PDA

View Full Version : Tervigon statistics - warning stats heavy thread!!!



Hades Alpha
02-16-2010, 03:29 PM
Hi everyone

I was wondering how long can a tervigon spawn termagants on average and how many would be spawned. So I made a computer script which simulate the whole spawning thing. Basically the program roll three dices, checks if there's any double and so on. I made a couple of simulations (more than a million) and here's the result:

On average, a tervigon will spawn for 2.22 turns giving you 23.36 termagants (total).

Now for those who wants more (heavy part):
95% of the time, a tervigon will spawn for 2.22 ± 3.05 turns giving you 23.36 ± 33.65 termagants (total).
or
68.2% of the time, a tervigon will spawn for 2.22 ± 1.55 turns giving you 23.36 ± 16.89 termagants (total).

Provides he survive this far.

That being said, you do what you want with this. I just though it was noteworthy.

'Now you know... and knowing is half the battle'
-famous G.I.joe quote!!!

fuzzbuket
02-16-2010, 04:07 PM
Now you know... and knowing is half the battle'
-famous G.I.joe quote!!!
or a TF g1 cheesy advert:p

I like the your math hammer but the dice are weird (i play better spacehulk with painted fig's and if the stealers are unpainted they win!:confused:

Mobious
02-16-2010, 05:12 PM
Good looks! I have been meaning to do the numbers, but you seem to have it done nicely.

Dark_Templar
02-16-2010, 05:24 PM
Now you know... and knowing is half the battle'
-famous G.I.joe quote!!!
or a TF g1 cheesy advert:p

I like the your math hammer but the dice are weird (i play better spacehulk with painted fig's and if the stealers are unpainted they win!:confused:

It's a stealth thing. Less friction, more light absorbing. The Marines are dead before they have a chance to power up their rocket pants (that's a wargear option right?)

Anyways, this thread is handy, as I have been wondering whether it was worth taking a Tervi considering the possibility of rolling a double on the first roll and having him sit there looking ugly the rest of the time.

Darkwynn
02-16-2010, 08:13 PM
great, you pointed out the bell curve with it :)

Tynskel
02-16-2010, 08:57 PM
That is a FAT SD. You should calculate out to 95% (two Standard Deviations for those that don't know statistics).

Nice that you ran a million iterations... I guess you took into account for everyone who uses the tervigon, not just yourself! ;)


By the way, some of your math(s) are slightly wrong. There are no negative turns, and there are no negative termagants! However, it would be funny to setup models on the board, then remove some before the game started!

Fizyx
02-16-2010, 10:11 PM
I find it surprising that off-center probability curve has a 2-sigma exactly twice 1-sigma. Have you plotted a curve?

Fizyx
02-16-2010, 10:52 PM
Here, I just whipped this up real fast. As much as I hate MatLab, it is easy to use. This is why I HATE statistical means for probability distributions. Even though the statistical average is 23-ish, the fact is you are more likely to get less than that. The reason the average is so high is because of the fact that if you DO pull off three or four rolls you will roll significantly higher (1.5 - 2.0 times two rolls, obviously) Notice the peak of the probability curve is MUCH lower than 23, actually almost half that! This does nothing for the fact that even with 100k iterations, the distribution curve is hellza flat.

Also, notice my curve goes out to 100? That means it is including the small chances of 7+ successful non-double rolls, skewing the average far to the right.

In any case, I'll be bringing the Tervigon for it's capability to capture objectives, not for potentially bringing out 20+ Termagants.

Hades Alpha
02-16-2010, 10:56 PM
By the way, some of your math(s) are slightly wrong. There are no negative turns, and there are no negative termagants! However, it would be funny to setup models on the board, then remove some before the game started!

Yeah I know. That's the funny part of the math. Like the termagant would go back to their mother!!!

@ Fizyx
I'll get back to you with this.

Thanks everyone for your nice comments. Hope it helps you.

Tynskel
02-17-2010, 01:47 AM
Fizyx looks more correct to me. I just knew that the tervigon wasn't THAT good. This makes much more sense.

I haven't used Matlab yet- I have used Interactive Data Language (IDL) in the past.

