No replies to this topic

[Spinfusion]

MWO could be a much more intriguing game. Its primary problem is the perverse CBill incentive structure, which discourages quality team play. Pilots are rewarded for individual performance, primarily damage dealt. This results in a selfish mob of individuals who criticize each other in chat for not putting team interests first.

A weak incentive for team play exists, since winning teams do more damage and thus earn more CBills as a whole. (Although certain low-damage roles are still woefully under-rewarded.) Competitive groups thus mostly overcome the misalignment, but PUG play remains abysmal. There will always be a difference between premades and PUGs, but much of the gap could be easily closed, resulting in a deeper casual gameplay experience.

Large rewards for a losing performance make no sense. In lore terms, MWO PUG play represents mercenaries on one-off assignments in ad-hoc units. Specialists should be paid according to their pilot skill and combat record. Ultimately, the best indicator of skill is win/loss ratio. Therefore, I will describe a method for basing CBill rewards on that metric.

I propose changing CBill rewards so that half the rewards are based on the player's win/loss ratio, and the other half on the player's performance in that specific match.

Making this happen requires a little math on PGI's part, but nothing more complicated than averages, standard deviations, addition and subtraction. Begin math!

First, we must treat different game arenas separately. I will discuss PUG statistics. The same process I describe must be applied separately to data from group queue and Community Warfare. This is because a PUG win/loss ratio is not the same as a group queue win/loss ratio. The former should not influence the latter. No shenanigans.

** Defining variables:

acm = global average CBills per match for all players since latest tweak to rewards formula
w/l = a player's lifetime win/loss ratio for the specific game type (PUG in our example)
SD = standard deviation

CBill rewards fall into two categories:
1. Rewards for winning - "win bonus", "first win of the day" (c1)
2. Rewards for everything else (c2)

** Calculations

*** Adjusting absolute CBill reward values

PGI should calculate the total CBills rewarded in c1 and c2 for a period. Let's say the c1/c2 ratio is 1/10. We want that ratio to be 1:1. So multiply all c1 bonuses by 10, and divide all c2 bonuses by 10.

Voila. Now players are highly incentivized to win. Yet rewards remain spiky, because pilots only get the big payoff if they win.

However, there's still a problem - this system doesn't fairly reward skill. Matchmaker attempts to build evenly matched teams. If it accomplishes this goal, then pilot win/loss ratio will vary far less than variance in pilot skill would suggest.

*** The pilot skill multiplier for "win bonus" CBill rewards

**** Redefining PSR tiers

The solution is to scrap the PSR 5-tier system and base pilot skill rating on w/l statistics instead.

Doing so is quite simple. Take the standard deviation of all pilots' PUG w/l ratios. These are your new tiers: +1 SD, +2 SD, -1 SD, -2 SD, etc.

The precise tier boundaries can be recalculated occasionally. Player movement between tiers can be instantly calculated based on the static boundary values. (E.g., if the +1 SD boundary is a 1.1 w/l ratio as of last week's calculation, then when Bob the Pilot wins 5 matches and bumps his ratio from 1.09 to 1.1, he gains that tier without PGI needing to redo the calculation.)

PSR is currently a grindey experience system. Partly this is driven by a legitimate concern - the w/l ratio of a new player is unstable due to small sample size. Statistics has encountered and overcome this problem through the concept of a "confidence interval". Scientists prefer a 1% error rate; however for video games I think 5% error is acceptable. Adjustment is, after all, dynamic and ongoing. Thus the errors will be quickly weeded out when they lose a few matches by bringing down the team.

So, with a 5% confidence interval, the tier boundary for +1 SD will look something like this (totally made up numbers ensue):
10 games, w/l >= 1.5
50 games, w/l >= 1.3
100 games, w/l >= 1.2
200 games, w/l >= 1.15

If desired, one can refine this further by subdividing performance by mech type: assault, heavy, medium, light. This avoids penalizing players for straying from their core competency. Otherwise an elite light pilot won't want to lurn-2-Atlas because it will ruin his ability to access elite gameplay.

In fact, it might be nice to allow a player to "declare" in advance whether a match should count towards his official PSR w/l ratio. That would leave pilots free to screw around with odd builds or master new chassis without suffering a penalty. He would thus maintain two PSR scores - one for "declared" matches and one for "undeclared". After all, it's not fair to the other elite players that he should show up when he's not trying. Let him run around with the noobs in the trial mechs instead - they won't notice a difference!

**** Formalizing inequality

Now that we have sound tiers, we must define the desired "inequality coefficient". How much more should the best be paid? How much less the worst? It should be enough to incentivize players to move up in tier rather than throw matches by DCing to stay at an easier level. However, it should not be so much that crappy players quit out of impoverished frustration.

Although, we shouldn't worry excessively about the latter possibility. A game is driven by the depth of its elite gameplay. The hoi polloi can pay to play.

Prematurely DCing losses shouldn't be too much of a problem, given queue wait times and half-strength damage CBill rewards. One could also make DC'ed-before-death losses count double against the w/l ratio. That should be more than sufficient.

** Taking it to outer space

To further increase team coordination, PGI could add a dedicated drop caller position - the "Captain". He would be the 13th man, who spectates, chats, and issues minimap orders. This position could eventually be expanded, Call of Duty style.

It only makes sense. Obviously the mechs have some kind of satellite uplink that permits them to know each others' positions and communicate without line of sight. Team chat, kills and lance damage percentages even punch through ECM. Lore-wise, this should be due to coordination with the dropship they rode in on.

Lastly, bring back mechs falling down on collisions. That was hilarious. (Kidding.)