Tiger spotting

[poesel]

I wanted to make a test to determine how tanks spot. But it's too HOT and SUNNY (Ahem! ) here and the data collection is very tedious, so I just made a video.

The setup is 1 Tiger vs. 3 Battalions of US infantry (minus weapons and HQs - so just the rifle squads) - each square holds one squad. All regular, fanatic, sunny, 12h, no wind, Tiger buttoned, infantry facing the tank. Each square is 8 AS wide so the whole thing is roughly 600x600m.

The video shows several interesting things about how spotting works.

First: spotting works in intervals. There is a 2-3 second offset at the beginning and then spotting occurs every 7 seconds. The offset is probably different for every unit on the field to avoid spotting 'peaks' which would overwhelm the CPU.

The offset can be seen by the delay until the first spotting occurs. After that changes only happen on the 7 second marks. Between that its static.

Second: during a spotting event all visible AS's get checked simultaneously. Easy to see when several units pop up at once.

Third: the Tiger has blind spots where he will never spot a unit. I was a bit surprised that a buttoned tank could see so much especially backwards. But I know nothing about the viewports of a Tiger. Someone here can possible tell me.

http://youtu.be/gqmyEzs1bxU

Next tests would be if sound or movement triggers spotting and influence of skill and light on spotting.

One explanation for some weird spotting events can already be drawn from this: if unit A just had its spotting event and unit B comes into sight and has now its own event then B has - worst case - 7 seconds time to shoot at A before A could even spot B.

[Baneman]

That's very interesting, thanks for this.

...

One explanation for some weird spotting events can already be drawn from this: if unit A just had its spotting event and unit B comes into sight and has now its own event then B has - worst case - 7 seconds time to shoot at A before A could even spot B.

Actually that's the best case - worst case would be A fails to spot B when its spotting event next occurs, giving B 14 or more seconds to shoot A

[poesel]

Ok, I've tried to test if movement or sound would trigger spotting but it looks like it does not.

I've used the same setup but delayed movement and shooting of some units (two different tests) and observed if they would be spotted on different times than other non moving/shooting units. This did not happen, they were always spotted on the same time like the others. However the shots fired could be seen before the units were spotted - which makes sense obviously.

IMHO this rigid 7 second grid explains a lot of the perceived weird spotting incidents. The (natural) expectation that something like movement or sound would cause the immediate attention of anyone able to see/hear it and the game not modelling it causes surprise and thus disappointment by the player.

OTOH I very well understand the reasons for doing it like this. It saves resources because nothing has to be computed during the 7 seconds and it makes the resource drain more planable.

A way to make spotting more 'natural' would be to use an event based system that triggers spotting when 'something' happens. But this is quite the reverse of the current system. Now every 7 seconds a unit spots into all AS it can see. In an event system an AS with an event would cause units in other ASes with units to make a spot into it. This can create CPU spikes when things happen that many can spot at the same time.

Anyway I'm pretty sure BFC knows that very well and much better than me. But I hope that they will spend some development time on it because IMHO spotting is such a fundamental function of the game. Making it more 'real' would be very beneficial

Just to be clear: I don't think that spotting is broken or anything. There are just some cases where it does work in a surprising way.

[Baneman]

...

Anyway I'm pretty sure BFC knows that very well and much better than me. But I hope that they will spend some development time on it because IMHO spotting is such a fundamental function of the game. Making it more 'real' would be very beneficial

Just to be clear: I don't think that spotting is broken or anything. There are just some cases where it does work in a surprising way.

Agree, I don't think it's broken either, I just thought ( before your test proved a 7 second cycle ) that too much time passes between checks - in a previous post regarding this issue I suggested more spotting cycles ie. less time between them - if spotting checks were every 3 or 4 seconds, there's be much less time for enemy to "invisibly" move into view, see you and shoot you. Halving the time between spotting checks could eliminate well over half of the weirdness.