Hades Alpha
02-17-2010, 07:07 AM
Here, I just whipped this up real fast. As much as I hate MatLab, it is easy to use. This is why I HATE statistical means for probability distributions. Even though the statistical average is 23-ish, the fact is you are more likely to get less than that. The reason the average is so high is because of the fact that if you DO pull off three or four rolls you will roll significantly higher (1.5 - 2.0 times two rolls, obviously) Notice the peak of the probability curve is MUCH lower than 23, actually almost half that! This does nothing for the fact that even with 100k iterations, the distribution curve is hellza flat.

Also, notice my curve goes out to 100? That means it is including the small chances of 7+ successful non-double rolls, skewing the average far to the right.

In any case, I'll be bringing the Tervigon for it's capability to capture objectives, not for potentially bringing out 20+ Termagants.

Okay, you are right. I assumed a normal distribution and it ain't. The average remains the same (23 termagant) though. Median is 17 termagant. Frequency peak is somewhere between 10 to 15 termagant. Now, what I think is more helpful to say is that a tervigon will spawn between 11 to 31 termagant (50% of all cases).

Cheers

Fizyx
02-17-2010, 09:31 AM
Now, what I think is more helpful to say is that a tervigon will spawn between 11 to 31 termagant (50% of all cases).

Cheers

I don't want to sound like I am being contrary, but I would really have difficulty passing off this statement. While it is technically true (that is, integrating from 11-31 gives you roughly 50%) saying "You have a 50% chance of spawning between 11 and 31 termagants" implies you have an equal chance of both 11 and 31, which is clearly not the case. I only say that because most people assume a normal distribution, which this is not.

I did run the simulation again, capping the spawn rate to 7 turns, but the numbers didn't change that much. The Weibull distribution is still skewed slightly to the right, but it is not as bad and definitely better than when I tried to fit a double gaussian.

I think if anyone asks, an appropriate response would be:


The statistical average of spawning Termagants is approximately 23 per game, but the peak probability is roughly 14 Termagants based on numerical analysis. The fact is that the variation is so large that you can not reasonably assume you will spawn any more than 16-18 Termagants per game, even though the average is much higher than that.

The range of 16-18 Termagants is based on the back slope of what I assumed was the dominant gaussian.

Hades Alpha
02-17-2010, 10:11 AM
This thread is getting quite serious... That's cool with me.


While it is technically true (that is, integrating from 11-31 gives you roughly 50%) saying "You have a 50% chance of spawning between 11 and 31 termagants" implies you have an equal chance of both 11 and 31, which is clearly not the case.


That's why I pointed out that the median is 17. Yes, you are more likely to have 11 termagants than 31.

That being said, I think with both agree with each other. It's just a matter of how we expose the result. Lets settle this with a friendly hand shake shall we?

Fizyx
02-17-2010, 10:35 AM
/handshake

nice doing business with you.

Tynskel
02-17-2010, 12:08 PM
Noooooo!

Conflict!

War!

Must be grabbing at each other's throats!!!!

Stop this 'peaceful' nonsense!

xomntec
02-17-2010, 02:53 PM
Thanks for this. I was trying to figure out if I owned enough termagaunts to use a tervigon, without having to rely on a pool of casualties. With your help I have concluding that I just have enough to field a single tervigon as a troops choice, but I will need to acquire more gaunts to use 2.

Xas
02-18-2010, 05:29 AM
if you already have the programm, could you prolly make a run to show how the average changes if you give the tervigon less time to spawn?

from an educated guess the numbers should not decline much if you only gave it 3-4 turns as the dominant factor is that it runs out but I'd like to know as I plan to use tervigons for objective grabbing mostly and therefore staying a few turns in save reserve is the plan :)

Fizyx
02-18-2010, 08:16 AM
If you cap the spawns to three turns, the average does change a bit, but it is all from the far right edge of the graph (it looks like oyu just lobbed it off at the 50-ish point. I would expect the same number as without.

Which is why i usually won't be pooping them out until turn three anyways.

Chaosgerbil
02-19-2010, 04:54 AM
Hmmm, your math indicates that you could spawn a negative amount of termagaunts. ;)