(Review Completed) [SP] Un-Impossible Mission

[THobson]

·  a complete nightmare for the first half dozen times you attempt the mission

I am not sure this changes after the 6th attempt

·  sufficiently hard that a few players will give up at that point

I think you will have succeeded here

·  extremely difficult - but possible - to get away with your whole squad intact

It does not feel difficult - it feels random

·  tricky but perfectly possible to survive yourself and get away with only a few casualties*

it does not feel tricky it feels… well… random

·  relatively straightforward to get away yourself with most of your squad dead

I would restart.   Edit: but once I have finished the mission I might want to try it with a smaller squad

·  random in terms of the opposition's axes (and type, to some extent) but reasonably consistent in terms of difficulty

A success - except I must have dropped lucky to get clean away this time

·  sufficiently random that, even once you are familiar with the locations, you never really get a firm grasp on what happens at all of them

A total success

·  not a barrier to replaying the mission once you have been well into it once

Well I know how I feel now, but really the whole experience of the mission, once in, is such I can see myself relenting - there really is no other show in town (except a couple of missions of my own of course  ;D)

You get up to two replacements at the civvies for precisely this reason.    The first replacement is not conditional because, originally, I never thought anybody would ever get there without casualties.

It is not the replacements that is my concern it is the casualties.

But this is only one data point - I am telling how I see it.  If you see it differently then stick with it - it is your mission not mine

[macguba]

All data is useful, particularly from such a dedicated and helpful beta tester as yourself.     I am not much of a data point on this matter anyway, since it's not at all random to me - I know exactly what might be coming.     I'm not shy about restarting 3 seconds into the mission.    Sometimes you need to - I wish it didn't say "8 is down" the very second the mission starts but the only way to overcome that is put the attack groups further away, which in turn means you sometimes get time to have a cup of tea and read the paper before strolling off without meeting anybody at all.    

This is not so much important for the mission as interesting to me.    The problem of how to make it random yet consistent is one that I never really solved .... so you mentioning it feels a bit like somebody poking an old wound.   ;D   I'd like to do it better of course.

Sorry to be a bore about this, don't feel obliged to answer if you've had enough.   In what ways did it feel random?  

[Planck]

As far as the times I have had to restart when playing this mission, I found that once I managed to detect where I was, and in which direction safety for me and my men lay in, I could fairly consistently manage to escape with all my men intact 80% of the time.

You have to be quick most of the time in deciding where exactly you are though.

The other 20% is an acceptable amount of uncertainty for me.


Planck

[THobson]

Thank you Planck.  If it can be done 80% of the time then I relent on my comments.  My tactic (because I don't know the start points - nor the safe directions) is to go in with all silenced weapons, hit the deck and go to stealth immediately.  Deal with the initial attacks and then crawl away, in what for me at that time is a random direction, before the M1A1 and M2A2 (a deadly combination) turn up. {What a complicated sentence!}  I generally choose to go down the gradient - for no reason other than that it doesn't require me to decide at the time and I don't want the game to think I am going up the hill for a long time yet.

Mac - the randomness I refer to is that, give or take the odd glass of wine, my skill level is pretty much constant, my tactics are the same but the results can be very different.  I admire Planck's tenacity to play the mission so much he has developed a detailed experiential knowledge of the start points.  But the real intention is that a group of soldiers bail out of a chopper into the mist, they don't land where they expect and they don't know where they have landed.  To be attacked very soon is a reasonable expectation - to start taking casualties in the first couple of seconds feels unfair.  But hey life is unfair - I guess I am coming round to the way it is, after all it has been known for real para drops to land on top of the enemy.

One option might be to reduce the randomness of the initial attacks, but delay them slightly, but that works against what you are trying to do - random but consistent.

[THobson]

All data is useful, particularly from such a dedicated and helpful beta tester as yourself.

Thank you for the comment.  I am beginning to wonder though if a beta test by me is a kiss of death for a mission.  Several of them result in either total silence or - in one case - the complete abandonment of a campaign.

http://www.ofpec.com/yabbse/index.php?board=23;action=display;threadid=19626

http://www.ofpec.com/yabbse/index.php?board=23;action=display;threadid=19251

http://www.ofpec.com/yabbse/index.php?board=23;action=display;threadid=18946

http://www.ofpec.com/yabbse/index.php?board=23;action=display;threadid=18433

[macguba]

ROFLMAO!

