Amizaur

Thanks for the reply. I have to agree that "far from perfect" was too hard word. "Good but not perfect and having strange issues sometimes" - would be more fortunate. I hope you will continue work on improving it and eliminating those rare, strange cases, which frustrate many of us . Especially I would love to see that units waiting in "ambush" - sitting still and watching carefully some road turn or top of a hill, at least those with an cover arc, would have MUCH higher chances of spotting first and firing first on something that shows up in center of their cover arc, than that poor moving unit which shows up on this watched road or crosses that watched hill top. CHANCES is the keyword here - I don't expect that ambushers would always be first, just that that have an advantage and USUALLY spot first. Currently in my experience ambushes - sitting still and expecting that the enemy will show in some place, with guns directed - do not give much advantage over a moving unit. Very often this is the moving unit that spots first, fires first and kills the ambusher. I would also love to see that units (like tanks) that are surprised by a fire from the side, take a bit longer to spot the shooter - and do not start responding to the fire after few miliseconds from taking it, as I experience now. Flanking and attacking from the side doesn't seem to give so much advantage as in real life - after the first shot, the attacked tanks in CMBN seem to instantly spot the shooter, they quickly rotate their turrets and - if they were not killed by the first shot, which is rare - they manage to face the enemy and exchange next salvos on equal terms. In real life tanks attacked from the flank would be USUALLY blind and disoriented for several seconds - many formations of tanks that were attacked from the side (in Normandy, too) died never realising what killed them and where the threat was. Especially green and regular units should usually need several seconds to even locate the enemy, and then to respond to the fire. Sharing that info between tanks in formation should not be instant, too. This is what bothers me personally most, in the spotting system. Regards

I know that doing 1000 sample test is very time consuming. I did such testing, too. Your test is great and respect for you that you are testing things at all, trying to check/verify various things and recreate game mechanics. Only I want to say, that 100 samples is - unfortunately - not enough to draw conclusions from them. The differences in results up to about 10% may be as well something we look for, as a pure coincidence. On the other hand, if we were looking for something that gives a 30% difference, then a test of even 100 samples can show it. Fortunately, if 10 people using the same scenario done a 100-sample test, the results could be added up and treated as single 1000 sample test. Making 100 sample test is not that much time consuming, it's quite fun. Making 1000 sample test is really boring - so maybe few people would like to participate, to make a group of testers, which would do 100-200 sample tests that combined together would be 1000-2000 sample tests ? You would only make a good scenario and describe the methodology of writing down the results. I could participate sometimes, too. Let's see if 1000 sample test is enough - or rather, what kind of accuracy can we get with it? I modify my program to do a serie of 1000 sample tests, let's see... Results: --------------------------------------------- PercentOfNoDamage = 5.7 PercentOfSpalling = 40.1 PercentOfPartialPen = 50.4 PercentOfPenetracji = 3.8 --------------------------------------------- Results: --------------------------------------------- PercentOfNoDamage = 7.2 PercentOfSpalling = 36.4 PercentOfPartialPen = 53.3 PercentOfPenetracji = 3.1 --------------------------------------------- Results: --------------------------------------------- PercentOfNoDamage = 6.7 PercentOfSpalling = 39.8 PercentOfPartialPen = 49.7 PercentOfPenetracji = 3.8 --------------------------------------------- The "true" (expected) values were: NoDamage chance = 7 Spalling chance = 38 PartialPen chance = 52 Penetration chance = 3 As we can see, now the results are usually within ~+/-2% from the true value. The smaller numbers (NoDamage and Penetration) lie even within +/-1%. I did 20 rolls, the results spreaded a little more, within 2...3%. That means that with 1000 samples we have 0.95 certainity that the error in our test is less than about 3% (2% for small numbers). With 100 samples we have 0.95 certainity that the error is less than about 10% (5% for the smaller numbers). I hope this estimation could help in deciding what number of samples is needed in a test to get significant results.

