Greedy Goblin

Wednesday, March 14, 2012

With more experience they get ... worse

I wrote that I had 60% winrate with the top tier tank "Object 704". That alone should prove that experience does not make players better (after the first few hours). However even more disturbing data came up.

My girlfriend finally got her Object too and off we went for some massacre. Which never came. Playing together we achieved the terrible 46% winrate. Considering that I'm alone capable of getting 60%, she must be awful. Except she had 1.33 kills, 1770 damage and 610 XP/battle, where the average player who played 5000+ games has 0.77 kills, 1100 damage, 458 XP. To make things worse, in tiers 1-7 the two of us has much higher winrate than me alone while in tier 8 a bit better. The offical page places her to #2700 in winrate, which is really good, and above mine.

So we take a player who can do 60% alone, add another way above average player and get below average winrate. What the hell? We were really puzzled and upset about the results. We started recording battles to figure out how we started to suck. Of course we had facepalm moments but usually we did great and still lost. Like some evil force provided us the most utter idiots. However the idiotfest is not an answer. The enemy should be idiots too so we should win.

The first thing we recognized is the almost total lack of the most expected outcome. I mean if there are two lanes in the field one would say that one team wins in one and loses in the other, according to which team sent more or better tanks to that lane. The outcome would depend on which lane is finished faster allowing that team to cap or attack the other lane from behind. This is the outcome we almost never see. Almost always we see all-lane defeats or all-lane victories. The replays explained why: when the red dots started to show up on the minimap, the tanks in the other lanes ignored everything and started running towards them, usually to their sure death. The side that could show up with force in one place (even in a totally irrelevant place) usually won as they could snipe down the mindless lemmings running to them one by one.

Then I checked the tank statistics again and made this winrate-tier graph. Every point is made from the average of the tanks belonging to that group and tier (so the T9 TD point is the average of the winrate of obj704, T95, Jgdtiger):

One explanation is that TDs and SPGs are getting worse as we progress in the game, so in clan wars no one shall play TD or SPG. The other explanation is more believable: the TDs and SPGs don't fit into the moronic "kill all red lol" strategy. The heavies are the ones the others rush to. The mediums are rushing. The TDs and SPGs simply don't get a shot as they are constantly blocked by rushing ones and can't rush themselves.

But why do the lines have trends? Why not TDs and SPGs are all-time low while meds and heavies are high? Because on lower tiers the new players are cautious and aware of their shortcomings. They think twice before moving. You can reason with them. If I tell a KV that running on the field of Malinovka is bad idea, he'll probably listen. From a Löwe, the answer will probably be "lol". The top tier random players learned nothing but got strong ego. They just got worse.

How could I get 60% then, alone in a TD? Because I played late night and early evening when I couldn't sleep. I played with WoT fans. I could talk to them. They proposed strategies. They cared about my line of fire!!! With my girlfriend we played in the afternoons and weekends when the worst swarm of lolkids were on. They just rushed to the first red dot like moths to the flame. Figuring this out we started playing 1-2 matches at 7 AM, right before we went work. On the left when we stopped playing in the afternoon, on the right the results after two weeks of morning games:
Despite her kill rate dropped to 1.24 and the damage only increased a little, the winrate went up to 62%!

Figuring this out, the most obvious way would be switching to med tanks and have a massacre in the afternoon. However we came up with a much wittier solution, you'll hear about it in a few weeks.

But the result for today remains: the M&S get worse by gaining experience as they get more self-confident, but not an inch smarter. I can't even believe of more elegant proof that the M&S is not "newbie", "undergeared" or "unlucky", just a bunch of lazy idiots.

PS: Quick EVE market update cash+sell+buy 1.31 B (0.5B gifts)


Kuon said...

Have you heard on the Dunning-Kruger Effect.

It might give some insight into the actions of the M&S.

Gesh said...

"With my girlfriend we played in the afternoons and weekends when the worst swarm of lolkids were on."

Nice, I've exploited the same thing in Starcraft 2, with the small difference I played at the afternoons on 1v1. Easy wins.

Kelindia said...

The reason you were getting bad results was that you simply weren't playing the way the map was shaking out. If you see a lane open it is not your job as a TD to cover it. The majority of the time you will not be able to finish someone off before they spot you.

As you have said previously your style as a TD requires you to play long range support. This means that you need to either be supporting the lane that is being pushed or in a position to prevent rushers from eliminating your arty and gaining flanking positions.

World of Tanks is a game of entering your tank into a position where you give yourself or a teammate an advantage over the other team in what would normally present itself as a 1 vs 1 battle or 2 vs 2 battle. This is why the light and mediums that rush for your arty are morons. They in no way have an advantage over anyone during their mad dash.

The only other thing to remember is that camoflage training has a huge impact on TD play. If endgame is balanced for max skills it would likely explain why TDs show a significant decrease in win percentage.

Anonymous said...

"But why do the lines have trends?"

Because a linear regression always shows a trend, whether there actually is one or not. One outlying data point, when you have such small data sets, can easily create the illusion of a trend. If you level out TD level 3 or Medium Level 8, then these two classes are essentially flat. The TD data is so noisy that it is meaningless - Down, Steady, Up, Down, Up. There is no way to predict where it would be if there was a Level 10. The only real trend I can see in this is the upward one for Heavies, and that doesn't appear until Level 8.

Casares said...

TDs _are_ obsolete in clan wars: There's nothing they can do that a t10 heavy can't do plus some more.

Decreasing arty winrate can actually be interpreted as learning effect: once you know how they work you will make sure to always move or take cover relative to where you assume the opponent arty to be.

Arty _is_ useful in clan wars due to teamwork. Simple example: arty announces target, med/heavy destroys track of the target.


Your entire post doesn't explain why your winrate is _below_ 50%. No matter how you twist it, 15 lolkids versus 13 lokids + 2 experienced players should still give you over 50%.

Phelps said...

It seems to be an EU/RU server issue. Play on the US server is more like you would expect.

It really comes to light when the test server comes up and the US server people are thrown in with EU and RU players. The reaction is always the same -- "these people are terrible lemmings."

I often wonder if there isn't some sort of social commentary in there.

Antivyris said...

Your entire post doesn't explain why your winrate is _below_ 50%. No matter how you twist it, 15 lolkids versus 13 lokids + 2 experienced players should still give you over 50%.

Of course it does. You are assuming that it was a level playing field. It clearly is not. You see "2 Experienced players, 13 Lolkids vs 15 Lolkids", but what you should see is "13 Lolkids vs 15 Lolkids". The unfortunate truth when you deal with idiots in large numbers is that, sometimes, there is no way to assist the idiots with any meaningful way, and by their shear idiocy you become useless. Every tactic you could use won't work, and every tactic they use will make no sense. You can 'try' to assist, but at that level of stupidity the entire fight really does come down to 15 vs 13.

Never underestimate the ability for idiots in large numbers to cancel out much, if not all the work done by non-idiots.

Gen said...

Kuon's mentioning of the Dunning-Kruger effect is apt:

"The Dunning–Kruger effect is a cognitive bias in which the unskilled suffer from illusory superiority, mistakenly rating their ability much higher than average. This bias is attributed to a metacognitive inability of the unskilled to recognize their mistakes" (to quote Wikipedia)

What I want to know is whether it is just coincidence that this false confidence on one's own ability can be abbreviated as "The DK effect" ?!!