COGNITIKA
EN
Get in touch

Evidence of Educational Game Design

A familiar claim is made for this field: that games make people learn better. It does not survive contact with the evidence. The honest version is narrower, and considerably more useful: a well-designed game produces measurable learning gains, and a poorly designed one produces very little. The difference is the design, not the medium.

By Ilija Bojović, Founder / Lead game designer · 16 August 2026

This is the hub of a five-part series. It gives the whole argument in about two minutes. Each section below links to a longer piece that goes into the evidence properly, with its own references.

The gist

An ordinary educational game gives about six in ten learners an edge over old-style teaching. A well-designed simulation gives eight or nine in ten an edge. That gap is the entire finding this series is built on, and it is good news: it means the outcome is a matter of decisions a designer actually makes, not a property of the medium.

What the evidence found

Every line below is checked against the meta-analysis it comes from; none of it is an estimate.

What we looked atWhat it means
Educational games vs. ordinary teachingAbout 63% of learners come out ahead
Simulation-based training80 to 85% of learners come out ahead
Motivation, on its ownNo proven boost. Playing a game doesn't reliably make someone more motivated to learn
A structured debrief after the exercisePerformance rises by 20 to 25%
How much better learners do, by teaching method Three bars showing how many learners do better than under the old way of teaching. Educational games in general: about 63% do better. Simulation-based training: about 80% do better overall, rising to about 85% for hands-on technical skill. Educational games, in general Simulation-based training Simulation training, technical skills ~63% ~80% ~85% just over half do better about 2 in 3 do better about 4 in 5 do better
The longer the bar, the bigger the improvement. Ordinary educational games clear a modest one: about six in ten learners do better than under the old method. Simulation-based training clears a much bigger one: eight or more in ten do better, best of all for hands-on technical skill. That gap between the first bar and the other two is what this whole series is about, and it is a gap between designs, not between technologies. These figures are what researchers call effect sizes: 0.33, 0.85 and 1.06. See How large is the effect, really? for what that means.

The seven pillars

Everything the evidence supports reduces to seven design decisions.

The learning is the progression mechanic

What a learner must do to advance is the single highest-leverage decision in the whole design. Argued in Four principles that make a game teach.

Feedback carries the loop

The content of feedback, not its presence, decides whether it works. Same article as above, and connects directly to What happens after the game, since a debrief is feedback at a larger scale.

Difficulty must be desirable, and matched to the learner

Difficulty that slows acquisition can improve retention. Difficulty that defeats the learner helps nobody. Also in Four principles that make a game teach.

Practice must be distributed

Of all common study techniques, only two are reliably effective: testing yourself, and spacing practice over time. Games do both natively. Same article again.

Format follows the task

Neither hardware nor game elements predict whether training works. Task fit does. Set out in full, format by format, in Choosing a format.

Discussion consolidates it

Structured debriefing adds a substantial performance gain on its own. Covered in What happens after the game.

Testing decides, during development

Playtesting is formative evaluation under a different name, and it has to run while the design can still change. The full case is in How to tell whether a game teaches.

What the evidence does not support

Worth stating plainly, because these are the claims that get people caught.

Where this leaves us

Games teach under conditions that can be named, tested, and built for on purpose. None of them follow from the decision to make a game. All of them follow from the design.

References

The full list for this series. Individual articles cite only what they use; this page carries all of it.

Back to articles