Don't worry, this is mission is in a  "finished" state - I could post it now.    However there is a list of improvements - including the stuff you have mentioned - that I want to make before drawing a line under it.

A massive attack at the very start is an essential component of a mission like this.   However, I agree with you - casualties within the first two or three seconds is unfair.  

I have discovered two tactics that work:  silenced weapons and all round defence as you use;  and just legging it.    However that requires a good knowledge of which directions are relatively safe, which takes a bit of time (or an examination in the editor) to acquire.     If I'm testing with a vulnerable squad that's what I do, retrying as soon as I take a casualty.    In fact its even trickier because depending on what I'm testing for I'm fussy about which start position I'm at.    I always get away ok after a few minutes of trying.    And nine times out of a thousand there will be no opposition at all.  

I'm going to lay out the statistics here, as much to keep my own head clear as anything.  I haven't done this for a while.

++++++++++++++++++++++++++++++++
You don't have to read this section

The start positions are all different but most have three infantry attack groups.   These groups have probabilities of presence of 90%, 75% and 75% respectively.     (These probabilities are the result of many hours of playtesting and statistical analysis) Each group has a random composition of between 3 and 9 loons, (ditto) with an average of 6.3.   Thus you can be attacked by anything between 27 and 0 loons, which is indeed a pretty broad range.    The loons all have the same skill and are a broad mix of all infantry types - soldiers, medic, machine gunners, LAW soldiers, G36, Steyr, XMS, etc..    

However, when you look more closely, the likely numbers narrow down considerably.

Prob of 0 groups = 0.006  => virtually 0
Prob of 1 group   = 0.094  => 10%
Prob of 2 groups = 0.394  => 40%
Prob of 3 groups = 0.506  => 50%

0 groups =>  0 loons
1 group   =>  6 loons
2 groups => 13 loons
3 groups => 19 loons

Combining all that gives an expected number of attacking loons = 15

The possible angles of approach of each attacking group leader cover an arc of up to 45%:  with loons spread out either side of him, and obstacles like trees or buildings, that probably gets out to 60%.

For simplicity's sake, let us say that the number of loons is equally likely to be 4, 5, 6, 7 or 8.   (This understates it slightly but we can add back at the end.)   If we look at three groups that means that each number of loons from 12 to 24 is equally likely.   The standard deviation of that data is 3.74 so call it 4, (there's a spot of adding back) so in other words two thirds of the time when there are 3 groups attacking there will be between 14 and 22 loons.  (The average is 18 and it turns in statistics that the +/- range given by the standard deviation captures about 2/3 of the data.   I have used the stdev for the population rather than a sample, but it doesn't make a significant difference here anyway.)  

2/3 is 14-22   (2/3*0.5= 0.33)
1/6 is 23-27   (1/3*0.5= 0.08)
1/6 is 9-14     (1/3*0.5= 0.08)

When 2 groups are attacking there are between 8 and 16 loons with a mean of 12 and a standard deviation of ooooh 2.58 well I suppose we'll call it 3.  

2/3 is 9-15     (2/3*0.4=0.27)
1/6 is 16        (1/6*0.4=0.07)
1/6 is 8          (1/6*0.4=0.07)

When 1 group is attacking, well its only 10% of the time and there are fewer than 9 so we'll leave it at that.

Summing these 6 probs gives 90% which is good.   If we say that 14 is much the same as 15 and round everything off we come to the final table of probabilities of expected attacking loons

7%        23+
33%      15-22
43%      9-14
17%      0-8

Whew.  Well we've got there.     If anybody spots any significant mistakes in that lot please tell me!  


+++++++++++++++++++++=

OK you can start reading again.

