• Steven (unregistered)

    And I thought "military intelligence" being an oxymoron was just a joke...

  • (cs) in reply to Code Dependent
    Code Dependent:
    Vino:
    Not only is it spelled properly in other places, but it is misspelled improperly twice...
    What is the proper way to misspell a word?
    With a "(sic)" after it.
  • Jim Lard (unregistered) in reply to codemanque
    codemanque:
    They want it to be /sometimes/ repeatable, and /sometimes/ different. In other words, they want randomised but repeatable scenarios. The WTF here appears to be in the attitude of the submitter, not the customer, and possibly (seeing as we only have one point of view to go by) not even the developers.

    As some others have pointed out though, if the randomisation is initiated by the controller hitting a button while the simulation is running, it will always be random because the point at which the randomness is initiated - at which the scenario branches from its strictly-defined rules - is occuring at a random point in time. (ie a submarine veering to the left at one moment will result in a different scenario from the submarine veering to the left one second later.)

    Just seeding the random number generator doesn't cut it for reproducing these scenarios, it sounds like what's really needed is the ability to record the random events and replay them later.

    But this would only be any good if none of the agents in the simulation had changed. If there's human actors in there, for example, or if the simulation of the enemy rickshaw has been tweaked, it could lead the entire simulation to take a different route... argh nightmare.

    This is why things like this have to be chiselled in stone in the spec before you even start. (Or alternatively, you do what we do in our office, and just wing it and hope it all works out somehow, lol.)

  • buzzard (unregistered) in reply to LightStyx

    That you're 100% retarded?

  • Sinan Unur (unregistered) in reply to ATimson

    You are out of your depth here. Leaving aside computer games for a second, any serious simulation software must involve some randomization if it is going to be useful at all. Further, if the purpose of the simulation is training, then every sequence of random inputs must be replayable.

    That is why the pseudo in pseudo-random number generator is a feature, not a bug.

  • orlly? (unregistered) in reply to CDarklock
    CDarklock:
    pink_fairy:
    On the other hand, I'd purely love to find out what Americans use for "alternate" when they actually mean alternate...

    We say "alternative", of course. I mean, what would be the alternate?

    win! :D

  • (cs) in reply to PaladinZ06
    PaladinZ06:
    running the exact same test for everyone seems... dumb.
    You see things strangely. Giving everybody completely different tests seems dumb to me, because then you can't meaningfully compare the results, so you end up knowing precisely nothing about the cohort you've tested.

    Scene: A foxhole somewhere in a future war

    Sarge:
    Right men! I need someone to sneak forward under cover, infiltrate the enemy lines, drop a grenade into their supplies bunker then signal us to advance. This is a difficult mission and will require a skilled trooper to complete it. Who got the best marks in training?
    Pvt. Foo:
    Well, I got 78% in Maths, sarge.
    Pvt. Bar:
    I got 84% in Woodwork, sarge.
    Pvt. Baz:
    And I got 94% in Media studies, sarge.
    Sarge:
    Right, you're clearly the best man for the job, Private Baz! Off you go.
    .... 90 minutes later, same foxhole, but with half the platoon dead and the enemy closing in ...
    Sarge:
    What the hell went wrong, Private? You were clearly the best man in testing, but when it came to the actual mission, instead of shooting them all you gave them an analysis of their propaganda and communications strategy!
    And the moral of the story: "best" at anything does not equal "best at anything".
  • (cs) in reply to C4I_Officer
    C4I_Officer:
    Nightst4r:
    I bet the random seed was 42... for some odd reason.

    I knew I was not going to like that answer.

    Ok, but wait til you hear the question. It's a doozy!
  • (cs) in reply to RobFreundlich
    RobFreundlich:
    LightStyx:
    The developers have their head up their asses because they made something according to documented requirements and the customer didn't like it because they didn't get what they expected. It's true the developers could've asked for more clarification for what they meant by "random" which is a common fault for many developers no matter how much experience they have. However, the Admirals are displaying common customer behavior when they say they want one thing done one way, see it, and don't like it because that's not what they had in mind. So they scold the developers for being lazy and not following directions.

    The real problem here isn't a WTF so much as a very common anti-pattern in software development: the customer throws some specs over the wall, the developers implement them blindly and throw them back over the wall, and the customer don't like the results. The problem here is the wall, not the developers or the customer.

