Author Topic: Statistical analysis of machines (Read 403 times)

Puck · « **on:** August 21, 2006, 12:03:17 PM »

Since I started flying again, and lately have been looking in on the BBS, I noticed there has been no significant change in the concerns regarding certain aircraft and how they should be perked. Being a naturally curious sort I decided to look into the numbers and see how these aircraft stack up.

Methods: I immediately removed the data for Gunner and Chute because K/D with these entries in anomalous. This has the effect of changing the ratio of kills to deaths, but the overall change in results is negligible. I pulled the data for this analysis on August 20th at about 11:00 PM MDT, so it pretty much covers the first 20 days in August.

The initial analysis includes all remaining rides. I will break out Aircraft and Vehicles in separate sections.

Overall Kill Totals
Kills: 400,212
Number of Cases: 84
Minimum: 42
Maximum: 74,386
Mean: 4,764.4
Standard Dev: 9,154.6

The bell curve for the data here is very, very skewed to the left. The first standard deviation goes through the mean and median. Cases in the second and third sigma (standard deviation) are listed below. Entries with an asterisk (*) are perked.

Quote

Vehicle Kills K/D[1]
Second Sigma--
Hurricane Mk. IIC 10,646 1.4
Bf-110G2 10,765 0.8
Typhoon 1B 11,418 1.5
Tiger 1* 15,290 6.4
P-51D 15,330 1.1
Ostwind 18,056 1.8

Third Sigma--
N1K2 20,877 1.3
Spitfire Mk XVI 20,898 1.1
La-7 23,222 1.2
Panzer IV H 74,386 1.0

These ten rides ( 11.9%) account for 55.2% of all the kills earned in the main arena. while the other 74 aircraft and vehicles share the remaining 44.8%. In terms of overall K/D[1], the second and third sigma are responsible for 47.0% of all deaths which gives them a 1.17 K/D ratio.

Aircraft:
If GVs are removed the numbers are:
Aircraft Kill Totals
Kills: 276,543
Number of Cases: 73
Minimum: 42
Maximum: 23,222
Median: 2199
Mean: 3788.2
Standard Dev: 4895.9

In this case the second, third, and fourth sigma are:

Quote

Vehicle Kills K/D[1]
Second Sigma--
C.205 4,901 1.2
P-38J 5,083 1.4
Spitfire Mk. VIII 5,105 0.9
A6M5b 5,199 0.8
B-24J 5,644 0.4
Fw 190A-8 5,673 1.2
Ki-84-1a 5,792 1.3
Il-2 6,473 0.9
F6F-5 6,722 1.0
F4U-1C* 7,733 2.3
SeaFire 7,795 0.9
F4U-1D 7,885 0.8
Fw 190D-9 8,770 1.4

Third Sigma--
Hurricane Mk. IIC 10,646 1.4
Bf-110G2 10,765 0.8
Typhoon 1B 11,418 1.5

Fourth Sigma--
P-51D 15,330 1.1
N1K2 20,877 1.3
Spitfire Mk XVI 20,898 1.1
La-7 23,222 1.2

If we consider only the 4th sigma aircraft; the BIG 4 that are most commonly seen, the results are:

% of total aircraft available: 5.5%
% of total aircraft kills: 29.0%
% of total aircraft deaths: 23.8%
K/D for these 4 aircraft: 1.15

So it would seem the dominance of these four aircraft is due to their ubiquitous nature, not their superiority in the arena. Given these are all late war aircraft their conglomerate K/D is not impressive.

Vehicles:

If only vehicles are considered the results are:
Vehicle Kill Totals
Kills: 123,669
Number of Cases: 11
Minimum: 374
Maximum: 74,384
Median: 2,473
Mean: 11,242.6
Standard Dev: 21,817.1

It is interesting to note that the 11 vehicles (13.1% of possible rides) are responsible for 31% of total kills. The ground war would therefore appear to be a significant fraction of game play; far more than the 13% allocated to it in the hanger. Because there are so few vehicles they are all listed below:

Quote

Vehicle Kills K/D[1]
Jeep 374 0.14
LVTA4 547 0.30
LVTA2 796 0.33
M-3 944 0.18
M-8 1,021 0.44
T-34 2,473 0.70
M-16 4,602 0.65
PT Boat 5,180 0.40
Tiger 1* 15,290 6.40
Ostwind 18,056 1.79
Panzer IV H 74,386 0.96

Obviously the Tiger, Ostwind, and Panzer are the most common ground vehicles. The Ostwind has the highest K/D ratio of all non-perked vehicles, including aircraft. The Bf 109K-4 is second with a 1.58 K/D and 4,537 kills.

Conclusions:
I think everyone here can (and will) draw their own conclusions. While the Ostwind is arguably the most dominant of the high-use machines, it is limited by a pathetic rate of climb and very poor high altitude performance. In this case, high altitude is anything AGL...

I am working on a dominance model that includes such things as perks, K/D, kills, and possibly a few other factors that will enable me to rank machines based on how dangerous they are (on average) in the main arena. As always these models fall apart on edge cases; Drex in anything is always dangerous. Me in anything is always harmless.

Notes:
[1] K/D is calculated by kills/deaths, and is a true K/D ratio. K/(D+1) is kills/life, and not used here as none of the machines have never been killed.

zorstorer · « **Reply #1 on:** August 21, 2006, 05:32:10 PM »

Lordy puck, lots of free time or did you find a lost bag of weed?

g00b · « **Reply #2 on:** August 21, 2006, 06:36:36 PM »

Very interesting, keep it coming.

drkorf · « **Reply #3 on:** August 21, 2006, 08:35:10 PM »

Interesting. Please share more. I think K/D should be calculated as (K+1)/(D+1).

Krusty · « **Reply #4 on:** August 21, 2006, 08:42:30 PM »

K:D is really pointless because of all the newbies flocking to the "easy" rides. All the augers. All the afks. All the GVs killing them, and so forth.

You can't use kill:death. I think you might have to use kills alone (aircraft kills), and even then that's misleading due to the amount of folks with 2 accounts that get 150 kills in a p40E before landing, or the number of planes that are predisposed to vulching, and so forth.

Puck · « **Reply #5 on:** August 21, 2006, 08:51:05 PM »

K/D is actually a sticky wickett.

PURE K/D is meaningless until there is a death, which is why Hitech uses the Kills/Life ratio (the number of kills per "life" in the arena). Unfortunately K/L doesn't roll as nicely off the tongue, but it DOES give you a good idea how you're doing. If your K/L is greater than 5 (or greater than 10 if you're a leather panty wearing Luftwobble) you can call yourself an ace and everyone else will ignore you.

Ok, the REAL reason for using K/L is the whole division by zero thing. Until you die the formula blows up, burns down, and sinks into the swamp.

I'm not sure what Kills+1/Lives would tell us? Without ever having flown an aircraft you have 1 K/L? I think you're trying to neutralize the efect of that D+1; 2 kills, 1 death, and a K/D (K/L) of 1, which is annoying even if it's accurate. 1 kill for every life spent in the arena.

Puck · « **Reply #6 on:** August 21, 2006, 08:52:58 PM »

Quote

Originally posted by Krusty
K:D is really pointless because of all the newbies flocking to the "easy" rides. All the augers. All the afks. All the GVs killing them, and so forth.

One of the bits of missing information is number of sorties flown by each aircraft. K/S would be a useful addition.