Accuracy vs TU Efficiency - Revision history

Ant 222 at 01:08, 18 September 2016

2016-09-18T01:08:05Z

← Older revision		Revision as of 01:08, 18 September 2016
Line 86:		Line 86:

	:[[User_talk:Bomb_Bloke:Firing_Accuracy\|Here]]. And as I said, "misses" actually hit fairly often at close range, and "hits" failing to hit isn't unheard of. [[User:Magic9mushroom\|Magic9mushroom]] ([[User talk:Magic9mushroom\|talk]]) 22:20, 2 September 2016 (UTC)		:[[User_talk:Bomb_Bloke:Firing_Accuracy\|Here]]. And as I said, "misses" actually hit fairly often at close range, and "hits" failing to hit isn't unheard of. [[User:Magic9mushroom\|Magic9mushroom]] ([[User talk:Magic9mushroom\|talk]]) 22:20, 2 September 2016 (UTC)

			Thanks for the link. The problem with the algorithm is that two of its three stages are physically meaningless. First—the random choice of the equation for the divergence factor, and second—its high dependency on an another random number within the chosen equation. For any given combination of the firer's stats, the weapon's stats, and shot type divergence should be strictly constant, and the very generation of a single angle distributed normally with the aforesaid divergence (which I take to be a measure of dispersion) should be the only source of randomness, whereas that algorithm severely diminishes the role of the normal-distribution model by introducing way too much randomness in the determination of its parameters.

			Nobody seems to have understood that what [[User:Bomb Bloke]] so painstakingly logged is actually direct samplings of the cumulative distribution function of angular deviation. He need only plot his data correctly to see that.[[User:Ant 222\|Ant 222]] ([[User talk:Ant 222\|talk]]) 01:08, 18 September 2016 (UTC)

	=Data Analysis=		=Data Analysis=

Magic9mushroom: /* Discussion */

2016-09-02T22:22:00Z

Discussion

← Older revision		Revision as of 22:22, 2 September 2016
Line 85:		Line 85:
	Is the algorithm that you have described documented anywhere? [[Accuracy_formula]] does not mention it and seems to imply a single dispersion cone per shot...[[User:Ant 222\|Ant 222]] ([[User talk:Ant 222\|talk]]) 17:08, 2 September 2016 (UTC)		Is the algorithm that you have described documented anywhere? [[Accuracy_formula]] does not mention it and seems to imply a single dispersion cone per shot...[[User:Ant 222\|Ant 222]] ([[User talk:Ant 222\|talk]]) 17:08, 2 September 2016 (UTC)

	:[~~http://ufopaedia.org/index.php/~~User_talk:Bomb_Bloke:Firing_Accuracy\|Here]. And as I said, "misses" actually hit fairly often at close range, and "hits" failing to hit isn't unheard of. [[User:Magic9mushroom\|Magic9mushroom]] ([[User talk:Magic9mushroom\|talk]]) 22:20, 2 September 2016 (UTC)		:[[User_talk:Bomb_Bloke:Firing_Accuracy\|Here]]. And as I said, "misses" actually hit fairly often at close range, and "hits" failing to hit isn't unheard of. [[User:Magic9mushroom\|Magic9mushroom]] ([[User talk:Magic9mushroom\|talk]]) 22:20, 2 September 2016 (UTC)

	=Data Analysis=		=Data Analysis=

Magic9mushroom: /* Discussion */

2016-09-02T22:20:54Z

Discussion

← Older revision		Revision as of 22:20, 2 September 2016
Line 84:		Line 84:

	Is the algorithm that you have described documented anywhere? [[Accuracy_formula]] does not mention it and seems to imply a single dispersion cone per shot...[[User:Ant 222\|Ant 222]] ([[User talk:Ant 222\|talk]]) 17:08, 2 September 2016 (UTC)		Is the algorithm that you have described documented anywhere? [[Accuracy_formula]] does not mention it and seems to imply a single dispersion cone per shot...[[User:Ant 222\|Ant 222]] ([[User talk:Ant 222\|talk]]) 17:08, 2 September 2016 (UTC)