edit: I can't find correct formulas to calculate what number of samples would be needed to get results with specified accuracy and certainity. But I wrote a simple basic program that simulated your 100 sample test. The program just randomised a number between 0 and 100, and decided if that number means penetration, partial, spalling or no damage. The threeshold values were set in a way: Penetration 97 ... 100 PartialPen 45 ... 97 Spalling 7 ... 45 NoDamage 0 ... 7 ... to get chances for various events shown below: Chance4NoDamage=7 Chance4Spalling=38 Chance4ParialPen=52 Chance4Penetration=3 The above was repeated 100 times and the summarical results were printed. Here are results of just 3 runs of that program: PercentOfNoDamage = 4 PercentOfSpalling = 31 PercentOfPartialPen = 62 PercentOfPenetracji = 3 PercentOfNoDamage = 3 PercentOfSpalling = 35 PercentOfPartialPen = 58 PercentOfPenetracji = 4 PercentOfNoDamage = 6 PercentOfSpalling = 43 PercentOfPartialPen = 49 PercentOfPenetracji = 2 As you see the random error of measurement is quite big, the percent of Partialenetrations you get with 100 sample test may be as well 49 as 62 and this is pure statistical fluctuation in a random proces. Unfortunately 100 samples is way too little data to get reliable results and draw any conclusions from it. 100 samples is enough to notice big difference, like between let's say 20% and 60%, but with 100 samples one cannot see a small difference between for example 55% and 65%. The measurement error is too big. It would take at least 1000 samples of data to see such small differences in probabilities.

I'm not sure what the conclusion is. Is the shatter gap modelled, or not ? In a similar test done at LONGER range, would be the proportion of penetrating hits HIGHER that from 500m test ? Because this is how the shatter gap should work. My second though is - I wonder what it an average expected variation of results between two identical tests done with 100 samples, if chances of various events are for example 7%, 35%, 52% and 3% ?. I feel it may be well several percents. So, unfortunately, it may be just coincidence, and if the tests were done with for example 1000 samples, the results would be almost the same. I'll try to remind/dig out some statistical formulas and calculate how much samples are needed to get results accurate to xx% with xx% certainity.

I believe that there are no bugs in LOS system that could be fixed. The system works as intended. I would rather say, that the design of the LOS system is far from perfect. Sometimes there may me - technically - a LOS through the in-game objects (tree models, buildings, etc). The game finds it and units spot one another or take shots (if the gunner happens to have a LOS too). But as someone said, the "seeing" is happeing in the brain. Being able to _technically_ (by game engine means) SEE (trace to) for few seconds small part of something (a tank, a soldier) - from few hundred meters, trough only a small narrow gap in view-blocking map objects (like tree foliage), in foggy weather or against the sun, for few seconds - doesn't mean this thing would be spotted, identified and attacked by a soldier on a battlefield. It COULD happen sometimes, but rarely, in specific situations. The eye has to look in the right direction and the right moment. The brain has to recognise the (partialy seen) pattern from the clutter. It's much less probable than just "having a technical LOS" because there is a 1 meter gap for viewing (enough for a narrow laser beam to reach the target). I wonder, how the spotting mechanics works. How many "spotting points" a vehicle has, and how many of those points has to be "seen" for the unit to be spotted and identified depending on range, visibility, cover. If only a single "spotting point" has to be "seen" trough small narrow gap in a game 3D world, for the object to be spotted, identified and attacked, then this is not the optimal solution IMO. The probability of detection and IDENTIFICATION should depend on many factors, including "how big part of an object am I seeing" one. Does it work this way ?

