missing info Statistics regarding the damage inflicted by Guns

[amazingme]

 Hello pilots,

  I made a little python scraper that grabs stats from SoW web site and spits out .csv file with all the players' guns, guns hits and gun kills from the last 12 days. I uploaded the .csv file [sow.csv] resulted after running the script, together with a picture [see below] of the total results per gun.

 Credits and many thanks to the owner of the site. S!

 I find the numbers for MK 108.. disappointing. At least 8 rounds per kill on average is a too high number to reflect the real data.

 I hope the devs would find these numbers as odd as I think they are, for all Guns actually, and make the proper adjustments to the DM.

 Feel free to comment... 10x

[Nirvi]

I don't think these statistic can be used for any sort of damage model adjustments. 

You could hit a B-17 one or two times with the MK108, and then use a few 13mm hits and get the kill.
It would show as 2xMK108 hits and zero kills, and 2 or 3 hits with the 13mm, but 1 kill. 

You would need  more controlled tests to get any valuable statistics.

[amazingme]

4 minutes ago, Nirvi said:

I don't think these statistic can be used for any sort of damage model adjustments. 

You could hit a B-17 one or two times with the MK108, and then use a few 13mm hits and get the kill.
It would show as 2xMK108 hits and zero kills, and 2 or 3 hits with the 13mm, but 1 kill. 

You would need  more controlled tests to get any valuable statistics.

  The data used are obtained from the stats of ALL pilots that flew on SoW in the past 12 days (~1000). This is the purpose of the OpenBeta itself, in order for ED could collect data from the players. What more significant tests would be needed here? The Law of Large Numbers (LLN) theorem proves that the average of the results obtained performing a large number of trials approaches the expected value. In other words, the more tests you do the more the results obtained  gets closer to the expected value. In conclusion. those numbers reflect quite well what's happening in the sim. 

[grafspee]

@amazingme still MK108 performing very well almost twice efficient then 20mm Hispano, what strikes me is if the difference between Hispano and mg151/20, provided that both are 20mm.

And MG 131 which is 13mm as i remember, performing at 20mm level, very impressive.

My theory is that placement of the guns make up in efficiency.

Edited January 12, 2022 by grafspee

[Mr_sukebe]

Interesting stats, particularly the comparison of hits/kills for the 20mm cannons and machine guns. In both cases the Luftwaffe weapons are proving more effective, despite being the same calibre.

ref the hits/kill ratio of the 30mm, is it that “killed” aircraft have been hit multiple times, and were “over killed”?

 

[kablamoman]

10 hours ago, amazingme said:

  The data used are obtained from the stats of ALL pilots that flew on SoW in the past 12 days (~1000). This is the purpose of the OpenBeta itself, in order for ED could collect data from the players. What more significant tests would be needed here? The Law of Large Numbers (LLN) theorem proves that the average of the results obtained performing a large number of trials approaches the expected value. In other words, the more tests you do the more the results obtained  gets closer to the expected value. In conclusion. those numbers reflect quite well what's happening in the sim. 

I'm not sure where exactly you're getting the numbers, but it's been a well-known issue that the hit-to-kill counts are nowhere near accurate since the SoW admins updated the stat tracking to include those stats again.

They can only scrape the logs for the info and DCS seems to have a problem logging individual shots/hits with the rapid fire WW2 air-to-air weapons. 

Edited January 13, 2022 by kablamoman

[Grimleo]

@amazingme does the stats reflect player vs player only or does B-17 bomber hits count too? Can you add target type to the statistics? 

I think @Nirvi has a valid point in lower caliber rounds kill stealing. The .303 Browning (7.7mm) is more effective than the .50 Browning (12.7mm) but remember the .303 is paired with the 20mm Hispano. The 13mm MG 131 is paired with either the 20mm MG 151/20 or the 30mm MK 108. The MG 131 is shown by the stats to be as effective as the MG 151/20.

The statistics do show that the .50 Browning is underperforming. I think the 8 rounds average for the MK 108 on SoW is likely due kill stealing by the MG 131. This Wikipedia article on the MK 108 cannon states the Germans thought it took an average of 4 rounds to kill a heavy bomber while it took 1 round to kill a fighter.

[NineLine]

The numbers really don't tell a lot, are these all A2A shots? Vs AI or Vs player? Were the shots made by those guns alone? Does the "scraping" know if the gun alone made the kill or if it was in conjunction with another? What is considered a kill, and could an aircraft been lost from damage and not considered a kill by these stats? Could the cannons did all the damage, but the last hit was from MG and that got the kill?

