Tour 12 "Top 30 pilots" data

[bloom25]

I just spent the last 4 hours working on a little bit of statistics for the "Top 30" pilots judged by overall ranking.

Here's what I did:  I wrote down all of the kills each of these 30 pilots had in every type of fighter, the number of deaths in each fighter, the number of top 30 pilots getting kills in each fighter type, the number of top 30 pilots getting a death in each fighter type, and the overall k/d for the top 30 for each fighter.

(Note: I made no adjustments at all to the data.  I simply recorded the kills in and deaths in numbers for each fighter for each of the top 30 pilots as of 9:00 PST 1-27-1.  This means K/d will likely be slightly lower than the actual totals due to death by vehicles, bombers, and flak gunners.)

The data is VERY INTERESTING to say the least.

Here are the results:
(k/d, frequency of top 30 pilots gaining at least 1 kill and/or 1 death in this aircraft.)
a6m5 : 1.9484, 19
me109f4: 1.775, 10
me109g2: 2.2375, 10
me109g6: 2.1, 10
me109g10: 3.169, 22
c202: .88, 8
c205: 1.6154, 10
f4u-1c: 5.031, 27
f4u-1d: 2.148, 23
f6f5: 2.4775, 25
fw190a5: 1.976, 16
fw190a8: 2.506, 13
la5fn: 2.071, 22
n1k2: 3.1495, 23
p38l: 2.6758, 17
p47d25: 2.0625, 6
p47d30: 2.12037, 16
p51d: 3.8482, 22
seafire: 1.8244, 20
spit9: 3.6666, 11
spit5: 3.7155, 16
typhoon: 2.272, 21
yak9u: 2.9818, 13

Ok, you've all seen the raw data now.  Here's what makes this data special:  The data all comes from the same 30 players.  Theoretically if all the planes were balanced the kill to death should be near each other.

It's also interesting to note the number of "top 30" pilots who have flown a combat mission that resulted in either a kill or a death in each plane type.  (In other words, 27 out of 30 of the top ranked players flew the f4u-1c for at least 1 sortie.)

I've now developed a ranking system using the following formula:  (frequency/30)*(kills/deaths).  (Ex.  The f4u1c had 27 top 30 pilots use it so far this tour.  The formula would thus be (27/30)*5.031 = 4.5279.
)  The reason for the ratio preceding the k/d is to even out variances due to lack of use of one type over another.

The results:
a6m5: 1.234
c202: .2346
c205: .53846
f4u1c: 4.5279
f4u1d: 1.6468
f6f5: 2.0646
fw190a5: 1.0538
fw190a8: 1.0859
la5fn: 1.5187
n1k2: 2.4146
p38l: 1.5163
p47d25: .4125
p47d30: 1.1308
p51d: 2.8220
seafire: 1.2163
spitfire9: 1.3444
spitfire5: 1.9816
typhoon: 1.5904
yak9u: 1.2921

IMO the data is directly comparable at this point for the most part.  Since I didn't take ground, flak gunner, or bomber kills into account the cannon armed fighters are probably a little underrated.  (Very little though IMO.)

My thinking at this point is that if 2 pilots of generally equal skill met in these respective planes under basically co-e conditions, the average chances of winning for each pilot is (his planes ranking/opponents planes ranking).  If you actually try this out it generally predicts the winner.  In the cases where it fails I believe that is due to the fact that some planes are only flown by experten in this version for the most part.

Whew, that was a lot of work.  I hope we can draw some meaningful conclusions from this.  (Or maybe I'm completely wrong.   )



------------------
bloom25
THUNDERBIRDS

[Spatula]

Very nice work bloom.

What percentage of the top 30 players fly the same plane for at least 90% of them time? I would assume the top 30 would have a single ride, unless they need to do something their ride doesnt do as well.

[GRUNHERZ]

Hi

Im really sorry for doing this but why the hell does chog always manage to score the highest in whatever statistical category we come up with. Really maybe thats what can be drawn from this date, that there is just something screwey bout that damn plane and how it fits within the planeset. Compare that dhog, which is supposedly the exact same plane save armament, plus the fact that only 200 were ever made and saw limited service, sorry it just really pisses me off that such a abysmally insignificant plane in RL completly dominates this entire game. Its a Golly-geen pathetic disgrace. It really pisses me off.

thanks GRUNHERZ

[TailLights]

Quite a work, bloom.

Some Questions that arise in my mind...

1. (frequency/30)*(K/D).  I can't really agree on this (f/30)*(K/D). If we really wants to make ratio preceding for variance reduction,
 (sorties in the plane / total sorties of 30 pilots)*(K/D) should be used.
 
 The final ranking of F4u is far superior than spit9, Which I think wrong... (what if total sortie of F4u is less than Spit9? then the (persons/30) makes wrong result. )

 2. Total Kill numbers of plane X
    Total Death numbers of plane X  
      of 30 pilots ;
     are not enough? I can't really understand why we need to even out variances of a type on another... IF we really want to even it out, IMO, frequently used plane should be 'Less' counted. Top 30 pilots are of course good ones. If they fly more, they will surely get more Kills than Deaths.

