Can You Predict a Hit?
Visualizing successful trends to predict game success
Thousands of games ship every year, but most disappear within weeks. A small subset earn critical attention, build communities, and stay profitable long after launch. I intened to predict these allstars by observing the patterns. This dashboard pulls five years of release data across genres, mechanics, and reviews, scores every game against a consistent rubric, and lets you explore what separates the breakouts from the noise. At the end, you can test your own concept against the prediction model.
Results
The classifed panel below summarises the strongest signals in the past five years of game releases:
- the genre mix that tends to score highest;
- the mechanics that correlate with both critic and community approval;
- and the release windows where competition is thinnest.
I’m not highlighting something new. These patterns have existed for a while. I’m simply pulling them all into one spot. If you would like to see the results of my analysis, click the Classified panel below to reveal the Blueprint to a Successful Game. Additionally, you can scroll further to learn more about the process.
Data & Methodology
The dataset is sourced primarily from the IGDB API (the Internet Game Database, maintained by Twitch), which provides trusted critic ratings, canonical genre taxonomy, and structured metadata for titles across all major platforms. Additional records are supplemented with the RAWG API for games not covered by IGDB. The final dataset is enriched with the Steam API to review counts, sentiment scores, and live player data if a matching Steam app can be found. Together the three sources cover the vast majority of commercially released titles from the past five years.
Before scoring anything, it helps to see the landscape.Genre Volume Over the Last Five Years
The stream graph shows how many games were released per genre each year. This graph showcases the true volume of the dataset, and you can already explore some genre trends. Hover a band to inspect a specific quarter; click a genre to isolate it.
Getting that volume of data into a consistent shape required some heavy normalisation. Raw tags arrive inconsistently labelled across sources. For example, the same concept might appear as rogue-lite, Rogue Lite, or Roguelite depending on who tagged it. Before any analysis, each tag is normalized in three passes: first, regex cleanup standardizes case, splits camelCase, and strips punctuation. Second, a 316-entry hardcoded map folds known variants and synonyms into 22 gameplay categories and drops storefront metadata. Third, compact-key and n-gram similarity merge any remaining near-duplicates before platform-feature noise is filtered out of the analysis.
Algorithm Deep-Dive
Success Scoring
Every game in the dataset needs a single comparable number: something that reflects both the critic’s and community’s perspectives. I also consider a game’s total number of steam reviews as a metric for the community’s size. A game with a large community, even though it has negative reviews, highlights how a game has an attractive concept but poor execution. Lastly, a recency bonus (wr) scales Success scores linearly from 0.6 for a five-year-old title to 1.0 for a game released this year, which reduces the impact from older hits, providing a spotlight on the new releases.
The Success Formula- C
- Metacritic score
- Q
- (Positive / Total) Steam reviews
- P
- Log-normalized review count How P is calculatedP = ReviewCount: Steam reviews for this game.
MaxReviewCount: Total Steam Reviews in the dataset - wr
- Recency bonus: 0.6 (5 years old) to 1.0 (current year)
A single Success Score can rank games, but it does not explain which recurring traits are doing the lifting. This chart shows the score distribution behind the strongest factors in the current view. Hover a point for a quick read, or click one to pin the full tooltip.
Success Factors: score distribution in top 10 factors
Release Timing
Release timing is one of the few things a developer can actually choose, so it is worth checking whether the usual advice really holds up. I am not trying to overturn common wisdom here so much as double check it.
Average Success Score by Month
Each bar shows the average Success Score for games released in that calendar month. Darker bars mark months with denser release volume, while the label above each bar gives the exact game count, so read the height and the count together. The coloured background bands group months into quarters, with each quarter summary shown along the bottom axis. Hover a bar to see the month data more clearly.
Month-level averages can hide a subtler pattern: whether a genre’s fortunes have been rising or falling year-over-year. A high-scoring genre in 2021 may look very different in 2025 if the market has since become saturated with similar titles. The chart below keeps genre on one axis and year on the other, but turns each genre into its own trend line.
The chart helps to quickly compare current actual data versus simple trend prediction. The History cells show the last five full years, Now shows the current year-to-date result, and EOY shows the projected year-end score. That makes it easier to spot which genres are holding steady, slipping, or rebounding into the end of the year. Hover any cell to inspect the exact score and sample size.
Genre Trends: History vs Current-Year Outlook
Closing Note
Sandbox
Design your own game concept below and see how its projected Success Score compares against the dataset.
Pick your settings and generate a score to compare your concept against the dataset.