Why this site exists

I'm C. B. Zakarian, and I write SportsDataTutorials. There's no team behind the byline - it's me, my code editor, and a folder full of Python scripts that I actually run before anything goes on the page.

I started this site because of how badly I floundered when I first tried to learn this stuff. I'd find a tutorial, copy the code, and it would die on line three with a version error nobody warned me about - or worse, it would "work" but quietly print numbers that couldn't possibly be right, and I had no way to tell. So the rule here is simple: nothing is hand-waved. Every line of code on every page was run against a real data source, and every chart and table you see is the genuine output of that run. If a result looks weird, it's because the data is weird, not because I faked a clean answer.

That honesty cuts both ways. When the NBA's stats API blocks my build machine - which it does, because it doesn't like requests coming from data centers - I don't paper over it. I tell you exactly what happens, show the fallback I use, and explain how the same script pulls live data when you run it from your own laptop. I'd rather lose a little polish than pretend something works when it doesn't.

How I think about teaching this

I don't believe you learn analytics from a textbook or a 40-hour course. You learn it by shipping small, real things: pull a week of data, make one honest chart, notice three things you didn't expect, then do it again on a question you actually care about. So every tutorial here is built around one finished artifact - a leaderboard, a shot chart, a model - that you'll have working by the end. I try to write the guide I wish someone had handed me: opinions included, the gotchas called out before they bite you, and no pretending the messy parts don't exist.

Where the data comes from

Everything here runs on free, well-respected public sources:

• Baseball – Statcast via pybaseball (Baseball Savant), with caching turned on so I'm a good guest.
• Basketball – the NBA's stats API via nba_api, with Basketball-Reference as a documented fallback.
• Soccer – StatsBomb Open Data (attribution required, and given) and Understat's expected-goals data.
• American football – nflverse play-by-play, read with nfl_data_py.
• Hockey – the NHL's public API at api-web.nhle.com.
• Sample NBA shot data – the public NBA_Shots_04_25 shot-log dataset, which is where the bundled shot CSVs come from.
• Baseball standings and soccer fixtures – the MLB Stats API and ESPN's public scoreboard, which produce the bundled CSVs.

In the example code I cache requests, send polite headers, and keep request rates reasonable - partly so you pick up good habits, and partly because these sources are free and I'd like them to stay that way for the next person.

How a tutorial gets published

Every tutorial starts as a script, not as prose. I write the script, run it against the real source, and the build system captures its printed output and its charts straight from that run into the page — so a printed table here is stdout from a file you can download, and every chart is the image that file wrote. When the run changes, the page changes with it. Nothing is retyped from memory, and no figure gets published that no script produced.

Found a mistake?

Tell me. I'd genuinely rather hear it than not. Email me at contact@sportsdatatutorials.com and I'll fix it.