Welcome to Steam 250, helping gamers find good games on Steam.
Steam 250 is dedicated to making good Steam games discoverable. Since 2014, Steam's games library has been growing at an exponential rate, but Steam's toolset has not matured quickly enough to be useful for finding good new games on Steam today. We provide a collection of video game rankings based on a wide selection of nuanced criteria to match what gamers are looking for. Steam 250 aims to provide all the tools necessary to be the one-stop website for finding good games on Steam.
Steam tends to focus on sales data to promote games whereas most of our rankings use gamer reviews data to rank games instead. Comparing our annual sales versus ratings rankings shows there is a wide disparity between games people are willing to buy and those they're willing to recommend. Focusing on sales provides an incomplete picture of the gaming landscape and frequently misses games more gamers are enjoying, but we complete the picture by including approval ratings so gamers can find the games they really want.
Most of our rankings use player votes to determine rank order. These votes are the thumbs up or down given with each game review on Steam. Any gamer can vote, as long as they own the game on Steam and have played it for at least five minutes, by visiting the game's store page on Steam and logging in.
At least once per day, the latest positive and negative review totals are collected by scanning each game's Steam store page. When this completes, each ranking is updated with the latest data and scored using various algorithms. The countdown shown at the top of the page estimates when the next update will finish.
A ranking algorithm determines the order of each game relative to every other game. For each game, the algorithm takes two inputs: the number of votes and the approval rating (percentage of positive votes) and assigns a score. The score is then used to order the ranking. The quality of our rankings is determined solely by this mathematical algorithm, so we spend a lot of time thinking about and scrutinizing it.
During early development of Steam 250, we trialled seven different algorithms, each with their own behavioural characteristics. Each algorithm is further tuned by a weighting variable that biases the score in favour of either the number of votes or approval rating. Each algorithm's weighting exists on an unbounded, sliding scale thus creating limitless possibilities. We narrowed down the possibilities to 22 distinct combinations of algorithms and weightings and scrutinized them until a winner could be declared.
It is important to appreciate ranking games is a complex and subjective problem with no correct answers. If game A has a 90% approval rating with 1000 votes and game B has a 100% approval rating but only 100 votes, which game is better? The algorithm must consistently provide a definitive answer to many questions like this, even though this is a seemingly impossible task. Ultimately the algorithm is blind: it knows nothing about the games other than the two inputs we feed it, and we must maintain the purity of its ignorance by discarding our biases and only examining these same figures when making decisions about how we think a ranking should look.
On a few occasions we have adjusted the weighting to fix a specific problem. We prefer to modify the ranking algorithm and weighting as little as possible, because even small changes can cause a drastic shift in the rankings. An example of a specific problem we faced were two adjacent games both with the same approval rating percentage, but a game with half as many votes was ranked above one with double the number of votes. This happened because the game with more votes also had a lot more negative votes, but proportionally, they were almost the same. It seemed the algorithm was too sensitive to fractional differences in the overall approval rating so we modified the weighting to favour the game with a lot more votes in this situation. It is important to emphasize the specific games involved are not important; we only look at the numbers when considering making a change to fix ranking anomalies.
The entire Steam 250 website is open source. All the code and data used to run this site, both past and present, are publicly available. This means the whole website and its infrastructure can be cloned in just a few clicks!
To developers, open source means the code can be browsed to see how the site works. Changes can be submitted via pull requests on GitHub to make improvements to the site, although it's a good idea to discuss change proposals beforehand.
To most people, however, open source means a commitment to honest and transparent game rankings that are not manipulated by third parties. Steam 250 will never adjust rankings to favour a particular game. It is important the game ranking algorithms we use are impartial and trustworthy. To this end, open source provides a way for anyone to independently verify ranking integrity.