			:[http://ufopaedia.org/index.php/User_talk:Bomb_Bloke:Firing_Accuracy\|Here]. And as I said, "misses" actually hit fairly often at close range, and "hits" failing to hit isn't unheard of. [[User:Magic9mushroom\|Magic9mushroom]] ([[User talk:Magic9mushroom\|talk]]) 22:20, 2 September 2016 (UTC)

	=Data Analysis=		=Data Analysis=

Ant 222: /* Discussion */

2016-09-02T21:38:17Z

Discussion

← Older revision		Revision as of 21:38, 2 September 2016
Line 83:		Line 83:
	I fear that we disagree about the term ''dispersion cone''. I understand it as the normally distributed direction of the projectile, so that each shot is modeled by a single dispersion cone, which describes all the possible outcomes with their corresponding probabilities. This interpretation leaves no place for two dispersion cones. If UFO:EU chooses the wider or the narrower of some two dispersion cones depending on accuracy, then it largely defeats the cause of the normal model with respect to the non-linear dependency on distance, for the choice of either the ''hit'' or the ''miss'' dispersion cone is the definitive factor, whereas the subsequent modeling with the chosen cone is only secondary, because the ''hit'' cone almost always results in a hit and the ''miss'' one in a miss.		I fear that we disagree about the term ''dispersion cone''. I understand it as the normally distributed direction of the projectile, so that each shot is modeled by a single dispersion cone, which describes all the possible outcomes with their corresponding probabilities. This interpretation leaves no place for two dispersion cones. If UFO:EU chooses the wider or the narrower of some two dispersion cones depending on accuracy, then it largely defeats the cause of the normal model with respect to the non-linear dependency on distance, for the choice of either the ''hit'' or the ''miss'' dispersion cone is the definitive factor, whereas the subsequent modeling with the chosen cone is only secondary, because the ''hit'' cone almost always results in a hit and the ''miss'' one in a miss.

	Is the algorithm that you have described documented ~~somewhere~~? [[Accuracy_formula]] does not mention it and seems to imply a single dispersion cone per shot...[[User:Ant 222\|Ant 222]] ([[User talk:Ant 222\|talk]]) 17:08, 2 September 2016 (UTC)		Is the algorithm that you have described documented anywhere? [[Accuracy_formula]] does not mention it and seems to imply a single dispersion cone per shot...[[User:Ant 222\|Ant 222]] ([[User talk:Ant 222\|talk]]) 17:08, 2 September 2016 (UTC)

	=Data Analysis=		=Data Analysis=

Ant 222: /* Discussion */ Added a comment

2016-09-02T17:08:55Z

Discussion: Added a comment

← Older revision		Revision as of 17:08, 2 September 2016
Line 80:		Line 80:

	:UFO:EU does actually use dispersion cones, FYI. The displayed accuracy of the shot is the chance that it will use the narrower "hit" dispersion cone instead of the wider "miss" dispersion cone, but "hits" can miss (usually due to cover, such as trying to hit an alien in a building through a window, but at long ranges true misses are possible against small targets) and "misses" can hit (most obviously at point-blank range). The size of the cones is also adjusted based on that accuracy, so the dispersion of a "miss" from a shot with 10% displayed accuracy is much larger than that of a "miss" from a shot with 99% displayed accuracy. So the "improvement" you want is already there to some extent. [[User:Magic9mushroom\|Magic9mushroom]] ([[User talk:Magic9mushroom\|talk]]) 00:58, 2 September 2016 (UTC)		:UFO:EU does actually use dispersion cones, FYI. The displayed accuracy of the shot is the chance that it will use the narrower "hit" dispersion cone instead of the wider "miss" dispersion cone, but "hits" can miss (usually due to cover, such as trying to hit an alien in a building through a window, but at long ranges true misses are possible against small targets) and "misses" can hit (most obviously at point-blank range). The size of the cones is also adjusted based on that accuracy, so the dispersion of a "miss" from a shot with 10% displayed accuracy is much larger than that of a "miss" from a shot with 99% displayed accuracy. So the "improvement" you want is already there to some extent. [[User:Magic9mushroom\|Magic9mushroom]] ([[User talk:Magic9mushroom\|talk]]) 00:58, 2 September 2016 (UTC)