I'm saving my money for something that has T-34s and IS-2s

"Yep. 76(W) Shermans have significantly better damage control than many other models of Sherman, so the chances of a non-killing penetration are higher" Anyway, single penetration from Panther's high velocity gun in Normandy enviroments (even without any serious damage and injuries, which is unlikely, but even with only pure light-and-sparks-and-sound-and-smoke type inside effects) - would cause MOST of crews to jump out of the tank - from psychological reasons. Single penetration means that someone has them in sight and the next penetrating round is just being loaded, and is going to arrive in seconds... And - in Normandy enviroments - it would work the other way, too. A Panther getting a surprise penetration, from unseen enemy, in 80% of cases would the mean crew bailing out as fast as they can, and only then cooling down and wondering what was that. NOT staying in the tank and trying to figure out where the enemy is and who's shooting. They may stay inside if they know the threat, think it's not that bad and they don't expect any further penetrations. Or if they are SO shocked by the explosion and gore effects (some beheaded crewmembers or something like that) that - even if uninjured - they are paralyzed and unable to act on their own, for some moments. That happened too. Or they are extremaly well motivated, that happened too... But even SS crews mostly bailed out if surprised. It's an instinct of self preservation. Unfortunately, tank crews in CM seem to not have any self-preservation instinct. As long as the vehicle is working, and they are not panicked, they would stay and fight - very effectively. Just like nothing had happened. Many times I've seen them killing the threat that has penetrated their tanks few seconds before, but didn't manage to reload and penetrate it again, before the retaliation came. Maybe tank crews should panic more easily.... They are a bit too "robotic" or terminator-like. A tank that has spotted an enemy SHOULD NOT immediately start rotating the turret in that direction, aim and engage like a terminator....!!! There is a HUMAN MULTI-CREW inside, a team, that is unable to ALWAYS act optimally and in shortest possible time. It DOESN'T work like that in real life. Same for things like reaction times, spotting times, reload times, engage times, time to start moving - they should be a lot more random in my opinion. Anyone who been there, or have read some war diaries, or watched some authentic documentary films, can notice that. Only very well trained and experienced crews in optimum conditions can be close to such "optimum" performance and act almost always as quickly as possible, almost like a robot. Most of the "human" or "soldier" performance is far from that. Such semi-random variability is quite easy to code, so I hope we will see it someday in CM.

I second that . He heh ! Now it seems that people who insisted that the lowered accuracy IS A BUG were actually... right... ? And those who defended the state of things and treated it as a "feature" - were actually... well... somewhat... wrong ? .

And that should be remembered! Pivoting in place in a field, even more at battle - was dangerous, rarely used, and probably prohibited - could easily throw off a track, or clog the suspension with dirt/mud. It could also just break something in the drive train - for example the final transmission gears, because pivoting in place strained the mechanisms. A breakdown like that would often mean a lost tank in 1944/45. And at last - pivoting in place was SLOW. It would be much faster to just move forward - even for few meters - gain some speed and then turn sharply. Pivoting would be used only, if for some reasons it was not possible to turn while moving (lack of place). It was not "new cool way of turning". Just a rarely used feature of the Panther's gearbox, with lot's of restrictions and risks connected with it's use in terrain other than a flat concrete/stone surface.

I'll try to make tomorrow test with tanks actually moving vs stationary and paused. Only about 50 shots to get general idea of that is going on - not precise statistically meaning results. About the Panther gunner having no 1x periscope and resulting long target aquision times - well, I would think that Germans developed some kind of crew communication / training / tactic to use a tank with such set of observation devices effectively. Such layout was worse than in Sherman, but probably Germans get used to this, trained to get best of it - some kind of good coordination between commander and gunner - because they were using Panthers quite effectively. But only well trained and experienced crews were really effective. Maybe Germans would be even more effective if the gunner had a periscope - in some cases and situations (like close quarter battles) Panther was at big disadvantage against Shermans, but generally Germans were not doing bad with it . It's possible that allied tankers testing the Panther tried to use it in a way they used american tanks, they way they trained and were used to. It didn't work too well and it took the gunner 20-30 seconds to acquire the target and take the shot. If there was a trained veteran German crew instead, maybe they would acquire the target and take the shot in 10 seconds. If they trained it for a month or so, and developed a way to use it effectively - they would be quicker. On the other hand, an inexperienced and poorly trained crew - like Germans had at the end of the war - would be not good, they could be as slow as the "test" crews - 20-30s - or even slower. Inexperienced and/or poorly trained crews would be much more effective in Sherman, than in Panther. The conclusion would be, that Panther needed very good crew communication and coordination so the commander could quickly and effectively guide the gunner to the target. Only well trained and experienced crews could fight effectively in Panther. Germans had such crews in 1941-1943 so maybe they didn't notice much problem with observation devices, tanks were effective anyway. But they didn't have much experienced (and trained) crews in 1944-45.... Some improvements were planned for late German tanks (like periscopic sights replacing telescopic sights, and maybe observation periscopes) but they didn't see production lines. In the Sherman, the gunner could more easily locate and acquire targets on it's own, it didn't need so much help from the commander.

