Posted By: mvp7
About gunner accuracy and precision - 10/26/19 07:14 PM
Rear gunners in WOFF are extremely deadly. Gunners in WW1 were indeed very dangerous but I think that their current efficiency is way too high at long ranges and high relative speeds.
For starters here's the difference between Accuracy and Precision:
In terms of air combat, estimating a correct lead for shots falls within Accuracy. It comes from the gunner's ability to estimate, observe and correct lead for their shots. This is far more important than precision and in practice, low precision actually compensates for less than perfect accuracy (See the two upper cases in the picture, low-low still hits the bulls-eye once).
Precision comes from the mechanical precision and zeroing of the weapon (both have fairly negligible effect in this case) and the shooter's ability to maintain a steady and consistent aim through a burst, which is the main factor for precision of WW1 gunner.
Currently when WOFF mentions "Accurate" and "Less Acc" in workshop, it seems to be referring to precision rather than accuracy. The actual accuracy of gunners seems to be nonadjustable and extremely high, especially for long ranges and high relative speeds.
I have some experience in these matters, having done my national service as an NCO in an Anti-Air battery of Finnish army. I have used 12.7mm NSV anti-air machine gun (12.7 ITKK 96) and modernized ZU-23-2 anti-air twin-autocannon (23 ITK 95) in training (the latter has an automatic lead calculator and is less relevant to this subject anyway).
PRECISION
First I'm going to get my thoughts on the WOFF gunner precision out of the way. When shooting at aircraft, precision is usually very unimportant factor compared to rate of fire and accuracy, AA weapon can actually suffer from being too precise as it would miss with perfect precision with anything less than perfect accuracy. With less than perfect accuracy (which it is most of the time) less than perfect precision will more likely produce at least some hits which is better than no hits.
Mechanical precision and sight zeroing a weapon would have almost negligible effect here. Even on the simple pintle mount, the effect of recoil would likely be fairly minor. The main challenges for precision would come from the gunner standing in the more or less unstable aircraft while being torn by the air flow. The gunner would constantly have to adjust his position and shoot at weird angles, struggling with the limited pintle angles and heavy gun-ring. In many cases it's unlikely that the gunner could maintain any kind of sight picture, not to mention a steady one, aiming mostly with the tracers instead.
Overall, I think the WOFF gunner precision (even with Less Accurate setting) is very optimistic in anything but the most optimal situation where the gunner gets to shoot at a relatively stationary target at very convenient angle (like an aircraft chasing the two seater on level and from behind for example).
ACCURACY
Accuracy is where I feel the gunners are being way too effective and where player has very limited ability to customize things. Shooting at moving target is generally really hard.
During training in the Army we were shooting the machine gun at a target RC plane: maybe 1.5 meters in length, flying maybe 200 meters away, sideways or towards the weapon on direct course at speed around 60 to 80 km/h I'd say. Even after dozens of passes and hundreds of shots from two machine guns, not one RC plane had been shot down. They actually hadn't even brought more than couple RC planes for a three day exercise because that's how unlikely they were to be destroyed (they weren't the only target though).
The distance, relative speed and angular speed are all factors that increase the difficulty of accurate shooting pretty much exponentially. The fact that the rear gunner is on a moving and maneuvering plane himself would add yet another variable to the equation. These are all things that the WW1 gunners would have to learn on the go while really being shot at by the real enemy.
Shooting at flying target is an endless process of trial and error. You make your initial (usually awful) estimate of the lead, shoot a burst and try to see where it misses. Just seeing where you missed quickly gets hard as the distance increases. In worst cases you don't even know which way you were missing. Based on those observations you correct your lead and shoot another burst. This is repeated until the target is destroyed or (more likely) until the target has left effective range.
What makes that process of trial and error even harder is that the variables are constantly changing: Distance, relative speed, directions and angular speed are all in constant flux and the more they change the less useful the trials become. In practice, outside of some particularly simple circumstances or very short ranges, shooting at quickly moving and/or distant target has very little chance of hitting it.
Observations from WOFF
Shooting at a target flying directly towards or next to you at hundred meters at slow relative speed would be fairly easy: i.e. you are likely to score a few hits and the target is likely to evade because it and hence shoot at you less effectively. In these short range situations the AI is good but not that outrageous. Low skill gunners should probably miss a bit more though.