    No, the problem IS the developers. They do this job every day of their lives and it is THEIR responsibility to know about the existence of the wall and deal with it, not to blindly march ahead as if it didn't exist. It is not right to expect the customer to understand and deal with this systemic problem, because the customer does not do this for a job, and you (as a developer) do.

    Lightstyx's "common customer behaviour" comment makes it pretty clear he's the incompetent-and-blame-everyone-else kind of developer. Before you can be any good at your job, you have to learn that it is about more than just writing code; that's only the start of it.

  • Sane Person (unregistered)
    Jim Lard:
    Just seeding the random number generator doesn't cut it for reproducing these scenarios, it sounds like what's really needed is the ability to record the random events and replay them later.

    But this would only be any good if none of the agents in the simulation had changed. If there's human actors in there, for example, or if the simulation of the enemy rickshaw has been tweaked, it could lead the entire simulation to take a different route... argh nightmare.

    Of course there is the unstated assumption here that the enemy forces in the simulation respond to the players. Maybe they just continue to march in a staight line, shooting straight ahead, no matter what the player does. The "March of the Wooden Soldiers" scenario.

  • (cs)

    I think "Randomizaion" button is much cooler than Entropy button...

  • (cs) in reply to Jim Lard
    Jim Lard:
    But this would only be any good if none of the agents in the simulation had changed. If there's human actors in there, for example, or if the simulation of the enemy rickshaw has been tweaked, it could lead the entire simulation to take a different route... argh nightmare.

    It's actually really hard. What needs to be done is for the PRNG to determine the behaviour/characteristics of the actors in advance.

    So, at the start of the simulation it decides that the destroyer will turn 10 degrees to the left after 5 miles or whatever. Then, if the destroyer gets destroyed before that point, it doesn't mean that the PRNG is one step out for other actors.

    However, this doesn't take into account that you may have:

    • destroyer turns 10 degrees to the left after 5 miles.
    • 1 minute later, the accompanying A/C carrier turns 10 degrees to the left as well to follow its destroyer escort.

    Now, if the destroyer gets destroyed, what does the A/C carrier do? Does it still follow the ex-destroyer, or does it carry on in a straight line (or does it retreat). Do other nearby ships come to assist, or what? What if those other ships had had random actions predetermined? Do they still do those, or not?

    Basically all you can do to be 'realistic' is let the whole scenario be repeatable identically as long as all the input actions are also repeated identically. If any input actions are different, then the whole scenario should not repeat identically at all.

    So, the PRNG should set things like 'trigger-happiness of commander', 'reliability of engines/weaponry', 'accuracy of helmsman' etc, and that's it - not precisely setting things that are going to happen, but the way things behave in general.

    Or you have humans controlling it (which is actually probably the cheapest and best way of doing it).

  • (cs) in reply to Vino
    Vino:
    I CANNOT BELIEVE that nobody yet has complained about the terrible misspelling in "Randomizaion"
    No, no, you're missing the point completely. The real WTF is that the developers heard "randomizaion" and assumed that it meant "randomization". While in reality - as is apparent from the article - "randomizaion" is just Military speech for "alternativation".
  • Dave (unregistered)

    A randomizaion is the stone portico of a corinthian greek temple, used for votive offerings. Presumably the Navy wanted a facility for lobbing these at the enemy as some sort of offensive measure.

  • Objective (unregistered)

    War has very few random elements. Just some basic logic would do... #

    IF MyWeapons > ENEMY then

    DoCmd.Attackandkill

    Else

    DoCmd.Runlikemad

    End

    Seriously any random element in a sim like this sucks, I mean you'd have to be an idiot to think God plays dice :-)

  • Anon (unregistered) in reply to DaveK
    DaveK:
    PaladinZ06:
    running the exact same test for everyone seems... dumb.
    You see things strangely. Giving everybody completely different tests seems dumb to me, because then you can't meaningfully compare the results, so you end up knowing precisely nothing about the cohort you've tested.

    That's why you repeat the test several times for each trainee. It'll play out slightly differently every time and for different people, but, if it's well designed and the starting conditions were the same everybody (but slightly different for each repetition), it should, on average, give you a pretty good idea of the abilities of each trainee. So maybe a trainee got lucky the first time through, it's doubtful that they'll be so lucky the next 9 times. And if they are, well, it's better to be lucky than good.

