For anyone interested, here's how a plot of the Bravery Experience (the number of times Panic/MC is resisted) versus how often it results in a Bravery increase:
EC Min Ave +/- SDs Max N 0 0 0.00 0.00 0 2004 1 0 1.00 3.02 10 100 2 0 0.90 2.88 10 100 3 0 2.91 4.55 10 179 4 0 4.47 4.99 10 150 5 0 4.97 5.01 10 185 6 0 5.13 5.01 10 150 7 0 6.80 4.68 10 150 8 0 6.80 4.68 10 150 9 0 8.53 3.55 10 150 10 0 8.92 3.12 10 249 11 10 10.00 0.00 10 250 12 10 10.00 0.00 10 50 13 10 10.00 0.00 10 85 30 10 10.00 0.00 10 133 90 10 10.00 0.00 10 129 250 10 10.00 0.00 10 50
Where EC is the experience count value hacked into the Unitref.dat counter for Bravery (UR), and Minimum, Average +/- Std. Dev. (sample), and Maximum are shown, together with the number of times tested. (The Ns are often nice round numbers because I would e.g. hack 10 soldiers in a savegame, then end that combat 15 times.) Don't be confused by the average; that is, you only get a 10 (or not), so the ave+SDs is actually a weighted average, as it were, of how often you get 10 or not for a given UR. (That's why the SDs are huge and pretty much =5 in the middle.) For practical purposes, it's also a percent, e.g. 10 ECs result in a Bravery increase 89.2% of the time.
As you can see, the progression is not particularly smooth. But that's how the data was. Don't ask me why they chose 11 for experience counters and especially this, but they did. As you can see, I zeroed in around that 10/11 border with a bit of additional testing, but it held up.
Since Bravery is a pretty boring statistic all in all, I didn't do anything more with it except to make that simplified statement on the Bravery page.
For the record: The above was for varying values of Bravery, as long as they were less than 100. (Which is actually stored as an "inverted" single-digit value which equals a x0 value - see Unitref.dat and Soldier.dat.) As for other primary stats, I found that the current Bravery value (0 to 90) had no effect on the increase, as long as it was below the cap of 100 (data not shown).
---MikeTheRed, 2 July 2006
For anyone interested, here are a few observations from the Panic testing I just did (see stats under this page's Bravery Training). For lack of a better place, I've put them here:
- 27 of the 104 soldiers never moved or dropped their gear, when going from Morale 0 to 60. (They always froze or berserked, their four times each while going from 0 to 60.) This little bit of datum may help piece together odds of Berserk (36%) versus Panic Run and Panic Freeze (read on). I did not track what happened each Panic/Berserk though; that would have been a lot more trouble than just getting Bravery experience counters (and noticing whatever else I could) for 0 to 60.
- Every soldier who ran, always dropped their gear. Both hands (all soldiers had both hands full).
- Only on fairly rare occasions did soldiers stay still but drop their gear. Also on rare occasions, they only ran a short distance. Most of the time when they ran, they ran to the limits of their TUs. (I had modified everyone to 40 TUs so they could run some, but not real far.)
- Therefore there probably are three possible results, from the programmer's side of it: Berserk (never drops), Panic Run (always drops), and Panic Freeze (never drops). And apparently the distance to run is usually full out, but occasionally is short or nothing.
- Soldiers who ran, always ran to the south and/or east, where southeast is "down your screen toward your keyboard". True, there was a UFO directly "up" (northwest) from their start point, but they never went west or north, even after they were far from the UFO. Or could it be they were running away from the UFO?
- On at least three occasions, I saw soldiers run through solid walls of other soldiers. This probably happened when a "Panic Run" order was given, but all paths south and/or east were blocked. (The only times I noticed it happening was on the first turn when everybody was Morale 0, bunched together with some soldiers blocking others' route to the south and east. After that turn, there were always "holes in the wall of soldiers" that they could go through, and they would.) This could be tested fairly easily, if anybody wants to. (Put 8 soldiers around a central one; set Morale of center one so only he'll Panic.) I wonder what they'd do inside a UFO?