Shooting at a distant moving target around 200-400 meters away is always tricky. It might not sound like that much but consider how far it would be on a shooting range and how unstable platform the plane would make. Even if the target is moving at low relative and angular speed, it's extremely hard to see and correct your fire. Accuracy starts to depend of very fine adjustments so precision also becomes an issue at this kind of ranges. Hits would be very rare and mostly thanks to bad precision. Landing too consecutive shots on the target would be a very lucky fluke. Shooting linearly moving distant targets like this with radar or laser lead calculator equipped weapon would be very easy but with just the limited capability of our eyes and brains its very ineffective. However the AI is highly accurate when shooting this kinds of targets.
Shooting at an enemy doing a hard banking turn at you side with the relative speeds, distance and angle speed all being in constant flux is very hard, especially if you consider the limited movement range of the gunner's pintle mount (you are not going to move the actual ring in a hurry). In this kinds of situations the AI can pull off some insane hits. When you are dog-fighting a Strutter or zooming and turning away from formation flying two seaters the AI will consistently land multiple hits on your plane in situations where a competent human gunner wouldn't even think of wasting the bullets.
Conclusions and recommendations
In my opinion the gunner AI in WOFF is maybe slightly too precise even with the 'Less Accurate' setting and definitely too accurate, especially with quickly changing large relative speeds and/or long distance, even when considering the relatively low speeds and distances of WW1 air war,
Having a dynamic gunner accuracy that improves with every burst and worsens when the variables are changing drastically would be great but tricky to code no doubt. To clarify, I don't mean just the inaccuracy that naturally comes from the AI making linear lead calculation against target that is not moving linearly. What I mean is an actual floating degree of accuracy that quickly get's better in static situations and similarly becomes more inaccurate when the situation is changing rabidly.
I suspect simulating "human" error and inaccurate perception like that is beyond what can be reasonably achieved with cfs3 engine though. Having more random errors in accuracy, and at the same time reducing the precision and maximum range of the gunners, would be a great as well. Basically anything that would prevent the AI gunner of a maneuvering plane from landing half of their first burst on another, distant, gently maneuvering plane even with "Less Accurate" gunners setting on.
For starters here's the difference between Accuracy and Precision:
In terms of air combat, estimating a correct lead for shots falls within Accuracy. It comes from the gunner's ability to estimate, observe and correct lead for their shots. This is far more important than precision and in practice, low precision actually compensates for less than perfect accuracy (See the two upper cases in the picture, low-low still hits the bulls-eye once).
Precision comes from the mechanical precision and zeroing of the weapon (both have fairly negligible effect in this case) and the shooter's ability to maintain a steady and consistent aim through a burst, which is the main factor for precision of WW1 gunner.
Currently when WOFF mentions "Accurate" and "Less Acc" in workshop, it seems to be referring to precision rather than accuracy. The actual accuracy of gunners seems to be nonadjustable and extremely high, especially for long ranges and high relative speeds.
I have some experience in these matters, having done my national service as an NCO in an Anti-Air battery of Finnish army. I have used 12.7mm NSV anti-air machine gun (12.7 ITKK 96) and modernized ZU-23-2 anti-air twin-autocannon (23 ITK 95) in training (the latter has an automatic lead calculator and is less relevant to this subject anyway).
PRECISION
First I'm going to get my thoughts on the WOFF gunner precision out of the way. When shooting at aircraft, precision is usually very unimportant factor compared to rate of fire and accuracy, AA weapon can actually suffer from being too precise as it would miss with perfect precision with anything less than perfect accuracy. With less than perfect accuracy (which it is most of the time) less than perfect precision will more likely produce at least some hits which is better than no hits.
Mechanical precision and sight zeroing a weapon would have almost negligible effect here. Even on the simple pintle mount, the effect of recoil would likely be fairly minor. The main challenges for precision would come from the gunner standing in the more or less unstable aircraft while being torn by the air flow. The gunner would constantly have to adjust his position and shoot at weird angles, struggling with the limited pintle angles and heavy gun-ring. In many cases it's unlikely that the gunner could maintain any kind of sight picture, not to mention a steady one, aiming mostly with the tracers instead.
