GG(x) is YOUR.GG’s proprietary model that computes win contribution(Performance Rating). Performance Rating is derived from machine learning and a comprehensive statistical algorithm taking rank tier and position into consideration. We use the following methods.
- Determine features for evaluation. Ex) Laning phase solo kill, vision score & etc.
- Calculate average value(expected value) and win contribution(correlation) of features determined from hundreds of thousands of data points categorized in tier and position.
- Calculate whether game data has met expected values for the specific tier and position. Ex) Was Support’s vision score higher than average? Did the ADC deal higher than avg. damage?
Let e be the fulfillment rate of a specific feature and let w be the contribution to a win. Performance Rating is calculated as shown below.
Performance Rating = e1*w1 +e2*w2+e3*w3 … en*wn
* Details have been left out, as it is just an illustration.
Computed Performance Rating follows a normal distribution as the following image.
Below are some examples to help better understand GG(x).
- Tier consideration: Bronze and Master tier games are different in many aspects and therefore have different expected values. For example, if an ADC had a 110 CS at 20minutes, it could be a good value at Bronze, but would be a poor value at Master tier.
- Position consideration: Supports have a higher expected value for vision and ADCs have a higher expected value for damage dealt.
- Time consideration: For example, growth is a high priority during the early phase of the game, but when everyone is near level 18, the importance(contribution to a win) of growth drops off significantly.
- Performance Ratings of Top, Jungle, Mid, ADC, and Support all converge towards 5.0.