It seems to me ( yes, just guessing here ) that we're at the mercy of RT framerate issues, since in WeGo I'd assume that more spotting cycles would just make the calculation phase take longer. Of course, it would be nice if WeGo could get more cycles even if RT can't - since WeGoers probably see 95% of the anomalies anyway.

[ZPB II]

I would hazard a guess that this test might be affected by computing power.

Some games act funky when they run at full load.

[Offshoot]

Interesting findings. As well as replicating the results, perhaps this could also be done for other types of units to see if all types follow the same rules. It would be also be interesting to see if it works this way for 'unspotting', where a unit goes out of view.

[c3k]

Nice test. BF.C has stated that spotting is done, as you've written, at intervals. I have never seen what that interval is. I wonder if it's the same for all units?

[Michael Emrys]

I wonder if it's the same for all units?

Good point. I wonder if, for instance, more experienced units have shorter spotting cycles. That would be neat.

Michael

[poesel]

Ok, tested conscript Tiger, elite Tiger and elite FO-Team - all the same 7 second cycle. The initial offset for the conscripts was the longest but this may have been coincidence.

Unsurprisingly the elite Tiger spotted more units faster and had by the end of the minute nearly spotted all enemies (except those far left/right back). The conscripts had roughly seen half of that.

I don't think that computing power has anything to do with the cycle. My guess would be fixed no matter what or number of units on the field (more units -> longer cycle).

I agree with baneman that the limit is RT mode. For Wego it doesn't matter much. I have a rather slow machine and waiting for the computation never bothered me (btw: it would be nice if it would ask for the filename first and then do the computation - that way no user input is necessary until the start screen comes back. I usually tab away and if the music comes back I know its finished).

If anyone is interested here's the scenario:

https://dl.dropbox.com/u/8811801/001%20Tank%20Spotting%20Test.btt

[Steiner14]

Wow, excellent test.

Does area firing on a suspected unit (but not spotted) restart the spotting-timer?

My impression is, that in the case of area firing, the spotting-cycle-timer immediately becomes restarted and that way a new round of spotting can be forced by the player.

7 seconds can easily be two shots of deadly cannons.

An estimation of the worst case (spotting-timer of unit A sadly begins in the moment right before the enemy unit B comes into LOS and starts it's own cycle and spots A) shows shocking random results that can be almost purely influenced by the spotting-timer:

If we assume a 90% hit ratio for both units: means 10% of being not hit.

This radom factor can reduce the chance of the unit with the unlucky bad timed spotting-timer, which starts almost 7 seconds too late, to 1% (0,1*0,1)! :eek: Only because it's spotting cycle starts different. A factor of 10 worse for the one with the unlucky timer when things should be equal and only dependent who draws first!

At a 75% hit probability (25% not being hit) it's chances are still reduced to 6% instead of 25%! A factor of 4!

At a 50% hit probability this still reduces the chances of the 7 seconds later spotting unit to 25%!!! A factor of 2 when everything should be equal!

This is only under the assumption, that both units would have exactly the same conditions, but the unit with the lucky random timer get's the benefit, while the unit with tactical equal chances is reduced to almost dead.

Ofcourse this is only valid, if the spotting-cycle-timer is not restarted due to other factors we don't know yet.

Another example:

Unit A outmaneuvers B and has a hit probability of 90% (i.e. hull down).

Unit B has only a hit probability of 50% but has the lucky spotting-cycle-timer: Unit B spots A almost 7 seconds earlier.

This reduces for unit A, the one with the huge advantage, the survival rate to 25% (!) before it can spot unit B - only because of this random factor.

Without this factor the survival rate of outmaneuvered unit B after two shots would be 1% and of unit A, the one outmaneuvering B, it would be 75%.

Under the assumption that both units shoot at the same instant. In reality A would shoot first and gain an additional advantage which can totally vanish due to the spotting-cycle.

If the spotting cycle would be faster than the ROF of guns, this random factor quite dramatically could be reduced since not two rounds could be shot before the unit gets a chance to recognize what is happening.