I'm also not sure if I understand correctly the test results, Vanir post #142 - were the tanks actually moving ? What was the range ? Was it the same like in first test, so 300m ? The conclusion is that Move order and Fast order have both the same penalty ? post #143 - the Cromwell and Sherman were actually not moving, with move & pause order. The conclusion is that test shows no sign of any gyrostabiliser effect for Sherman, right ?

Good point. Flak 36 should turn much faster than even lighter AT guns. It has a comfortable AA mount. Fast turn rate could at least partially compensate it's bigh size and height.

Then the current state of "shooting accuracy while moving or paused movement" has a room for improvement. You think current state is good, I think it's bad, we'll both agree that it can be better.

I start to think that JonS doesn't exist. It's a BF company under this name, posting on forums Every thing that is in the game, is good at present state of things. But if it turns out to be wrong and be corrected in next patch, that will be good too . "... unless the choice is the full movement penalty, or no penalty at all." So you believe the current state of things is THE only available solution or the best compromise ? That it CANNOT be programmed to have no penalty while paused while retaining the penalty while moving, or to have 1/5 of movement penalty when paused, or the penalty depending on the speed of movement & lenght of pause, or anything like that ? Those men coded the WHOLE GAME, and you really think they can't code better solution for such a simple problem than that? I think it's just a bug, something that was simply omited, very easy to fix and w'll see it fixed in the next patch.

Maybe a tank that is just going to start moving should be a little bit less accurate than a fully stationary and "not going anywhere soon" tank. (If this is worth to reflect this in the game, it's another question). But some of you guys seem to forget about the proportions of the things you discuss. Because the accuracy of "starting moving soon" tank should not be FIVE TIMES WORSE than for fully stationary tank !! We are talking here about difference of 6% misses and 30 percent misses FROM 300m !! Accuracy in second case is 5 times worse than the first. 70% of accuracy from 300m ? I would expect such accuracy from a tank that just stopped moving and is firing a shanpshot, not from stationary tank firing shot after the shot from the same position against same target at 300m - even if the tank is going to move soon. The difference is WAY to high to reflect the hipothetical psychological effects that a "soon we are moving" command could have on the gunner.

Maybe like - "Fast move - don't stop when detecting a target" and "Quick move - stop and engage after detecting a target". Or "Quick - don't stop" and "Move - stop and engage". Combined with "Target Arc" and "Armor Target Arc" settings this could also handle our preferences what kind of target are we interested in. I mean - should it stop to engage every target, or maybe only for armored threats and AT guns ? Add some fuzzy logic to handle exeptions like AT team close range ect .

When I think about accuracy of fire on the pause, then: 1). if the tank is actually moving and then stops, there should some (gradually decreasing) penalty to the accuracy for few seconds - reflecting the time it takes for suspension to stabilise the tank. After that time, the accuracy should be at stationary level. The time it takes for the accuracy to change (gradually) from "moving level" to "stationary level" could depend on the speed of the tank before it stopped. I mean - if a tank moved at slow, it could take only 2 seconds after stopping to achieve maximum accuracy. If the tank moved fast, then it could be longer like 5-6 seconds to achieve full accuracy. 2). let's assume a tank is firing few shots. The first shot is not very accurate, but each shot after that increases it's accuracy (against that target). Now - each time the tank moves, the accuracy should be "reset" to the level of "first shot". So every "fire from short stop" would work like first shot. I believe it already actually works this way in CM. P.S. Sgt Joch - there is no need to "fine tune the whole engine" - it's enough to use some simple smart alghoritm that emulates real life. Like the time it takes for suspension to stabilise the tank. Even VERY simple alghoritm (like 3 lines of code) would give MUCH better results than simple ON/OFF, or - worse yet - ALWAYS ON or ALWAYS OFF.

I hope it's just a bug. Really not easy to notice one. Good that you caught it !

We are talking about CM here, not UFOs, ideologies ect. I found many of John's posts interesting and informative, his opinion on few (CM related) subjects is similar to mine. So I do care.