Its a neat chart, but not sure it gives enough info to make any changes. 

**Edit** I did a simple quick and dirty test, lined up about 10 AI P-51s, and just shot them from straight behind, one round, always a bail or RTB, many times the tail came right off. 

[amazingme]

18 hours ago, NineLine said:

The numbers really don't tell a lot, are these all A2A shots? Vs AI or Vs player? Were the shots made by those guns alone? Does the "scraping" know if the gun alone made the kill or if it was in conjunction with another? What is considered a kill, and could an aircraft been lost from damage and not considered a kill by these stats? Could the cannons did all the damage, but the last hit was from MG and that got the kill?

Its a neat chart, but not sure it gives enough info to make any changes. 

**Edit** I did a simple quick and dirty test, lined up about 10 AI P-51s, and just shot them from straight behind, one round, always a bail or RTB, many times the tail came right off. 

 The data collection is done by parsing the Server's log by the site back-end, probably using 'Core.Event' object methods and 'EventFunction' function calls from MOOSE.Core.  I only scraped the site and used the data that already have been parsed. The server has AI's in form of bombers and a couple of wingmen available for each side. I can conclude that most of the stats are from PvP engagements, so the AI can be neglected in this case.

 That being said, what these data tell us is that, in case of the 30mm, every other 8th round, on average, did the kill for each pilot that hit a target and that's around 9% of ALL 30mm spent. If the damage done by the 30mm had been much higher, let's say, 1 hit/kill, then the percentage hits/kill would've have been, of course, 100%. I expect this percentage to be at least 70% to be accurate and to reflect the damage done by the shell in real life, which was close to 1 hit/kill against fighters and 2-3 hits/kill for bombers. The so called 'kill steals' would average out each other as it can be the other way around where the 30mm would still the kill from 13mm. In case of the Bf 109 and 30mm, usually the pilot would shoot the MK-108 separately from the 13mm due to the different trajectories of the rounds (bullet drop).

 Overall, the canon shells (both 20mm and 30mm) do seem to behave like simple bullets or to have too less 'punch'. From my own experience in online engagements, I very rarely made a kill using only one burst, on the contrary, I may say. I usually have to spend at least halve of my rounds to kill the double tailed P47 and also noticed that AI bombers are much less resistant to 30mm than any player's aircraft.

[NineLine]

Can you share some tracks where these cannon rounds are not doing damage, or enough to you? Surely if you dont have MP tracks you can make simple ones with a buddy?

I cant seem to reproduce the issue, sure there is times a cannon round doesnt impact like I had hoped, but I chock that up more to either hitting a part of the aircraft with not much going on, or a glancing blow. Most times when a cannon around or 2 are right on the money, the aircraft in question is junk.

[PawlaczGMD]

On 1/11/2022 at 5:26 PM, amazingme said:

 Hello pilots,

  I made a little python scraper that grabs stats from SoW web site and spits out .csv file with all the players' guns, guns hits and gun kills from the last 12 days. I uploaded the .csv file [sow.csv] resulted after running the script, together with a picture [see below] of the total results per gun.

 Credits and many thanks to the owner of the site. S!

 I find the numbers for MK 108.. disappointing. At least 8 rounds per kill on average is a too high number to reflect the real data.

 I hope the devs would find these numbers as odd as I think they are, for all Guns actually, and make the proper adjustments to the DM.

 Feel free to comment... 10x

These statistics are not in-depth enough to draw conclusions about the DM. 

Hits without a kill inflate the rounds per kill numbers unevenly, and in this setting, you can often tank a lot of damage and make it back home.

Some of these guns are often fired in combination, sharing a kill (.303 + Hispano, MG 131 + MK 108). This might explain why the MG 131 has such an impressive stat, it is likely inflated by a lot by the 30 mm. Obviously, the .303 is much weaker than the .50 cal, but your data shows otherwise - because it is inflated by the Hispano that it's paired with.

MK 108 often overkills, many shots in a long burst go into a guy that is effectively already dead.

Proper testing would involve setting up a target AI, and actually measuring the nr of hits from each weapon, required to obtain either an immediate kill, or delayed kill (engine will die in X minutes). I agree with NineLine that you can kill an aircraft with significantly fewer than 8 MK 108 hits most of the time.