  • Edward Royce (unregistered) in reply to Anon
    Anon:
    He should have just replaced with the button with a pop up that says "Who is the best Admiral? You are! Yes you are! Yes you are!"

    I think that would have gone over better.

    Don't forget the picture of a lolcat.

  • Beavis (unregistered)

    They definitely should have called it the "Crazy Ivan" button.

    "Crazy Ivan! Crazy Ivan!"

  • Artie M (unregistered)

    Sounds like that movie the 'Pentagon Wars'.

  • ath (unregistered) in reply to Timothy
    Timothy:
    Wouldn't it have been a better idea to generate a random seed and display this number?

    Yep, you nailed it. This is a common request. Simulations must be reproducible with predictable behaviour, just like the rest of your code. It is a common newbie mistake to use "random(clock())" or similar instead of using a fixed/user controlled random seed.

    When running stress-tests, simulations or whatever, you often find unexpected bugs/features after the 1001st test run. If some idiot used a "random(clock())" construct, you can not reproduce it. I'd even go as far as calling it an antipattern.

    Of course the general wants to reproduce and reuse those random scenarios that turned out interesting.

  • Buffled (unregistered) in reply to Anon
    Anon:
    Wait a minute, so the button was set up to use the same seed every time it was pressed, but since the admirals were the ones pressing the button and the state of the simulation at the time they pressed the button depended on exactly when they pressed the button, wouldn't the scenario play out differently if the pressed the button say 1 minute later the second time they ran it and therefore still be practically unreproducible? Guess it depends on the time resolution of the simulation.
    Perhaps you should learn how PRNG's work before posting?
  • Anon (unregistered) in reply to Buffled
    Buffled:
    Anon:
    Wait a minute, so the button was set up to use the same seed every time it was pressed, but since the admirals were the ones pressing the button and the state of the simulation at the time they pressed the button depended on exactly when they pressed the button, wouldn't the scenario play out differently if the pressed the button say 1 minute later the second time they ran it and therefore still be practically unreproducible? Guess it depends on the time resolution of the simulation.
    Perhaps you should learn how PRNG's work before posting?

    Perhaps you should read the article before posting? I'm fully aware of how PRNG's work, but the pressing or not pressing of a button and when exactly that button gets pressed isn't determined by a PRNG, it's controlled by a person who adds a element of randomness that is not reproducible.

  • methinks (unregistered)

    "...the paradoxical request..."

    What? Why paradoxical?

    This is a core feature for any kind of training system and would have been implemented much better by some sort of "scenario editor" where one can create a scenario, add the desired randomness if required, save the scenario and replay it whenever one wishes in order to track the trainee's progress.

    No WTF here. Except in the requirement analysis...

  • methinks (unregistered) in reply to Herman
    Herman:
    AT:
    ["Unexpected" and "military" are two words that usually don't go together. In the military, every plan has contingencies and every contingency has plans, and each of the plans have contingencies all their own, ad infinitum.] - Daily WTF

    [No battle plan survives contact with the enemy] - Helmuth von Moltke the Elder

    I'm going to have to side with the German on this one.

    look out, a real nazi! not like the grammar ones that normally lurk around these parts

    Dumbass... von Moltke was a prussian field marshal who lived from 1800 to 1891.

    You do know when the nazis ruled Germany, do you? Well, apparently not. You read a german name and think "nazi"...

    Dork.

  • methinks (unregistered) in reply to Nightst4r
    Nightst4r:
    I bet the random seed was 42... for some odd reason.

    Even though 42 is even...

  • methinks (unregistered) in reply to Pim
    Pim:
    Vino:
    I CANNOT BELIEVE that nobody yet has complained about the terrible misspelling in "Randomizaion"
    No, no, you're missing the point completely. The real WTF is that the developers heard "randomizaion" and assumed that it meant "randomization". While in reality - as is apparent from the article - "randomizaion" is just Military speech for "alternativation".

    Google: "Did you mean: randomization" "Results 1 - 10 of about 324 for randomizaion."

  • methinks (unregistered) in reply to Objective
    Objective:
    War has very few random elements. Just some basic logic would do... #

    IF MyWeapons > ENEMY then

    DoCmd.Attackandkill

    Else

    DoCmd.Runlikemad

    End

    Seriously any random element in a sim like this sucks, I mean you'd have to be an idiot to think God plays dice :-)

    Au contraire!

