Bio Engineer Archive

Oh gawd - I hate to say this, but they guy has absolutely no idea about statistics I'm afraid.

Let's see - typical creature build is:

1 to make the DNA

10 experimentations

1 to make the critter

That's 12 rolls.

Chance of success = 0.95^12 = 54%

But each experimentation is 'rolled' against 2 stats, which means there are really 22 'rolls':

Chance of success = 0.95^22 = 32%

Or to put it another way -

Whenfully experimenting and building a creature there is a 68% chance of it critically failing!

That's with a 5% failure rate - which the dev seems to think is OK.

That's quite apart from the fact that there is a bug with tool/crafting station quality and how it affects success.

<sobs>

---

Tal-N wrote:

Err, no. Technically if you use all 12 points induvidually or all at once then your chance of critical failing is still 5%. Just like if you flip 12 one at a timecoin orall 12 at onceyou still stand a 50/50 chance of each one landing heads up.

---

---

marsx wrote:

Its probably not much of an issue for any other crafter ...

We feel it too. Try crit failing on a piece of RIS (wollymander fragments are razy hard to get, kimo scales, albatross feathers). Weapon smiths have their rancor teeth, armor smiths have krayt scales and night sister layers..

Droid engineers .. well we have our final product...

---

---

Tal-N wrote:

Err, no. Technically if you use all 12 points induvidually or all at once then your chance of critical failing is still 5%. Just like if you flip 12 one at a timecoin orall 12 at onceyou still stand a 50/50 chance of each one landing heads up.

---

What they said Tal .

Don't feel bad about getting this wrong - this is the mistake the Devs have made and.more seriously, is a mistake frequently made in courts of law involving evidence from 'expert witnesses'.

I only got into this stuff as part of my physics degree course, although it is prett fundamental!

The problem is that when an event happens often enough, even a 'small' chance of something happening makes it a near certainty. In this case one critical fail means a failure of the entire experiment (it's built so that further experimentation will not recover from a crit fail). With 22 'rolls'the oddsof not getting a single crit fail are against you (60-70% chance of failure overall).

In a semi-conductor junction the chances of an electron jumping through the gap are infinitesimally small - but computers seem to work anyway.................

Your chances of winning the (UK) lottery are 1 in 14 million. The chances for anyone of winning the lottery are 1 in 14 million. The lottery is won pretty much every week becuse there are about 20 million tickets bought.........

The other thing I keep banging on about, with number runs in pseudi-random number generators, compounds the problem even further, making a failure a near certainty. Unfortuanetly this is even more arcane and I won't even attempt to summarise it fully here! (I wrote a white paper on the subject 10-15 years ago which, unfortuanately, I don't have a copy of). The meaty bit if it is that the number distribution from these means that in 'small' samples, number-runs are a certainty. Common coding errors favour extreme results. Across large samples the distrubtion is even with metronomic prescision (out of 1000 coin tosses 500 WILL be heads and 500 WILL be tails - this does not happen in RL!)

Suffice to say, between the probabilities 'admitted' by the devs, the pseudo-random number generator pitfalls and the crafing tool quality bug, the odds are severely against you producing a creature that isn't borked in some way...........

(BTW in regard to the court evidence thing: In the UK a storm has blown up because a couple of hundred women have been locked up for 'killing' thier cot death babies on the say-so of flawed research suggesting that multiple cot deaths imply murder. Both the 'experts' and the courts failed to interpret the statistics correctly...........)

If we want to work on getting this changed we should start tracking Crit failures.

I sugest we only track failures in template and final clone stages not failures durring experimentation.

If Talthazar can put up a sticky of a post titled Crit Failure Tracking, Then people can add posts to it.

Posts will included:

Date:

# of sucesses in Template

# of crit fails in template

# of sucess in final critter

# of Crit Failures in final Critter

If this goes up please don't post any other data or comments, it will be easier to sort through the data if the only posts there are stats. If everyone pitches in in about a week we should have enough data to make a case.

Eede

I agree that we need hard numbers to prove our point to the developers. I would recommend that we also document the quality of the tools used in our attempts. There is a lot of anecdotal information being posted. Rather than posting number of successes and number of failures for each phase,I recommend posting number of attempts and number of failures. My results from last night:

Personal Crafting Station 42+, Crafting Tool 14+ (I need to get the exact numbers)

Templates attempted: 13

I agree that we need hard numbers to prove our point to the developers. I would recommend that we also document the quality of the tools used in our attempts. There is a lot of anecdotal information being posted. Rather than posting number of successes and number of failures for each phase,I recommend posting number of attempts and number of failures. My results from last night:

Personal Crafting Station 42+, Crafting Tool 14+ (I need to get the exact numbers)

Templates attempted: 11 Template critical failures: 4

Experimentations attempted: 70 Experimentation failures: 10

Creature assemblies attempted: 7 Creature assembly critical failures: 3