The very rough overall conclusion is that

- about 20% of the time you are attacked by 20 or more loons and you are toast

- about 20% of the time you are attacked by fewer than 10 loons and you should be ok

- about 60% of the time you meet between 10 and 20

- there are other significant factors such as the direction and timing of the attacks, this analysis covers only one aspect of the randomness


The interesting thing is that THobson, Planck and I are all correct.     In spite of all my efforts THobson is right, it IS still too random.    From Planck's experience it would seem that about 20 loons is the maximum you can deal with.   However, the good news is that you only have to make about 5 attempts to meet with fewer than about 10 loons.  While this is more than enough to toast you early on - not least since they are probably coming from two directions - once you know what you're doing it shouldn't be too bad.

A large number of loons attacking is always going to be hard to deal with.   However, a small number can still be devastating if they catch you quickly or happen to come in on an axis that is not well covered.    In other words, you won't get an easy run as often as these figures suggest.    

I was concerned yesterday that I would have to do a major rethink on the starts.  However, I'm more relaxed about it now.  THobson, I suspect you were just unlucky and had a tough sequence:  next time it should be easier.

I can also now reveal that previously, the percentages were 90%, 70% and 70%:   I put them up to 75% because I wasn't getting enough complaints about the start being too hard.  

[THobson]

next time it should be easier.

Radom events have no memory so I can't count on that.

If I get time I will have a look in the editor and  work out exactly the probabilities of each possible number of loons.  Sad I know but it is the sort of thing I like to do.

It is not just the number though, 3 loons from one direction are easier than 1 from each of 3 different directions.

ROFLMAO!

Means what?  I am the wrong side of any age you care to guess at to know what this means.

[macguba]

Roll on floor laughing my ass off.   It's internetspeak you know, the Young use it.    It's the superlative of lol (laugh out loud), ROFL (roll on floor laughing) being the comparative.    Actually, lol means "I smiled" and ROFL means "I chuckled out out".    

I should also add to my previous post that this analysis doesn't cover all the start positions, some are different although playtested to be about the same difficulty.

[THobson]

Thanks.  I needed that.

During my lunch-time walk by the Thames today I was thinking about randomness with consistency.

My thinking so far is: if you make everything random then you cannot guarantee consistency.  There will always be a finite chance of one extreme or the other occurring,  but if you introduce some dependencies you might be able to do it.  For example, have one group with a probability of existence of 50% and have another that is certain to exist if the first group does not and certain not to exist if the first group does.  So it is consistent and it is random.  I am not saying do exactly that in your mission but if you haven't already, it might be worth playing with this concept.

[macguba]

It's a good idea and I played with it some time ago.   It isn't actually used in the mission though I can't really remember why:  it does put pressure on number of groups (since you are using two groups to create one) unless you start having some really convoluted logic.     Using Resistance to reinforce West would of course be a total no-no.

It's really only relevant at the start position - over the rest of the map the rest of the randomness pretty much balances out.     However, when you have so many possible start positions, it gets really messy really quickly.   I decided quite early on that what happened at a start position would not be influenced by whether the player was present or not.     Though this is not so much a fundamental design rule as an excuse for laziness.....

[THobson]

I had the 'convoluted logic' option in mind - not the two for one option.  Anyway just so long as you had thought of it.

I might experiment with this a bit, it seems to have some useful possibilities.

[macguba]

It has excellent possibilities.    Very useful, once the player knows the mission, for inducing the fear factor:  you know (because you haven't met them yet) that there is somebody out there looking for you.

Or for creating pseudo random opposition.   If you combine it with setPos commands for the whole group and some nice convoluted logic, the possiblities are endless.   In fact it would save you groups in the long run, since you could have say 100 possible locations but need only 30 groups to cover them, if you wanted only 30 used.    No extraneous guards where the player isn't.

[THobson]

I now have a lot of ideas for a mission - I just don't have a 'plot'.

Anyway, I said I would work out the exact probabilities of every possible number of starting loons.  I lied.

Instead I have written a program that simulates one million occurrences of the mission starting at BMP Ambulance & BRDM (it is now a simple matter to feed in the probabilities for the other starting locations and check them as well).  In the table below I show the % number of times that a particular number of loons were generated in the one million simulations (give or take a bit of rounding).

