1
MWO today has many game modes with unique fight sizes, from 1vs1, 2vs2 in Solaris to 4vs4 in FP scouting and 12vs12 in QP and FP (4 spawns). A common feedback across all these modes is that the teams are not fair, one side often starts with an overwhelming advantage in skill or mech choice or both, and this does not lead to a fun match. People want fair, balanced matches where they feel challenged when winning, and feel they had a chance when losing. This is important to enjoying any game.
The reason the teams tend to be unbalanced is how the Match Maker (MM) creates teams. The MM uses a Tier system which doesn't accurately reflect a player's skill. It does not look at mechs or loud-outs selected, it does not consider the map or mode, it does not look at synergies between group members, and in the case of Solaris it does not give any consideration to the relative strengths and weaknesses of a match-up. The bottom-line is that it is not putting teams together smartly.
To fix this problem once and for all, I want two changes:
1. A Match Maker that looks not only at a pilot's past performance, but also the mech selected (it's load-out, armor, skill status, heat efficiency, effective range), how this mech+load-out performs on the current map, how large a group the pilot is in and their synergy together.
2. Dynamic team numbers, so that for QP and FP, the Match Maker is allowed to make teams of greater than or less than 12 in order to create balanced matches in a reasonable amount of time. This may mean updating maps so that turret and base health or capture time can be dynamically adjusted to team size.
The impact I expect to see in QP and FP is that teams are much more balanced because the MM knows not only how skilled a pilot is, but also whether they are using a good or bad mech whether it performs well on the current map. In group queue, a team of 8 great pilots can be matched against 16 intermediate pilots if that is what it takes to create a fair match. (If 12vs12 can be done, the MM will still do so, but its hands are no longer tied to 12 players per team nor 24 players in total)
The impact I expect to see in Solaris is that all these division of mechs that don't really work can be re-merged into one. The MM knows if two mechs are a really bad match-up, for example, a locust with a streak maddog, or a streak maddog with an atlas, or an atlas with a locust. It will take both mech chassis and load-out into consideration to avoid making lopsided matches. Matches with clearly a stronger mech versus a weaker mech would only be made if the MM sees a sufficiently more skilled pilot in the weaker mech.
As far as whether this MM is even possible to implement, it is. It requires a statistical general linear model with all of the predictors listed above, and it will figure out how much to weigh each of the predictors in a way that allow an MM to assign a value to each pilot and mech. This is something I do every day at work, even if you don't understand it, only the programmer implementing it needs to.
Discuss?
Edited by Nightbird, 14 June 2018 - 03:23 PM.