---MikeTheRed 13:25, 1 November 2006 (PST)
It may be the UFO the troops run from, or it may be an alien? The closest or furtherest one maybe?
Fwahaha! New grey detector!
As for why the troops sometimes don't move, my bet is a random location to run to is set, which sometimes turns out to be where the unit already is.
- Bomb Bloke 05:53, 6 November 2006 (PST)
It could be, BB. It's hard to draw conclusions on these other things since I wasn't really testing them. There was only one alien left alive, and some troops were "above" him (to north and/or west), but everybody to a man still went south and/or east. So I'd tend to think it's either south and/or east regardless, or they're running from the UFO. Maybe they run from whatever's the biggest (group of) alien thing(s) they can see? So maybe it's a new UFO detector, lol. Good point. Or maybe it's not.
That could be true about a random location to run to. But that may be bound up with how far they run, making it be two ways to say the same thing. Who knows? --MikeTheRed 07:10, 6 November 2006 (PST)
I always thought that panicked soldiers ran towards the nearest prox grenade? In any case, you can use a wall of soldiers to pen panicked soldiers, in the back of the skyranger for instance. - Egor
- Just did a simple test on this to check a theory of mine. Edited 25 soldiers to have 0 initial Bravery (makes it easier to see differences), crammed them all on an Avenger and changed their Bravery Experience Counter to 1. Then I ended the mission and tallied the results. Did this 40 times for a total of 1000 separate data points. My theory was that the expected count should be (1/11)*1000=90.90 and thus the percentage should be 0.9090%. From the data I get an actual count of 91. So that's close enough. Basically if this holds (and it should) the breakdowns are the following:
EC Improve % Fraction 0 0.00% 0/11 1 9.09% 1/11 2 18.18% 2/11 3 27.27% 3/11 4 36.36% 4/11 5 45.45% 5/11 6 54.55% 6/11 7 63.64% 7/11 8 72.73% 8/11 9 81.82% 9/11 10 90.91% 10/11 11+ 100.0% 11/11
I'm going to collect 1000 more values for 1 EC (for a total of 2000), then move on to 2 EC to see if soldiers improve Bravery 18% of the time. Most likely, the actual is going to be very close to the anticipated so the above table should hold. --Zombie 16:52, 26 March 2009 (EDT)
With 2000 values, the %improvement rose a little bit to 9.35%. Still within range though. For the EC of 2 I have 1000 values and the improvement is 20.4%. Still a little high so I'll gather another 1000 which should even everything out. By the way, the article page wasn't really clear on this aspect which is why I'm testing. The ~9% increase per EC point is correct, but it doesn't directly spell it out. I'll add the above table to the article page to clarify it (hopefully). --Zombie 20:02, 26 March 2009 (EDT)
- Hi Zombie, nice work. You had more patience than I, smile. My data must be more variable (above) simply because of the smaller groups. Your first 1000 was incredibly tight... there is a rule in probability that as sample size increases, you get closer to the true value in relative terms (percents), but farther in absolute terms (counts). For example, consider flipping a coin 100 times. Then 1000. Then 10000. It's quite surprising when one hits the nail on the head with large groups, but of course, it can happen.
- One thing that I did was to record the results as batches of soldiers. This let me get variability estimates (standard deviations). It's helpful in understanding, especially if something doesn't agree with what is expected... is there so much variability that the expected result is explainable by ("lies within") the variability? I'm sure your 9.35% was well within the standard deviation... but you didn't post it. FWIW, in hindsight, I shouldn't have tallied my results (above) by the average Bravery increase per se (averages of +10), but simply by the number of increases. Because otherwise, it messed with the values of my standard deviations, as I pointed out.
- Anyway, good work! -MikeTheRed 14:57, 6 August 2012 (EDT)