BFC pls fix or do somefink!

[c3k]

Nicely done. What happens if more than one tiger is present? Say, 10?

[poesel]

Nicely done. What happens if more than one tiger is present? Say, 10?

Sorry, don't get the question. What do you want to find out with 10 Tigers?

10 Tigers in one unit? In C2?

[Vanir Ausf B]

Interesting to see that experience level affects ability to spot other units. I wonder if altering the experience level of the infantry would affect how quickly they get spotted.

[A Canadian Cat]

This experiment certainly is enlightening, @poesel71. Thank you.

I think many of us knew that spotting calculations had to be happening in some kind of heart beat. Very interesting to see what this experiment shows.

Knowing this will probably change some of my decisions in game. Here are two examples that either were a result of this or could have turned out differently if the spotting calcs had landed differently.

It means that if you set a 10s pause on the shoot part of a set of shoot and scoot orders you are giving your tank at most two chances to spot the enemy but most of the time they will only get one chance. This happened to me just he other day I had a Sherman do a shoot and scoot with a 10s pause facing where I knew a Tiger tank was (that's why I did not go for longer than 10s). The Sherman saw nothing and backed away.

Note: not saying the outcome was not possible. The Sherman had not seem the Tiger before and to get to its shooting position it had to go uphill and crest plus there were a tree or two between them. So, I never though it was a sure thing.

For moving in and out of smoke this as serious ramifications because once your unit clears the smoke they can sit there blind for a long as 6s. This explains what happened to me just this morning. I had a Panther moving out of a smoke screen right next to where a Sherman Firefly was sitting. The Panther cleared the smoke and then sat there like an idiot for a long time (I did not time it) not even 10m from the Firefly. Finally the Panther spotted it and shot it. But the Firefly could have easily spotted the Panther 1s after it cleared the smoke and then it would have a 6s drop on the Panther. Yikes. It all depends on what which unit is on what schedule.

Note: again not saying this could not have happened. The Panther was in the smoke buttoned up and did not spot the Firefly earlier before the smoke was dropped. So real life could have turn out just like that.

[c3k]

Sorry, don't get the question. What do you want to find out with 10 Tigers?

10 Tigers in one unit? In C2?

Actually, 10 identical, independent Tigers: the purpose would be to see if the spot "polling" period (spp) changes from 7 seconds. With more units present does the game change the spotting routine? Is it load-leveling the processing based on the number of LOS origination points, or is it fixed at 7 seconds?

Heck, 100 may be a better number for an initial hack.

[Vanir Ausf B]

Interesting to see that experience level affects ability to spot other units. I wonder if altering the experience level of the infantry would affect how quickly they get spotted.

It would be cool to know if higher experience units are more difficult to spot.

[Offshoot]

Actually, 10 identical, independent Tigers: the purpose would be to see if the spot "polling" period (spp) changes from 7 seconds. With more units present does the game change the spotting routine? Is it load-leveling the processing based on the number of LOS origination points, or is it fixed at 7 seconds?

It could do of course, but is it likely? Assuming that the polling time is a limitation mainly from realtime, if the polling time changes, there will then have to be a system in place that ensures these changes are synchronised between the two players' game instances.

Knowing the polling time, this set up would make it easier to derive quite solid info on the effects that, for example, unit experience and terrain type has on spotting. I'm not convinced yet that movement and sound do not affect spotting; it might be that spotting in this setup was easy enough that the effect of movement and sound was lost.

So how long before this gets exploited? Throw out a sacrificial unit to check when the polling for an enemy unit occurs (a twist on recon by fire), then start counting out lots of seven seconds. Though I wonder if in WeGo whether the counter is reset at the beginning of each turn?

[akd]

I think the people attempting to exploit this will get exploited.

[poesel]

I've added another Tiger from another company and ran the test again. The spp (! ) is still 7 seconds but the offsets are different between the Tigers. So both units spot at different times.

