Greedy Goblin

Tuesday, May 21, 2013

World of Tanks cheat: lucky and unucky shots

This is part 2 of a series, see part 1 first.

When you fire a shot, you can hit or miss. A hit can penetrate or not. A penetrating hit can damage or not. Damage can vary according to the random generator. Lot of variables, but all can only depend on tank stats, player skill and random generator.

I gathered data from 374 matches. All from the same tank, same gun: AT15, B-barrel. During these matches I faced practically every tank randomly, the large number (15*374 = 5610 enemies) guarantee that I had an "average target tank". My skill couldn't change much as I've had more than a thousand matches with British TDs and more than a hundred with AT15 itself. The only factor remaining is random luck. Let's see how random it is.

The average damage/hit is 141.3 (gun nominal value: 230). The standard deviation is 39.6. Now, according to the distance from the average I classified matches as "unlucky" (below average damage/hit), "normal" (around average) and "lucky" (above average). The boundaries were symmetric of course and were chosen as wide to make 1/3 of the matches "normal". With random distribution it would mean 1/3 lucky and 1/3 unlucky matches. It wasn't true, I had 29% unlucky and 38% lucky matches but let's ignore it now. It's important to know that the above classification ignored if I won or lost.

Let's see the average damage/hit for the three categories, separating wins from losses:
The lucky wins aren't luckier than the lucky losses, so far, so good. But the extreme differences are troubling at best. I mean in the luckiest 1/3 of the matches I damaged almost twice as much as in the unluckiest. While random generator should be part of the damage/hit calculation this is a bit large, don't you think? Our skill should have larger effect on the outcome than RNG. But this is no cheat.

Let's see how being lucky affected the outcome of the matches?
Oops! Why is it asymmetrical? I mean if my damage/hit increases from 100 to 140, it increases winrate by 6%. If it increases by an equal amount, it has no effect on the winrate. Luck is strange in World of Tanks. It has a binary "yes or no" form. Like it would be universal, deciding only one thing: are you meant to win or lose.

As a tank destroyer, especially a 20km/h slow one, I can shoot only when the enemy can kill the rushing faster tanks and break through. If they are lucky, I don't get shots. So my number of shots is a good measure of their luck and obviously should be uncorrelated to my luck.
Uncorrelated heh? The team spirit is strong with this game: we are lucky together or unlucky together. You can see it yourself all the time: winning on one side of the map should be uncorrelated to winning on other side. Yet you rarely see games where your team reaches their flag on the left and they reach your flag on the right. What you see is your team being obliterated on both sides or winning on both sides.

Enough of stats, let's see how far "luck" can go to support a player! The infamous scout tank, T-50-2 can only trust in its speed as its 37mm armor on all sides can't stop anything that Tier 9 tanks throw at it. If it is hit, it's probably oneshotted due to its 500HP. Seriously how many shots and potential damage can this thing take before exploding?
OK, there is one way to survive penetrating hits: component damage. The tank loses some part or crew but no HP. But even that wouldn't help a scout much: lose track or engine and it can't dodge more hits. Get a hit to the fuel or ammo and it burns or blows up. A hit in the radio or the periscope would stop it from spotting and relaying the location to the others. Where can you hit the damn thing without disabling it?

Want to be as lucky as him? Wait until the exploit post on Friday and I tell you how! Tomorrow we'll see that the legendary "I lost because I got into a fail team" has a strange meaning in World of Tanks.


Anonymous said...

It's become very obivious on artillery. Same map, same spot, same arty, same firing distance other and other, yet one battle it fells like i'm firing guided missiles, other still guided, but away from enemy. It's almost never in the same battle, it's either ALL shots missing/no damage/10hp or ALL shots one-shotting/crippling/burning. Very interesting RNG they have here.

Anonymous said...

You have made one valid point, and one flawed point.

The Valid one = gun dispersion. This factor is random. WG have now admitted that although the math worked for how the game was set up in 2010, it is now time to change the the Sigma setting in the Gauss dispersion, see;

The flawed point = player skill. Whoever shot the T26 in the lower glacis (a) did not know the tank, and (b) shot it in the wrong place, and (c) the T26 is slower than most heavies and so cannot usually death circle other tanks.

Does this mean a skilled player either (a) remembers the weak spots and characteristics (such as speed and reload time) of every vehicle? or (b) uses a mod such as Hitzones to paint tanks different colours to assist with aiming at the correct weak spot, cf;

Lastly, have you tried using XVM in it's different versions, and looked at the statistics it produces about different players and their skill and effectiveness?

Cf ;

Trespassers W said...

As owner of S Pershing and T-50-2 I can say frontal armour of SP have some plating, according to dev posts after each penetrating of platings shot direction, penetration and so on is affected, so, your shot can land on actual armour after penetrating platings in wierd angle. About your T-50-2 case, it make me laugh. Such cases rare, very rare. Also do u recording matches? It'll be interesting to download some rare cases and analyse them.

Trespassers W said...

Oh, okay. I found some collected data, and it seems that WG cuts damage normal distribution in 2-sigma, so ~20% of shots are min or max damage. Thats applied to gun dispersion as anonymous said, and be fixed in 0.8.6 (only gun dispersion)
2-sigma seems to be hard Belorussian way to make games. Link with math (in russian)

Anonymous said...

You have a flawed premise. You say that the game will let you win regardless of your skill over time to balance WR.

Then you go to show that it ups your level of skill through increasing your luck. (very poorly btw)

But these two statements are logically incongruous. If your first was true, then you would be able to AFK through this game and have a 50% WR, as your play has nothing to do with winning. Have you performed that experiment? If not, then you have not tested your first proof.

Only when your first proof is dis-proven should you start chasing your second point as to what other factors could be determinant to account for the behaviors you believe you are seeing.

Devin said...

Wait, you think the negative correlation between shots fired and damage per shot is evidence of WG cheating? Maybe look for confounding factors first.

1. Let's say one match is a close-in city brawl. Let's say another is a wide-open field. (Himmelsdorf vs Steppes, say). In the first case, you probably don't get many shot opportunities because the buildings block most of your lines of fire, but when someone does turn a corner, you are very likely to hit and pen. In the second, you will get a lot of chances to shoot, but at longer ranges, against faster-moving targets, and with fewer opportunities to aim for weak spots or take square shots on side armor.

So we'd expect to see more shots for less damage per shot on wide-open maps than on restricted ones.

2. Likewise, consider the effect of enemy armor. If you shoot at an IS-8, you are very likely to pen every time and kill them in eight shots. If you shoot at an E-75, you could easily bounce half your shots if he's well-angled and at a bit of range. Assuming he's rocking the stock turret, they have almost the same HP, so your damage per shot just got halved and your shots taken just doubled.

So no, I would not expect your number of shots to be uncorrelated to your average damage. I would expect certain tactical situations and certain opposing tanks to affect both statistics.

Anonymous said...

Another thing--how many times is a tank left with <100 hp. I believe that WG manipulates the damage so tanks can take more damage in some circumstances.

Novoskaya said...

A major thing not taken into account is the match rating. Are you in a match full of Tier 8 tanks, Only Tier 9 tanks, or 50% Tier 10 tanks. The difference between shooting a T8 tank and a T10 tank is huge. If your matches are split 29% with T8s, 33% with T9s, and 38% T10s, then your "luck" is fully explained. But we have no visibility there, do we.