#Loons       %
0             Â Â Â 0.6
1                0.0  (this is an impossible outcome)
2            Â Â   0.02
3             Â Â Â 0.27
4             Â Â Â 1.1
5             Â Â Â 2.3
6             Â Â Â 2.8
7            Â Â Â Â 2.2
8           Â Â Â Â Â 1.8
9              Â Â 3.1
10          Â Â Â Â 5.6
11          Â Â Â Â 8.1
12        Â Â Â Â Â Â 8.8
13          Â Â Â Â 7.7
14          Â Â Â Â 6.5
15          Â Â Â Â 6.4
16         Â Â Â Â Â 7.6
17         Â Â Â Â Â 9.0
18          Â Â Â Â 9.1
19         Â Â Â Â Â 7.6
20         Â Â Â Â Â 5.1
21         Â Â Â Â Â 2.8
22         Â Â Â Â Â 1.2
23         Â Â Â Â Â 0.4
24         Â Â Â Â Â 0.1
25          Â Â Â Â 0.02
26          Â Â Â Â 0.0024
27          Â Â Â Â 0.0

Interesting that there was not one ‘full house' in the 1 million simulations.

Interesting shape if you plot it.

Would you be interested in the result for any of the other locations?

Edit:  Actually it is not too surprising that there were no occurrences of 27 loons.  The probability works out at 1 chance in a million.  Or to put it another way - it will happen 14 times for each time you win the UK lottery!!

Edit2:  I ran it again, this time for 10 million mission starts and got 27 loons 9 times.

[macguba]

Now that's cool.   (If you're a spod.  ;D)

This is really interesting stuff.    0.6% for 0 is exactly what I got on my steam powered spreadsheet, which is reassuring.    The figure for 2 is odd - it should be 0, since that is also an impossible outcome.

Also good to know that my figure of 20% for ten loons or fewer was spot on, but that 20% for 20 or more was flat wrong - the actual figure is 10%.     In between is then 70% rather than 60%.     I was assuming that both tails would be equal whereas of course the high end is much less likely.   It's noticeable that if you make it below 10, rather than 10 and below, the figure drops to 14% from 20%, and down to 4.5% on the high end for 20+.   Between 10 and 20 inclusive is 81.5%.     Planck's figure of struggling 1 time in 5 actually suggests that 18-19 loons is where it gets really hard.

I've plotted these on a graph and it shows 3 clear peaks, which is what you would expect with 3 groups.  They are at 6 loons, 12 and 18 - again, pretty much what you would expect.  The trough between 12 and 18 is the most undesirable feature in my view.  

You will have noticed by now that most of the starting locations are much the same.    The ammo depot in the north with the Abrams looks interesting, but in fact it isn't and in any case has been changed in the new version.    You are now guaranteed an Abrams and a Bradley:  the Bradley is guaranteed to have 3 dismounts.   The only uncertainty here is whether there will be any more dismounts (who all start mounted) and the singe infantry group which is on 50%.   In other words, the only real uncertainty here (apart from precise axes of attack) is whether that infantry group is present or not.

If you wanted to examine one more closely the jeep/mg groups at the ammo/fuel station/repair junction in the south is probably your best bet.  

However, the key question is how to smoothe it out a bit.    The ideal is of course a more or less normal distibution.


Edit:  and yes, on a test run this afternoon I got a 0 attack.    ;D   A 1 in 167 shot, which was particularly annoying as I wanted to be shot at.

[THobson]

The figure for 2 is odd - it should be 0, since that is also an impossible outcome.

No it is not - unless I have read it wrong.  At infil6 I think you have only 2 loons per group that are certain if the group exists.  In fact that is the case at most locations.

You will have noticed by now that most of the starting locations are much the same.

I didn't get that feeling when I was keying in the probabilities, but I have not run the result yet.

Do you have a C compiler?  I would happily give you the source code - which now has all the start locations entered.  All you need to do to change the start location is change a variable to infilx where x = 0 to 7  (I used infil0 for the Ammo & UAZ, the others follow your marker names).  You woudl also easily be able to change the other probabilities to match your current version.

Your graph looks exactly like mine - it even has the same colours.

I take it spod is good.