This was an expected outcome as the purpose of the offset is to spread the spotting over time. Btw the better unit doesn't get a shorter offset. Smaller offset is no advantage anyway so its probably more or less random.

I didn't test if the spp is reset after each turn but I'm pretty sure it won't. Wego is RT with pauses and resetting the spp doesn't make sense in RT. Would also be quite unfair in Wego.

I tried to add another three battalions of infantry to see if the spp changes but that is too much for my machine. Then I deleted 2 of the 3 battalions and the spp is still 7 seconds.

So the spp is most probably a fixed value. Maybe someone with a faster machine can test this, too.

...

I'm not convinced yet that movement and sound do not affect spotting; it might be that spotting in this setup was easy enough that the effect of movement and sound was lost.

Well, please try and prove it different. I gave a quarter of the infantry a 5 second delayed move order. The Tiger spotted first after 4 seconds so those units he spotted stood up and started walking. But none of those who were not spotting were spotted standing up on the 5s mark. The Tiger spotted some of the after 11s while walking. Similar for sound.

Please note that this tests does NOT determine if spotting moving units is easier than prone ones. Its just to see if movement CAUSES spotting.

[Offshoot]

Please note that this tests does NOT determine if spotting moving units is easier than prone ones. Its just to see if movement CAUSES spotting.

Ah, I see - sorry, misunderstanding on my part. I get it now - no spotting events occur between the polling events even if the status of a spottable unit changes in that time. You've mentioned sound resulting from firing - were the units firing actually firing on the tiger or area firing elsewhere? I just wonder if receiving fire triggers a spotting check?

[Steiner14]

Poesel,

could you check with area fire light from the tiger on a unspotted unit, if the spotting-cycle is restarted as soon as the weapon is fired?

[poesel]

No to both questions. Looks like neither receiving nor giving fire triggers spotting. Firing does, no surprise, vastly increase the chance of being spotted.

I have to say that I'm a bit disappointed that no spotting happens when you are shot at. IMHO that would be a very natural reaction to at least look into the direction were it came from. Or maybe my test is skewed here - its quite a synthetic setup. But everything I have seen so far points to a rigid 7 second cycle no matter the circumstances.

There are probably very good reasons to do it like this. But this, I feel, is the 'uncanny valley' of spotting which creates weird results sometimes. I hope this can be fixed.

As for exploiting: the amount of work and uncertainty of the results combined with the expected yield - well, I'd rather spent the time thinking about tactics or checking the terrain.

[Steiner14]

Thanks.

As for exploiting: the amount of work and uncertainty of the results combined with the expected yield - well, I'd rather spent the time thinking about tactics or checking the terrain.

A factor of 2 of better chances in engagements up to a factor of 10 (+1000%!) for exploiting IMO will not be ignored. The more precious vehicles in battles are, the higher the impact of this purely random-factor on the outcome of the whole battle and therefore the higher the impact, if it can be exploited from one side.

IMO it's only a question, how difficult it is, to find the restartime of the spotting-cycle or how interested a player is to win. Since it seems that it's a constant cycle anyway, it could be enough to write down the second in the game the last time the vehicle spotted an enemy.

E.g. before engaging a tank, spot infantry with this tank. Write down the battle's second and you know when it will occur in future seconds. Then time the engagement against the enemy tank accordingly.

[Steiner14]

And it get's even better:

Move infantry into LOS of the enemy tank and you could get the time, when it's cycle begins!

Once you have your tanks' restart-times and the one from the enemy tank, choose the tank with the restart-time closest after the enemy tank's restart time and voila, you grab the biggest time window, during it's time the enemy tank will see - nothing.

Ugly, very ugly.

[Paper Tiger]

A 7-second cycle of spotting would certainly explain some of the LOS oddities that crop up from time to time. But isn't this what a second processor could be doing? I've got one core doing absolutely nothing while I'm playing this game.