WAR was a fun attempt at a stat - but it had its issues. So here's the new and improved version:
Average Impact RatingTwo of the biggest issues with the WAR stat were 1) concerns about the effect of sample sizes and 2) dealing with the Score statistic in Waste's leaderboards. AIR deals with both of these problems. Here's how:
Take this hypothetical statline.
AIR takes into consideration three things: KDR, FBR and TKs. The columns labeled ">" and "<" ask whether KDR is greater than FBR and whether FBR is greater than KDR respectively (which is why you can see the FALSE there, as FBR in this case is higher than KDR). Since in this case FBR is higher, the formula
adds (FBR/10), which is 0.2, to the KDR - 0.84 + 0.2 = 1.04.
If it were the case that the FBR were lower than KDR, like in this case,
then the formula would
take away (FBR/10) from the KDR, which is 0.84 - 0.05 = 0.79.
The finishing calculation factors in TKs - it takes the amount of TKs, in this case 2, divides that by 100, and subtracts that from the previously calculated number - giving us a 1.02 for the first stat line presented and a 0.77 for the second.
So how does this solve the two issues with WAR? Well first of all, by dealing directly with ratios, amount doesn't matter - all that matters is the ratios. This solves the issue with sample sizes and removes the need for Score entirely! Since this is going to be used per match rather than per season it doesn't matter that some matches are longer than others - all that matters is the ratios put up within the rounds played.
How would this scale? Well, taking a random smattering of KDs and evaluating those numbers leads us to think that
0.79 or less is bad
0.8-1 is average
1-1.3 is above average
1.3-1.5 is spectacular
1.5+ is elite
I'll be doing those for each match during the season and periodically updating on neat trends and stuff. And I won't be the only one doing this for a stat - Russian is building a stat as well, which will also be revealed in due time!