I care. (10 characters)

The "weapon hits" are always shown on the gun barrel, in case of the Tiger there are two "hitboxes" - one is the muzzle brake, second is the place where the diameter of the gun tube increases. Every "weapon hit" is shown as explosion in one of those two places. On the other hand, I could swear that sometimes the hit locations were shown in wrong places... It would be SO EASIER if we could just ask how is it coded... and help to improve that system.... I remember years ago being betatester for Dangerous Waters sim, when there was a bug in sonar propagation model. But the company said there can't be a bug because they have the knowledge, they have experts, and they are working for military too, so their formulas are ok. They didn't said any details of the sonar model. So I did series of tests collecting statistical data, plotting graphs and recreating the formulas they have used from that data, then finding the bug on the graphs, and at last recreating the bug in the formula (where spherical spreading of sound was switched to cylindrical spreading: logx^2 is NOT the same as log(x^2) ;P) that produced exactly same kind of bad results. Some other guy knew a bit about assembler and found another way to rectreate the formulas and has found the bug independanty. Only then they... no, they didn't acknowledged that. Just fixed it quietly in next patch... Life would be SO easier, if people and companies could cooperate a bit more....

76-85mm shell fired at 800m/s to a range of 2000m and arriving there with terminal velocity of 600m/s has a trajectory height over the aiming line of about 10m (a parabolic trajectory that is 10m high and 2000m long) and a descent angle of about 1deg 20min, maybe 1.5deg. That's really not much. Don't know what is terminal velocity of US 76mm shell at 2000m but wouldn't expect descent angle greater than 1.5-2deg. I would not expect much of "trajectory effects" here... P.S. The theory about commander's cupola hits being counted as "front turret" hits turns out to be true... Tested specifically for that in RT: But I also seen hit exactly like that being shown as "front left turret"... Also seen hits almost in center of the mantlet shown as "front turret hits" I tried to put the Tiger tank half behing a stone building, so the one half of the turret with the cupola would be impossible to hit. And check, if I would see any "front turret hits". Unfortunately, if a tank is half-hided behind a building, it also can't be spotted by the Shermans..... When I moved Tiger a bit in a way that it's center of mass was just visible from behind of the building, it was spotted and targeted by the Sherman but it's cupola was visible too and it was hit several times, one of the hits (catched by pausing the RT game) is shown on the screenshot.

Could you please drop also the scenario file of your test ? I played the saved game but could not modify it. So the first thing I did at the start was to rewerse the Tigers from 1000m range to 2000m range. They were not hit during reversing (the dust raised by rewerse movement was enough to hide them from Sherman's view). Then I counted hits from 2000m. I played several turns and counted up to 138 hits - not as much as you did, but I don't have time for more... I'm also not sure if that was hull down or not, as the raise in the ground was now at 1000m range when Tigers were reversed to 2000m - so probably there was not much hull-down. The results of 138 hits at 2000m are: hit / % hits Front lower hull /5 /3,62% Front upper hull /2 /1,45% Superstructure front hull /21 /15,22% Front turret /35 /25,36% Left/right front turret /23 /16,67% Weapon mount /43 /1,16% Weapon /6 /4,35% Top turret /3 /2,17% Top hull /0 /0,00% skirt /0 /0,00% There were 8 spallings and 1 partial penetration from superstructure front hits, and 6 spallings with 2 partial penetrations (one turned the tank to destroyed state) from "front turret hits", there was total one casualty (after one of spalling hits). Weapon Mount / Front turret ratio: 1000m - 1.31 (your results) 2000m - 1.23 (my results) As the spread pattern gets wider, the number of left/right front turret hits increases, and the ratio of WeaponMount / Front turret hits seem to decrease. But what surprises me, is that there is ONLY 20% of HULL HITS from 2000m range, with supposedly not-too-much hull down setup...

Maybe that is because of 500m shooting range ? The aiming of the gunner is so good and the circle of spread so tight, that very rarely any shot lands on the roof or on the sides of the turret ? Trying the same test on 1000-1500m should generate much more "side turret" hits and give a much more uniform hit distribution pattern along the front turret area. Simply, the spread circle would be 2-3 times larger, covering more than the whole turret area.

There were no "side turret" hits ? Tiger turret is round, and the turret side is visible from front aspect... Maybe the somewhat high number of "turret front" hits comes from side turret hits being counted as front turret hits ? Second thought: so we have 22% of hits landing on a Tiger's turret disabling it's main gun ? (22.7% -- weapon) Isn't it a bit high number, considering the relative area of the muzzle brake (that is "collecting" almost all weapon hits) compared to the area of the whole turret ?