rate. The value opportunity is gone.

How much return per unit of risk? (return minus risk-free rate) divided by volatility. Higher is better. Below 0 you lose money after inflation. Above 1 is good. Above 2 is exceptional.

Sharpe makes different strategies comparable. 20% return at 30% vol is similarly poor to 10% return at 15% vol (both Sharpe 0.67).

Like Sharpe but only penalizes downside volatility. The idea: upside moves are not "risk", you want them. Only downside spikes hurt.

The largest percentage the portfolio has dropped from a previous peak. Drawdown is the pain you feel in practice: portfolio from €12,000 to €9,500 is -21% drawdown.

Many investors quit at -30% drawdown ("I cannot watch anymore"). My system targets max -30% drawdown, preferably under -20%.

Percentage of trades closed at a profit. Surprisingly not the most important metric. A 40% win-rate system can be profitable if winners are much larger than losers.

My system targets 40 to 55% win rate with asymmetric outcomes: average winner larger than average loser.

Running the system on historical data. What would have happened if I had started 4 years ago with these rules? An estimate of performance, not a promise.

Backtests can mislead through look-ahead bias, survivorship bias and overfitting. A good backtest checks how the system reacts to periods it has not seen before.

A classification of overall market conditions: risk-on (calm), neutral, cautious, or risk-off (panic). Determines how much exposure the system accepts.

In risk-off (like March 2020 or October 2022) exposure drops to 25% and the system aggressively closes positions. In risk-on the portfolio runs at 100%.

An index measuring expected S&P 500 volatility over the next 30 days. Often called the "fear gauge". VIX under 15: calm market. Above 25: nervous. Above 40: panic.

What share of stocks trades above its 200-day moving average. Above 60%: healthy market, broad participation. Below 40%: narrow, possibly trouble under the hood.

The yield gap between risky corporate bonds (high yield, HYG) and safe ones (investment grade, LQD). When the gap widens, bond investors see more risk. An early warning for problems not yet visible in stock prices.

The AI tool that writes the actual code based on my prompts. I describe what is needed, Cursor produces the Python or TypeScript. Then I review and we test together.

My AI assistant for thinking, designing and reviewing. In this project I call him "Harry the helper" for a distinct identity in the collaboration. Writes the prompts for Cursor, helps me decide, flags when something is off.

My workflow: three perspectives on every decision. Me (design and review), Claude (sparring and prompts), Cursor (code). No perspective may be skipped. It forces me to think before code is written.

The backend runs behind the scenes: pulling data from FMP, computing scores, generating signals, sending orders to the broker. Written in Python.

The frontend is what you see on the site: the pages, tables, charts. Written in Next.js, a popular web framework.

The broker where my real trading happens. Worldwide market access, low commissions, programmable. My Python backend sends orders directly to IBKR via their API.