First off, I want to give a bit of info about myself. When I played WoW, my team was ranked 10th in battlegroup and played competitively with the teams ranked above us. The only reason I bring that up is to give myself some credentials. I am not a pro when it comes to PvP, but I am pretty good at it.
For a little more info about myself, I am also a soon-to-graduate math/engineering major. One of my interests is statistical analysis applied to sports. I’m bringing this up because my suggestion is actually not original. I am going to suggest a stat familiar to anyone who follows sports analytics.
Currently, as I understand it, player ratings are based on wins and losses. If and when a ladder system is implemented, I am sure wins and losses will continue to be the sole input to the rating.
My concern is that this does not work for solo queues. Team games are never won or lost by single players. If you are rating teams, it is fine to look no further than wins and losses. If you are rating players, it is a major problem. The problem gets compounded by GW2’s small population. Players are repeatedly paired with each other, one game after another – something that should rarely happen.
There is a stat that can solve all these problems and more. It is called Win Percentage Added (WPA). What WPA measures is the contribution of a player to his or her team’s chances of winning.
For those not familiar with this stat, it is simple to calculate. All you need is a database of game states paired with outcomes. Then, every time a player changes the state, the difference between the old and new expected win percentage is calculated. In GW2, the state is: time remaining, score, team composition, points controlled, players defending points, players attacking points, and players pursuing objectives. States are saved throughout the game, and the eventual winner and loser is stored along with the states.
What this lets you do is build an expected chance of winning, based on what the current game state is. You simply look at how many teams went on to win from a given state. This is a very simple, quick calculation. For those of you who are software engineers, it is actually a constant time operation. Since this database is constructed ahead of time, all the calculations are done offline only to be stored in a hash table. (More importantly, the memory requirements can be carefully tuned by the size of your “time remaining” bins.)
The point of all this is to accurately determine a player’s skill. A terrific player on an abysmal team will get spotted. Sure, such a player’s team may lose 500-100. But a good player will still earn a good WPA because their contribution need not change in a loss. Conversely, a poor player on a great team will also be spotted. If they contribute little to the win, they will not be rewarded for it.
This is a stat an MMO could implement very easily. The memory requirements are trivial for a server, and the computation itself is constant time. But as easy at it is to calculate, it adds a world of sophistication to a rating-based system.
A common complaint is that WPA seems vulnerable to inflation. In GW2, this might mean abandoning points to capture new ones. This is understandable, but ultimately misguided.
A player who attempts to inflate their WPA must change the game state to do so. In other words, lowering your team’s chance of winning just so you can raise it again causes a net WPA-gain of zero. You get penalized for hurting your team, and fixing it just returns you to where you started.
For example, imagine a player is the last defender at a point. Hoping to inflate their WPA, they leave it to contest another point. This changes the state from “One Defender at A/B/C” to “No Defenders at A/B/C”. This will undoubtedly inflict a negative WPA for the play.
I don’t want to spend to much time justifying WPA as a stat. It is a time-tested, universally accepted tool. There has yet to be a game or sport where WPA could be abused. What makes it immune to abuse is its definition: the only way to earn it is to help your team win. Whatever potential abuse you might think of, closer inspection always reveals that either your would-be abuse actually contributes to a win or always results in negative WPA.
This isn’t the end-all stat, of course. It can’t be abused, but it doesn’t tell the whole story either. That said, it is a vast improvement over attributing entire wins and losses to a player. And for a ladder system using a solo queue, it is all you need.
(edited by zilcho.7624)