			I fear that we disagree about the term ''dispersion cone''. I understand it as the normally distributed direction of the projectile, so that each shot is modeled by a single dispersion cone, which describes all the possible outcomes with their corresponding probabilities. This interpretation leaves no place for two dispersion cones. If UFO:EU chooses the wider or the narrower of some two dispersion cones depending on accuracy, then it largely defeats the cause of the normal model with respect to the non-linear dependency on distance, for the choice of either the ''hit'' or the ''miss'' dispersion cone is the definitive factor, whereas the subsequent modeling with the chosen cone is only secondary, because the ''hit'' cone almost always results in a hit and the ''miss'' one in a miss.

			Is the algorithm that you have described documented somewhere? [[Accuracy_formula]] does not mention it and seems to imply a single dispersion cone per shot...[[User:Ant 222\|Ant 222]] ([[User talk:Ant 222\|talk]]) 17:08, 2 September 2016 (UTC)

	=Data Analysis=		=Data Analysis=

Magic9mushroom: /* Discussion */

2016-09-02T00:58:46Z

Discussion

← Older revision		Revision as of 00:58, 2 September 2016
Line 79:		Line 79:
	That being said, I don't think serious improvement is possible unless someone can change the hit-chance formula so that it will approximate, if loosely, the varying effect of distance at different ranges described above. As distance increases the effectiveness of auto- and snap shots should become lower compared to that of an aimed shot. [[User:Ant 222\|Ant 222]] ([[User talk:Ant 222\|talk]]) 15:28, 1 September 2016 (UTC)		That being said, I don't think serious improvement is possible unless someone can change the hit-chance formula so that it will approximate, if loosely, the varying effect of distance at different ranges described above. As distance increases the effectiveness of auto- and snap shots should become lower compared to that of an aimed shot. [[User:Ant 222\|Ant 222]] ([[User talk:Ant 222\|talk]]) 15:28, 1 September 2016 (UTC)

			:UFO:EU does actually use dispersion cones, FYI. The displayed accuracy of the shot is the chance that it will use the narrower "hit" dispersion cone instead of the wider "miss" dispersion cone, but "hits" can miss (usually due to cover, such as trying to hit an alien in a building through a window, but at long ranges true misses are possible against small targets) and "misses" can hit (most obviously at point-blank range). The size of the cones is also adjusted based on that accuracy, so the dispersion of a "miss" from a shot with 10% displayed accuracy is much larger than that of a "miss" from a shot with 99% displayed accuracy. So the "improvement" you want is already there to some extent. [[User:Magic9mushroom\|Magic9mushroom]] ([[User talk:Magic9mushroom\|talk]]) 00:58, 2 September 2016 (UTC)

	=Data Analysis=		=Data Analysis=

Ant 222: Added the attribution

2016-09-01T15:28:13Z

Added the attribution

← Older revision		Revision as of 15:28, 1 September 2016
Line 77:		Line 77:
	Precise aiming takes a certain time determined by the weapon and the firer, after which accuracy reaches a plateau and then starts to degrade due to fatigue. During the effective aiming stage the rate of accuracy increase is highest at the beginning and becomes lower at it approaches the plateau, but in the end the probability of defeating a remote target is improved out of all proportion with regard to the time spent aiming.		Precise aiming takes a certain time determined by the weapon and the firer, after which accuracy reaches a plateau and then starts to degrade due to fatigue. During the effective aiming stage the rate of accuracy increase is highest at the beginning and becomes lower at it approaches the plateau, but in the end the probability of defeating a remote target is improved out of all proportion with regard to the time spent aiming.