    A battlefield is most likely not "random" - it's "chaotic"!

    CAPTCHA: validus - yes. yes it is.

  • (cs) in reply to BeenThere
    BeenThere:
    ...Amusing story but you have to get pretty pedantic to make TRWTFs out of it.

    What rock have you been hiding under all these years? You do realize this is possibly the most pedantic site on the Internet aside from Slashdot, right?

    Note: I have inserted at least one intentional mistake to keep the grammar nazis interested in this thread.

  • Herman (unregistered) in reply to methinks
    methinks:
    Herman:
    AT:
    ["Unexpected" and "military" are two words that usually don't go together. In the military, every plan has contingencies and every contingency has plans, and each of the plans have contingencies all their own, ad infinitum.] - Daily WTF

    [No battle plan survives contact with the enemy] - Helmuth von Moltke the Elder

    I'm going to have to side with the German on this one.

    look out, a real nazi! not like the grammar ones that normally lurk around these parts

    Dumbass... von Moltke was a prussian field marshal who lived from 1800 to 1891.

    You do know when the nazis ruled Germany, do you? Well, apparently not. You read a german name and think "nazi"...

    Dork.

    heh, good one. no, seriously, nice comeback, way to get all serious. a) i think it was quite clear that my comment was in jest, b) i think my further comments made it even clearer, c) i can put things in bold to, d) the Nazis have ruled germany since the country was found in 1972, e) if i actually thought someone was a nazi, I wouldn't let on that I know, in case they tried to convert me, f) I didn't read a german name and think nazi, i read a sentence that said "im going to side with the german", and everyone knows those were the last words of the fuhrer, g) no one uses the word "dork" any more (except you).

    i think i've made my point. and now, back to the "real" world

  • Joel (unregistered)

    It seems like there was a strong overreaction on both sides here. The admirals wanted unpredictability so they could test how the people they were training could think on their feet. They also wanted reproducibility so they could test how different people react under the same scenario. The simple solution would be to add a little box next to the entropy button that would allow the instructor to enter a random seed. This way they could have unpredictability, but could also save seeds so they could test the same scenarios on different people. If the developers in this WTF didn't have such a smug attitude about it, they would've seen that and just put that option in.

  • (cs) in reply to Anon
    Anon:
    DaveK:
    PaladinZ06:
    running the exact same test for everyone seems... dumb.
    You see things strangely. Giving everybody completely different tests seems dumb to me, because then you can't meaningfully compare the results, so you end up knowing precisely nothing about the cohort you've tested.

    That's why you repeat the test several times for each trainee. It'll play out slightly differently every time and for different people, but, if it's well designed and the starting conditions were the same everybody (but slightly different for each repetition), it should, on average, give you a pretty good idea of the abilities of each trainee.

    Uh, well, yeah, you could do that. Or you could repeat the same set of tests for each trainee, and avoid any freak fluctuations by making sure that what you are measuring for each different trainee is commensurable. (What if your PRNG gives trainee #1 three scenarios where the enemy do something random and stupid that makes it trivially easy for him to prevail, while trainee #2 ... #299 all get an even balance of hard and easy scenarios? If every trainee gets the same set of scenarios, this cannot happen.)
  • (cs) in reply to kastein
    kastein:
    BeenThere:
    ...Amusing story but you have to get pretty pedantic to make TRWTFs out of it.

    What rock have you been hiding under all these years? You do realize this is possibly the most pedantic site on the Internet aside from Slashdot, right?

    Note: I have inserted at least one intentional mistake to keep the grammar nazis interested in this thread.

    Wrong!

    ... err...

    ... I mean right! ...

    ... no, wait a minute, wrong!

    .. no, hang on, right! ... wrong! right ! Wrong! RIGHT! WRO- OH SHI-

    <explodes in a puff of logic>
  • (cs) in reply to Herman
    Herman:
    heh, good one. no, seriously, nice comeback, way to get all serious. a) i think it was quite clear that my comment was in jest, b) i think my further comments made it even clearer, c) i can put things in bold to, d) the Nazis have ruled germany since the country was found in 1972, e) if i actually thought someone was a nazi, I wouldn't let on that I know, in case they tried to convert me, f) I didn't read a german name and think nazi, i read a sentence that said "im going to (...) side with the german", and everyone knows those were the last words of the fuhrer, g) no one uses the word "dork" any more (except you).

    i think i've made my point. and now, back to the "real" world.

