Some time ago I did Battlefield 4 kill distances post which aimed to answer question at what distances Battlefield 4 (BF4) soldiers manage to kill other soldiers (and by rough extension, what are the engagement distances). Following this we decided to do the same for Battlefield 1 (BF1).
Compared to BF4, the methods used to gather the data are more sophisticated. The post-processing/analysing also includes additional steps to remove e.g. bias caused by some maps being more common than others.
Details of the data:
- Using total of 2000 recorded kills per map (BF1 vanilla maps + Giant's Shadow) from multiple rounds. 2000 samples were randomly sampled from larger set of records per map.
- Only kills by primary and secondary weapons were counted (no vehicles, gadgets or melee)
- Total of ~15000 records used for plots and analysis below
We could also "patch around" bias from uneven distribution of the weapons, but the goal of numbers shown here is to reflect average game of BF1, including the players' taste for certain weapons or weapon classes.
I.e. these numbers reflect what a BF1 player should experience on average while playing any of the maps.
General kill-distance density
Following plot displays density function of the kill-distances, i.e. how large fraction of players were killed at given distance.
Mean, median, 75th and 90th percentile are also included in the image. "Xth percentile" means point where X% of the density is below that point.
E.g. In this plot 75th percentile is distance below which 75% of the kills happen.
(In case image is not visible: Mean: 30.6m, Median: 15.6m, 75th percentile: 37.0m, 90th percentile: 72.9m)
More than half of the kills happen at under 20m, with high number of kills focusing at around 5-10m and a long tail going well above 100m (max was 428m). 90% of the kills happen at under 73m.
The popularity of SMG (close-range) weapons could partly explain this.
Comparing to BF4, BF1 shares very similar kill-distance density graph and summary values. See more on this few headlines down.
Kill-distances per map
Some of the maps have vividly different themes: One has lots of open area while another takes place in a city. To study these ranges we can calculate statistics per map.
Note: Here we use all available samples per map (more than 2000 in most cases)
Map Mean Med 75th 90th
MP_Forest: 17.9 11.6 21.4 42.7
MP_Amiens: 20.7 13.9 29.6 48.9
MP_MountainFort: 27.5 13.1 30.8 63.9
MP_Chateau: 23.5 14.5 31.2 54.1
MP_Giant: 30.6 17.0 40.1 70.9
MP_Scar: 34.7 17.1 41.7 95.4
MP_ItalianCoast: 33.2 14.9 44.9 87.7
MP_Desert: 37.0 19.2 46.0 86.4
MP_Suez: 39.1 22.0 56.4 101.3
MP_FaoFortress: 53.2 25.3 64.5 153.8
There's clear difference between some of the maps: Fao Fortress (mountains) has roughly 20m higher 75th percentile than other maps (minus Suez). Interestingly Suez (desert urban-ish) has the second largest kill distances, beating Desert map (desert with lots of open area) by 10m in 75th percentile and 15m in 90th percentile.
At the other end of the spectrum, Amiens (city) and Forest (thick forest) have more than 90% of the kills happening at under 50m, which is below optimal range of longer range snipers such as M1903 and Gewehr 98.
BF1 kill distances vs. BF4
Note: BF4 analysis was done slightly differently, e.g. it included gadgets and melees, and did not 'fix' map unbalance. As such these results should be taken with a pinch of salt.
Since we have very similar analysis done on BF4, we can compare these results. Along with comparing summary-statistics like mean and median, we can plot the densities in same plot.
Studying the graph, BF1 has lower fraction of kills at distances 20-40m but slightly more at above 40-80m.
Using the numbers from BF4 post, mean distance for BF4 is 35m and median is 17m. BF1 has mean of 30.6m and median of 15.6m, both of which are lower than for BF4. While this could indicate that kills in BF1 happen at closer range than in BF4.
It should be noted again that BF4 had different methods for gathering and analysing data. However there does not seem to be any exceptional difference between BF1 and BF4 kill distances.
Horizontal vs. vertical aiming ("Vertical gameplay")
We can also analyse what is the vertical angle between shooter and the target. This "vertical gameplay" where soldiers could have very different heights (e.g. player on a skyscraper and another on ground) spurred some discussion during BF4.
Following plot illustrates density of angles between shooter and the target, from the point of view of shooter. Negative angle means shooter had to aim downwards.
50% of kills happen between vertical angles -1.7 and 1.5 degrees, while 90% happen between angles -12.8 and 14.1. This was calculated by removing (1-0.5)/2 and (1-0.9)/2 of the density from both sides.
Slightly more than half of the kills happen inside 2 degree vertical deviation from horizontal line. At 37m (75th percentile of kill-distances) this translates into total of 2.6m deviation which is roughly 1m more than soldier's height.
The median of vertical angles is 0 and mean 0.45. Interestingly mean is slightly positive and the 50%/90% boundaries have larger absolute value for positive boundary. If we assume players always aim at same part of the body (e.g. torso),
this would mean kills happen slightly more often uphill (shooter aimed upwards) than downhill.
Conclusions and possible future work
Quick summary of conclusions:
- Most BF1 infantry-on-infantry kills happen at under 20 meters from shooter to target. 90% of kills happen at under 73m.
- No exceptional difference from BF4, there seems to be less kills at 20-40m range and more on <20m and 40-80m range.
- Clear differences in kill-distances between some maps. Median ranges from 12m to 25m, while 90th percentile ranges from 42m to 153m.
- Majority of kills happen at "horizontal" level. I.e. Shooter and target were on similar height. More than half of the kills happen inside -2 to 2 degree vertical angle.
- One could plot out "kill locations" what was demoed in BF4 kill distances post. Maybe a more refined visualization could show something interesting.
- Kill distances per weapon (class). Problem here is the possible lack of datapoints for some weapons, but this could show if weapons' have overlapping "kill ranges".
- Engagement distances / effective engagement distances. For example, we could track damage done instead of kills. This could be used to analyze weapons' effective range.
- Refining this kill-distance study by removing biasses caused by weapons and whatnot.
Credits:
elementofprgess - For doing a lot of the work behind all this. <3!
shellBullet at unknowncheats.me - For providing info that made our job easier.
Update 7.6.2017: Distance per class
Thanks to list provided by lope_a_dope we now can calculate kill-distance statistics per class. Here are summary statistics and density plots:
Class Mean Med 75th 90th Sidearm: 9.0 7.0 11.8 17.4 Vehicler: 11.0 8.2 13.1 22.4 Assault: 11.4 8.9 14.8 22.0 Support: 23.1 16.4 30.2 51.0 Medic: 28.2 19.3 36.9 60.8 Pickups: 52.3 14.8 50.1 176.5 Recon: 68.4 54.2 88.6 140.7
Class "na" can be ignored, as it contains all non-infantry weapons and gadgets. Sorry for the confusing colors!
Sidearms and vehicle-r (pilot and tanker) weapons kill most at close range, which is to be expected given weapon types (pistols, small SMGs). These are followed by Asssault class, which is also focused to close range (SMGs, Shotguns).
Medic (semi-auto rifles) and Support (LMGs) share very similar ranges in both summary statistics and density plot. It would be interesting to know what kind of differences there are between engagements of these two, as their weapons differ quite a bit.
Recons have flat density curve compared to other weapons, soft peak around 30-50m. This flatness could be partly explained by how some snipers have high damage at ranges 40m-120m, but some start with highest damage like rest of the weapons.