I love data.
More specifically, I love the kind of data that changes how people understand a game.
That is why I have always had a crush on the sabermetrics community. Baseball fans took a sport built on tradition and intuition and slowly transformed it into a laboratory of probabilities, leverage and decision-making.
My first love in that world was Win Expectancy.
Given the exact state of a baseball game, you can calculate the historical probability that either team eventually wins.
Using tools like FanGraphs or SaberTables, you can plug in a game state:
- inning
- score differential
- outs
- runners on base
…and instantly see how likely each team is to win.
Quick Example: Bottom of the 4th. Tie game. Bases loaded. No outs.
Historically, the home team wins roughly 80% of the time from that position. If the batter strikes out, the probability drops to 73%. If they hit a grand slam, it jumps to 95%.
That single at-bat becomes measurable leverage. You can literally visualize the moment the game swung.
I loved this idea so much that I built WarningTrack.co just to explore it.
Can We Build the Same Thing for Magic: The Gathering?
Baseball has innings. Magic has turns.
Baseball has home and away teams. Magic has being on the play or on the draw.
Baseball has score differential. Magic has life total differential.
In baseball, every out advances the game state. In Magic, every phase does:
- beginning phase
- precombat main
- combat
- postcombat main
- end step
What Data Points Should We Be Tracking?
Some of the options are straightforward:
- turn number
- life totals
- cards in hand
- cards remaining in library
- mana available
- creatures in play
- total power/toughness on board
- mulligans taken
- cards in graveyard
A player with six cards in hand and stable life totals is usually in a much stronger position than a player top-decking with an empty board.
A missed third land drop dramatically changes win probability.
Being on the draw versus the play matters immediately before the first spell is even cast.
The Hard Part: Context Changes Everything
A runner on first base in baseball means roughly the same thing every time.
A 3/3 creature in Magic means almost nothing by itself.
Maybe it is lethal next turn. Maybe it is blanked by a deathtouch blocker. Maybe it has flying. Maybe it is enabling a combo. Maybe the opponent is representing a combat trick with two untapped mana.
That complexity is exactly why a true MTG “win expectancy” engine does not really exist yet.
But it is also why the problem is fascinating.
Because unlike ten years ago, we finally have the data.
Enter 17Lands
17Lands collects enormous amounts of gameplay information from users running their tracker:
- draft logs
- decklists
- game outcomes
- card performance
- opening hand data
- full replay sequences
We can let the data tell us.
Instead of creating arbitrary heuristics, we can train models directly on historical game states and outcomes.
And this is where the project stops being theory and starts becoming data science.
Join next week for part two!