    In the real world, nobody would have that many grammar errors in one paragraph, Herman. I marked them for you.

  • John (unregistered)

    This is another of those WTFs that only seemed like a WTF to someone who didn't understand everything that was going around.

    Writing up the story from first person perspective just feels like a trick designed to encourage confirmation bias in the reader, so that we will perceive the same WTF.

    I will call these "Only WTF when Reality Not Comprehended" or OWTFRNC.

    This story is OWTFRNC because prima failed to understand user requirements. Is it a wtf that military trainers want to surprise trainees without surprising the instructor? Of course not.

  • Anon (unregistered) in reply to DaveK
    DaveK:
    Anon:
    DaveK:
    PaladinZ06:
    running the exact same test for everyone seems... dumb.
    You see things strangely. Giving everybody completely different tests seems dumb to me, because then you can't meaningfully compare the results, so you end up knowing precisely nothing about the cohort you've tested.

    That's why you repeat the test several times for each trainee. It'll play out slightly differently every time and for different people, but, if it's well designed and the starting conditions were the same everybody (but slightly different for each repetition), it should, on average, give you a pretty good idea of the abilities of each trainee.

    Uh, well, yeah, you could do that. Or you could repeat the same set of tests for each trainee, and avoid any freak fluctuations by making sure that what you are measuring for each different trainee is commensurable. (What if your PRNG gives trainee #1 three scenarios where the enemy do something random and stupid that makes it trivially easy for him to prevail, while trainee #2 ... #299 all get an even balance of hard and easy scenarios? If every trainee gets the same set of scenarios, this cannot happen.)

    But the point is that you can't give two people the same scenario if it is interactive. If the PRNG causes an enemy to do something unexpected for one trainee, but the next trainee blows up that enemy before that event happens, then they don't get the same scenario. The idea of having strict repeatability in an interactive simulation is impossible. You can only set up everything at the beginning (including seeding PRNGs) and let things play out as they will. If you are going to rely on the results of a single test to assess your trainees then you really aren't being fair to those trainees. And as for trainee #1 in your argument, it's unlikely that they are going to get 3 easy scenarios by pure luck. And if that's a concern, give them 4 scenarios instead, or 5, or 6. You do know how averages work right? I think ultimately, TRWTF is that the admirals (and many other educators), what an easy and objective way to measure somebodies performance that doesn't require them (the educator) to spend time and really think about their trainees. This is why multiple choice exams are so popular, but despite what students think, they really aren't fair on the students.

  • Anonym (unregistered)

    This comment is random, but exactly the same every time you read it.

  • Anonymous (unregistered) in reply to pink_fairy
    pink_fairy:
    On the other hand, I'd purely love to find out what Americans use for "alternate" when they actually mean alternate...
    In American English, when alternate is meant as an synonym of alternative, the "a" is pronounced as a scwha, and when British meaning is meant, the "a" is pronounced long.
  • Cow orker (unregistered) in reply to Anon
    Anon:
    * Presumably there's a file with the simulation setup (number of enemas, type of enemas, location of scenario), but the seed in there.
    FTFY
  • jumentum (unregistered) in reply to Captain Crunch
    Captain Crunch:
    AT:
    ["Unexpected" and "military" are two words that usually don't go together. In the military, every plan has contingencies and every contingency has plans, and each of the plans have contingencies all their own, ad infinitum.] - Daily WTF

    [No battle plan survives contact with the enemy] - Helmuth von Moltke the Elder

    I'm going to have to side with the German on this one.

    [In preparing for battle I have always found that plans are useless, but planning is indispensable.] - Dwight D. Eisenhower

    Yeah! But what does this have to do with kangaroos?

  • Chris Doffer (unregistered)

    Random comment is not random.

  • (cs)

    TRWTF is the admirals did not request for an "I WIN!" button.

  • Skyhigh (unregistered) in reply to Kermos
    Kermos:
    The only WTF I see is on the developers side for not realizing this and taking the word random a bit too literally.

    I'm sure others have quoted this....but...random is random.

    Ask for what you want and you'll get it. Say what you mean and mean what you say.

    I think it's more a customer requirements/spec issues.

Leave a comment on “More or Less Random”

Log In or post as a guest

Replying to comment #:

Log Out

« Return to Article