Overall, I think the WOFF gunner precision (even with Less Accurate setting) is very optimistic in anything but the most optimal situation where the gunner gets to shoot at a relatively stationary target at very convenient angle (like an aircraft chasing the two seater on level and from behind for example).
ACCURACY
Accuracy is where I feel the gunners are being way too effective and where player has very limited ability to customize things. Shooting at moving target is generally really hard.
During training in the Army we were shooting the machine gun at a target RC plane: maybe 1.5 meters in length, flying maybe 200 meters away, sideways or towards the weapon on direct course at speed around 60 to 80 km/h I'd say. Even after dozens of passes and hundreds of shots from two machine guns, not one RC plane had been shot down. They actually hadn't even brought more than couple RC planes for a three day exercise because that's how unlikely they were to be destroyed (they weren't the only target though).
The distance, relative speed and angular speed are all factors that increase the difficulty of accurate shooting pretty much exponentially. The fact that the rear gunner is on a moving and maneuvering plane himself would add yet another variable to the equation. These are all things that the WW1 gunners would have to learn on the go while really being shot at by the real enemy.
Shooting at flying target is an endless process of trial and error. You make your initial (usually awful) estimate of the lead, shoot a burst and try to see where it misses. Just seeing where you missed quickly gets hard as the distance increases. In worst cases you don't even know which way you were missing. Based on those observations you correct your lead and shoot another burst. This is repeated until the target is destroyed or (more likely) until the target has left effective range.
What makes that process of trial and error even harder is that the variables are constantly changing: Distance, relative speed, directions and angular speed are all in constant flux and the more they change the less useful the trials become. In practice, outside of some particularly simple circumstances or very short ranges, shooting at quickly moving and/or distant target has very little chance of hitting it.
Observations from WOFF
Shooting at a target flying directly towards or next to you at hundred meters at slow relative speed would be fairly easy: i.e. you are likely to score a few hits and the target is likely to evade because it and hence shoot at you less effectively. In these short range situations the AI is good but not that outrageous. Low skill gunners should probably miss a bit more though.
Shooting at a distant moving target around 200-400 meters away is always tricky. It might not sound like that much but consider how far it would be on a shooting range and how unstable platform the plane would make. Even if the target is moving at low relative and angular speed, it's extremely hard to see and correct your fire. Accuracy starts to depend of very fine adjustments so precision also becomes an issue at this kind of ranges. Hits would be very rare and mostly thanks to bad precision. Landing too consecutive shots on the target would be a very lucky fluke. Shooting linearly moving distant targets like this with radar or laser lead calculator equipped weapon would be very easy but with just the limited capability of our eyes and brains its very ineffective. However the AI is highly accurate when shooting this kinds of targets.
Shooting at an enemy doing a hard banking turn at you side with the relative speeds, distance and angle speed all being in constant flux is very hard, especially if you consider the limited movement range of the gunner's pintle mount (you are not going to move the actual ring in a hurry). In this kinds of situations the AI can pull off some insane hits. When you are dog-fighting a Strutter or zooming and turning away from formation flying two seaters the AI will consistently land multiple hits on your plane in situations where a competent human gunner wouldn't even think of wasting the bullets.
Conclusions and recommendations
In my opinion the gunner AI in WOFF is maybe slightly too precise even with the 'Less Accurate' setting and definitely too accurate, especially with quickly changing large relative speeds and/or long distance, even when considering the relatively low speeds and distances of WW1 air war,
Having a dynamic gunner accuracy that improves with every burst and worsens when the variables are changing drastically would be great but tricky to code no doubt. To clarify, I don't mean just the inaccuracy that naturally comes from the AI making linear lead calculation against target that is not moving linearly. What I mean is an actual floating degree of accuracy that quickly get's better in static situations and similarly becomes more inaccurate when the situation is changing rabidly.
I suspect simulating "human" error and inaccurate perception like that is beyond what can be reasonably achieved with cfs3 engine though. Having more random errors in accuracy, and at the same time reducing the precision and maximum range of the gunners, would be a great as well. Basically anything that would prevent the AI gunner of a maneuvering plane from landing half of their first burst on another, distant, gently maneuvering plane even with "Less Accurate" gunners setting on.