	That being said, I don't think serious improvement is possible unless someone can change the hit-chance formula so that it will approximate, if loosely, the varying effect of distance at different ranges described above. As distance increases the effectiveness of auto- and snap shots should become lower compared to that of an aimed shot.		That being said, I don't think serious improvement is possible unless someone can change the hit-chance formula so that it will approximate, if loosely, the varying effect of distance at different ranges described above. As distance increases the effectiveness of auto- and snap shots should become lower compared to that of an aimed shot. [[User:Ant 222\|Ant 222]] ([[User talk:Ant 222\|talk]]) 15:28, 1 September 2016 (UTC)

Ant 222: Edited my comment in the Discussion section

2016-09-01T15:27:33Z

Edited my comment in the Discussion section

← Older revision		Revision as of 15:27, 1 September 2016
Line 68:		Line 68:
	The second reason is that snap shots are more effective at close range and aimed ones at long range. It happens because at short range the dispersion cone is much smaller than the target's angular area and aiming consists of pointing the weapon roughly at the target without the use of sights. The hit change is near 100% so the one who shorts first is the winner. At long range, however, the situation is reversed. The weapon must be aligned carefully to point at the small target and stabilized to narrow down the dispersion cone. The effect of distance is not uniform across all ranges but follows the behavior of the quantiles of the normal distribution:		The second reason is that snap shots are more effective at close range and aimed ones at long range. It happens because at short range the dispersion cone is much smaller than the target's angular area and aiming consists of pointing the weapon roughly at the target without the use of sights. The hit change is near 100% so the one who shorts first is the winner. At long range, however, the situation is reversed. The weapon must be aligned carefully to point at the small target and stabilized to narrow down the dispersion cone. The effect of distance is not uniform across all ranges but follows the behavior of the quantiles of the normal distribution:

	[[File:HitChanceVsDistance.png\|\|Hit chance vs. distance]]		[[File:HitChance1d.png\|\|\|center\|\|\|\|\|One-dimensinal formula for hit chance]]

			where Φ is the [https://en.wikipedia.org/wiki/Normal_distribution#Cumulative_distribution_function cumulative normal distribution function], ''d'' the distance to the target, and ''r'' the target's characteristic radius as projected in the firer's direction. The standard deviation σ determines the opening angle of the dispersion cone and thus the accuracy of the shot.
			Here are the graphs for three different accuracies:

			[[File:HitChanceVsDistance.png\|center\|Hit chance vs. distance]]

	Precise aiming takes a certain time determined by the weapon and the firer, after which accuracy reaches a plateau and then starts to degrade due to fatigue. During the effective aiming stage the rate of accuracy increase is highest at the beginning and becomes lower at it approaches the plateau, but in the end the probability of defeating a remote target is improved out of all proportion with regard to the time spent aiming.		Precise aiming takes a certain time determined by the weapon and the firer, after which accuracy reaches a plateau and then starts to degrade due to fatigue. During the effective aiming stage the rate of accuracy increase is highest at the beginning and becomes lower at it approaches the plateau, but in the end the probability of defeating a remote target is improved out of all proportion with regard to the time spent aiming.

	That being said, I don't think serious improvement is possible unless someone can change the hit-chance formula so that it will approximate, if loosely, the varying effect of distance at different ranges described above. As distance increases the effectiveness of auto- and snap shots should become lower compared to that of an aimed shot.		That being said, I don't think serious improvement is possible unless someone can change the hit-chance formula so that it will approximate, if loosely, the varying effect of distance at different ranges described above. As distance increases the effectiveness of auto- and snap shots should become lower compared to that of an aimed shot.
	~~[[User:Ant 222\|Ant 222]] ([[User talk:Ant 222\|talk]]) 11:58, 1 September 2016 (UTC)~~

Ant 222 at 11:58, 1 September 2016

2016-09-01T11:58:20Z

← Older revision		Revision as of 11:58, 1 September 2016
Line 63:		Line 63:

	: One additional benefit I can see for using the scout/sniper strategy other than the benefits already mentioned would be the fact that the sniper will not have to move as much, if at all. Therefore the sniper will have more TUs available to make aimed shots. Obviously this means nothing for some weapons that can only be fired once per turn on aimed, but for weapons that can be aimed several times, then every extra TU helps. Just as snaps and auto shots bank on extra attempts to improve their odds of success, the same can apply to aimed shots too. -[[User:NKF\|NKF]] 22:38, 12 November 2008 (CST)		: One additional benefit I can see for using the scout/sniper strategy other than the benefits already mentioned would be the fact that the sniper will not have to move as much, if at all. Therefore the sniper will have more TUs available to make aimed shots. Obviously this means nothing for some weapons that can only be fired once per turn on aimed, but for weapons that can be aimed several times, then every extra TU helps. Just as snaps and auto shots bank on extra attempts to improve their odds of success, the same can apply to aimed shots too. -[[User:NKF\|NKF]] 22:38, 12 November 2008 (CST)

			In real combat, both quick and carefully aimed shots are useful depending on the situation, for at least two reasons. In situations such as an ambush, when the target is unaware of the firer, it is beneficial to hit it with the first shot, regardless or how long it may take to aim precisely, because after that the firer's position, if only approximately, will be made known to the enemy, who will then be prepared for fire from the same quarter and will seek cover and possibly fire back. This consideration can hardly be accounted for in the classic X-COM (but must be in the games it inspired).

			The second reason is that snap shots are more effective at close range and aimed ones at long range. It happens because at short range the dispersion cone is much smaller than the target's angular area and aiming consists of pointing the weapon roughly at the target without the use of sights. The hit change is near 100% so the one who shorts first is the winner. At long range, however, the situation is reversed. The weapon must be aligned carefully to point at the small target and stabilized to narrow down the dispersion cone. The effect of distance is not uniform across all ranges but follows the behavior of the quantiles of the normal distribution:

			[[File:HitChanceVsDistance.png\|\|Hit chance vs. distance]]

			Precise aiming takes a certain time determined by the weapon and the firer, after which accuracy reaches a plateau and then starts to degrade due to fatigue. During the effective aiming stage the rate of accuracy increase is highest at the beginning and becomes lower at it approaches the plateau, but in the end the probability of defeating a remote target is improved out of all proportion with regard to the time spent aiming.

			That being said, I don't think serious improvement is possible unless someone can change the hit-chance formula so that it will approximate, if loosely, the varying effect of distance at different ranges described above. As distance increases the effectiveness of auto- and snap shots should become lower compared to that of an aimed shot.
			[[User:Ant 222\|Ant 222]] ([[User talk:Ant 222\|talk]]) 11:58, 1 September 2016 (UTC)

Hobbes: /* TU Efficiency Mods */

2013-07-23T02:53:16Z

TU Efficiency Mods

← Older revision		Revision as of 02:53, 23 July 2013
Line 127:		Line 127:
	* [[Media:OBDTU125.DAT]] fixes Aimed fire so it is 125% more efficient then Snap fire. This is the same ratio between Snap and Aimed fire that exists for Rockets.		* [[Media:OBDTU125.DAT]] fixes Aimed fire so it is 125% more efficient then Snap fire. This is the same ratio between Snap and Aimed fire that exists for Rockets.
	* [[Media:OBDTU150.DAT]] fixes Aimed fire so it is 150% more efficient then Snap fire. This is probably over-powerful, and too unbalancing in the other direction. Also, in some cases (Rifle and Plasma Rifle) it is necessary to reduce the TU cost of Aimed fire slightly, as otherwise Accuracy would need to be raised above 255% (not possible as it is held in a single unsigned byte).		* [[Media:OBDTU150.DAT]] fixes Aimed fire so it is 150% more efficient then Snap fire. This is probably over-powerful, and too unbalancing in the other direction. Also, in some cases (Rifle and Plasma Rifle) it is necessary to reduce the TU cost of Aimed fire slightly, as otherwise Accuracy would need to be raised above 255% (not possible as it is held in a single unsigned byte).

			[[Category:Enemy Unknown/UFO Defense]]