Greedy Goblin

Tuesday, January 31, 2012

Matchmaking fail due to player stupidity

Many players of World of Tanks claim that the matchmaking system is bad and believe that the company is intentionally messing with them, giving them impossible matchups.

To check this, I simply guessed which side will win before the match started. I had 68% chance to guess correctly in 350 matches since I started this experiment. This is extremely high. I mean that if we round it to 66.6, it means that only 2/3 of the matches are even, 1/3 is pre-determined (I'm wrong 1/3 by guessing, so I'm also right in 1/3 by blind guess, the rest 1/3 are not guessed). This is really a serious matchmaking issue.

The reason why there is only murmur on the World of Tanks forums and no uproar, is not only that you have equal chance to be in the "I win even if I'm AFK" team as in the "if I win this, I get a Kolobanov's medal" team. The reason is that the matchmaking system is technically not wrong. It's wrong in handling player stupidity. It doesn't misplace certain tanks (for example placing a Maus to a Tier 3-5 match would be such mistake). Sometimes the same tanks win, sometimes lose for no obvious reason. Also the same matchmaking is not wrong for organized teams, meaning if two equally good clans would play with the sides, the outcome wouldn't be pre-determined, so the problem can remain hidden front of experienced players and developers.

Let me show some schemes how ways matchmaking fail. The first is "alike teams win over diverse". This is totally counter-intuitive. If some clan leader would be given the task of "create a perfect team from tier 8 tanks", none of them would create an all-Type 59, all-ISU152 or all-Tiger II teams. Such teams would be devastated by a properly organized team. Yet in a random match I'd guarantee that either of the above teams would have 95% win rate against a diverse team. Why? Just look at the way how different teams position next to an obstacle (Meds have their cannons higher so they can shoot over the low side of the obstacle)
Tanks of the same type play similarly so they can naturally support each other. To support a different tank, you need to think, which is impossible for the average player. He ignores line of fire of other tanks, especially if they are away from him. Also, he is unable to determine the actions of the enemy team, even if it's obvious. If on Ensk one side has 5 Type 59, it doesn't take a genius to figure out that they will rush on the park side. Yet the enemy won't place heavies and TDs to the end of the park side and hide the arty in the city side. They will send their 2-3 mediums on park side and the arty will be in the park side corner as always.

The second is the scout rush disaster. For some weird reason people believe that the job of the scout is not scouting but to rush to the enemy base and kill the enemy arty. So they try. They might succeed, they might fail, but at the end of the first minute they will be dead. Or yelled at. Since artillery is extremely effective, their success or failure can make or break a match. It depends on the amount of scouts vs amount of tank destroyers and slow heavy tanks. The TDs camp, the slow heavies simply did not get far enough. So looking at the scout:TD/slow ratios of the teams can quickly tell what will happen with the arty. Obviously no game mechanic stops a fast medium from camping next to the arty for just one minute. But they just won't. For example we have 3 scouts, the enemy has one. If they have 7-9 slow tanks/TDs, we are outnumbered 14:12 after the first minute. If they have just 4-6, they have no arty.

The third is the "no arty: campers party!". A standing tank is less visible and aim much better than a moving. So in a 1v1 battle a standing tank always wins. To avoid everyone standing, artillery was introduced to the game in a very unnatural way: the arties in a WWII settings are capable of XXI century GPS-based indirect fire with XXI century accuracy. Their firepower is also huge. Practically they can oneshot everything that is visible and stands in the same place long enough to lock on. This feature counters camping, forcing tanks to move after revealing their position by firing. The amount of artillery varies in matches and strongly effect the value of different styles. Since players are usually dumb like a rock, their style is pre-determined and usually reflected by their chosen tank class. The fast medium will rush even if there is no enemy arty. Even better, the scouts will rush too and after death they will spam "ffs arty i spotted dem why no shoot". TDs will camp even when the enemy has 4 alive arties. So looking at the arty count and the camper count you can tell who will win. This problem is not obvious to a developer who would naturally play more defensively when no arty barrage is incoming and move more often when there are 3+. Of course this interacts with the second point which decides arty survival.

The fourth is my personal favorite: "kill them all XDDD". It happens when the highest level tanks on one side are fast while the other team has slow or TD as top tanks. This means that the first half of the battle the strong fast tanks massacre the weak enemy fast tanks, getting 5-6 kills ahead. Despite all the dead are from the lowest of tiers, they declare themselves winner and rush out to hunt down the few "remaining survivors". The battle between a camping 100% HP ISU-152 and an incoming 40% Type59 who races with some T6 med who gets to the "easy kills" first is pretty fun... if you like camping.

There can be other schemes which are unknown to me as I don't play those tanks, so probably after playing all classes I could guess who'll win with 80-85% accuracy. However the bottom line of all these can be given: since the average player can't adapt, certain compositions that support strategies needing little thinking (all rush left ftw) are overpowered, while other compositions that tempt into stupid mistakes are underpowered, despite the same compositions are OK in the hand of competent players.

Playing a game where victory is decided before start is not fun. This can be solved two ways: one is implementing a brain into players, but that's out of scope of a gaming company. The other is much more restrictive matchmaking system that enforces that both teams are equally strong in every tank classes (TD, medium, heavy, light, SPG) and not just overall. Also the most determining class, the SPG should be pre-set to 1-3, so there is no battle without SPG or with 4 and more SPGs. It would slow down the matchmaking.

World of Warcraft solved this problem by class homogenization and "bring the player, not the class". Being a mage or a rogue is merely cosmetic difference. The only real difference is between roles so the same problem exits on the random BGs: you can tell who'll win by counting the healers on both sides.


Anonymous said...

Sometimes you can count the healers.

Really it comes down the specs and HP pools on both sides. 5 healers with 120k HP each aren't going to last long against enemey DPS with 150k HP from a raw output perspective.

I tend to evaluate the HP of each team. If their team has substantially more players of higher HP than mine, I eat the 15 min debuff and do something else.

Bworf said...

Another thing that makes a huge difference is the number of and relative strength of platoons on the respective sides in a random battle. A team with 3 coordinated players in a platoon where they are on the upper half of the relative team strength has a huge advantage compared to random morons. While incompetent platoons occur, platoon members are far more likely to be coordinated and good players than others in my experience.