So If they fly a plane regularly, Its ranking should be reduced.

 Indeed, I think the MAIN difference of those ratios between planes are mainly from this reason ; they prefer some planes. The preferred planes gets more K/D ratio.
(Btw, still c hog seems to have much higher K/D)

 We cannot just sum up or compare those ones....

eh... just my 20 cents. Maybe I'm thinking on wrong way... :>

tailight
-lyun-

[funked]

What Taillights said.  There are some guys in the top 30 who only flew 1 or 2 sorties in the Chog or other planes.

[Dingy]

With regards to rankings I think there is something screwy.  When I sorted the ranked players according to fighter ranking the #1 pilot is lumpi96 who had 1 sorty, no kills and was killed during his sorty.  He hasnt been in a fiter since yet he is #1.

-Ding

[gatt]

The ranking and scoring system is completely wrong for some guys.

[eskimo]

Interesting stuff Bloom.

I don't think the % of use, in combination with K/D, helps the data much though.

The P-47d25 and the P-47d30 have about the same K/D among these guys, but the 30 is flown by 2.66 times as many folks as the 25.  That doesn't make it 2.66 times better, however.

Also, some folks fly different planes for different purposes.  
If people tend to use the Jugs for Jabo more often than other planes, it's going to screw up the Jugs K/D.
I (and my squad) often use the A6M for base defence.  As a result I(we) get vulched in it alot.  If many of the top 30 view the Zero in the same manner, its K/D will be somewhat screwed.
Also, one person with a so-so K/D can screw-up the overall K/D of a particular ride if they fly it often.  I.E. if one guy flew over 50% of the 205 sorties, and he had a K/D of 1.0, the 205's K/D would be misrepresented.

It would be interesting to look at the K/D of the favorite rides of the top 20 Fighter and top 20 Attack pilots.  I.E. The P-51 was chosen over 50% of the time by 3 pilots with a K/D of 5.14,  ...F4U1-C by 9 pilots with a K/D of 7.33, etc.

eskimo

[bloom25]

A couple of things:  1)  I used the overall ranks, not fighter ranks.  This was to avoid the "one sortie" wonder pilots.  2)  I think my ranking system better describes the perceived usefulness of each of the fighter types.  It does take into account the k/d and the number of top 30 pilots using the plane.  On second thought, I'm thinking perhaps the root(frequency/30)*k/d would be a little better.  This wouldn't effect planes like the p47d25 so badly.

There are some real oddities in the "top 30" pilots scores.  Some of them do tend to stick to a single type and do very well in it.  Then there are the people who get a huge k/d in a varient of a fighter and can't even manage a k/d over 1 in another varient of the same fighter.

I have all the data in a spreadsheet, but there is just too much to post here.  (117 columns worth.)

Here is the total number of kills scored in each of the types for the "top 30" pilots:

a6m5: 982
me109f4: 142
me109g2: 179
me109g6: 84
me109g10: 431
c202: 22
c205: 21
f4u1c: 3904
f4u1d: 464
f6f5: 1024
fw190a5: 247
fw190a8: 426
la5fn: 350
n1k2: 1222
p38l: 586
p47d25: 33
p47d30: 226
p51d: 461
seafire: 239
spit9: 352
spit5: 431
typhoon: 284
yak9u: 164

I'll give you one more number and you can draw your own conclusions:  The f4u1c scored 31.80707186% of all the TOTAL kills among the top 30 pilots.

I'll see if I can give you a graph of the number of kills scored in the f4u1c among the top 30 pilots.  (Note:  I didn't keep track of names, only stats.)




------------------
bloom25
THUNDERBIRDS

[bloom25]

Ok, here's some data generated by my Statgraphics software:

Percentiles for f4u1c kills

1.0% = 3.0
5.0% = 7.0
10.0% = 7.0
25.0% = 28.0
50.0% = 70.0
75.0% = 224.0
90.0% = 506.0
95.0% = 508.0
99.0% = 561.0

Basically this says that on average the top 30 pilots had 70 kills in the f4u1c.

(The max is 561 and the minimum is 3 out of the 27 pilots who used it.)



------------------
bloom25
THUNDERBIRDS

[eskimo]

The number of kills per plane is interesting.
I don't understand what the F4U percentiles represent... can you explain?
eskimo

[gatt]

Eheh, look at those C-Hogs and Nikis figures. Looks like we have tons of uber-ride-dweebs amongst the top scorers  

[Jekyll]

 

What Taillights said. There are some guys in the top 30 who only flew 1 or 2 sorties in the Chog or other planes.

So what you are saying funked is that guys who only fly a couple of missions in the CHog are still able to maintain an overall 5:1 k/d ratio with the aircraft, even though they are unfamiliar with it?

[BigBen]

Don't forget that the CHOG is a very popular jabo platform.  Combined with its large ammo load, this means that the C-hog is often involved in vulching scenarios.  This is bound to greatly increase the number of kills gained by the plane, in extremly low-risk non-ACM situations.  I think that this contributes greatly to the perception of the -1C's supremacy.

[eskimo]

CC that BigBen.
The Buccaneers refer to the F4U-1C as the
"VulchHog".

eskimo