Hi!

Re-read all threads here and tried simple PP in a test to try to see a power of PP (de)teraforming.

Quote:

Also PP packets can deterraform a planet on impact (50% chance of 1 Terra click per 100 and base terra per 1000kt or part of - this only counts on minerals actually impacted), and if a planets value can be lowered to below the point that the max pop (or actual pop present) of a planet is less than 250,000 (25% for a non-OBRM and non-JOAT race), then the number of defences allowed to be operated will be less than 100. 100,000 pop can operate 20 defences...

Yesterday I got that 1000kT PP mixed packet did... 4% of gravity permaforming!

I tested this a bit more, and it looks like chances for permaforming are calculated not for each 1000kT as discussed in threads here, but for each 300kT, or at least for each 333kT. I'm getting 2-3% of permaforming by 1000kT packets quite often.

Can somebody point me to the right source of information about PP packet teraforming?

Or did I just spot a secret that PP PRT fellows tried to hide as much as possible?

Or, maybe my JRC3 version went wrong during one of the poweroffs?

I'll do more tests soon (still sitting on AR test - large one, on MM). Maybe I'll spot more details about how PP teraforming and permaforming exactly works.

Report message to a moderator

On the contrary, your test meets the quote...

"50% chance of 1 Terra click per 100 and base terra per 1000kt or part of"

So on average, a 1000kt packet will typically do 5 clicks of normal terraforming, and 0.5 clicks of permaforming. Nice!

Your result of 4 clicks terra is actually ever so slightly unlucky!

Report message to a moderator

Dogthinkers wrote on Tue, 21 February 2006 01:07

Your result of 4 clicks terra is actually ever so slightly unlucky!

He say he got 4 clicks of PERMAforming, not terraforming.

Report message to a moderator

Kotk wrote on Tue, 21 February 2006 02:19

He say he got 4 clicks of PERMAforming, not terraforming.

Exactly! Thanks.

DId more testing and noticed that immune field teraforming chances is certainly not 50% - by 1000kT, never got more than 4 teraforming, and 4 is relatively rare. In addition, small packet (100kT or so) never does teraforming of immune field, while it does permaforming there (permaforming without teraforming at all). Quite strange. Never nooticed these before (though rarely used).

I made testing environment that allows me to repeatedly get Stars! client opened after re-generating the same turn results (simple bat file with cycled "goto" ), so I'll make more tests and calculate exact chances - I guess need at least 500 results or so. Will post results here.

Report message to a moderator

Report message to a moderator

Report message to a moderator

Report message to a moderator

Quote:

What are your terra tech levels ?

For mentioned tests - highest possible (+-30%).

I made PP race with TT for testing to see also whether there is something related to teraforming as well. It needs MUCH more tests than I did to make accurate assumption about exact formula or algorithm for teraforming, and tell for sure what it is related to or not related.

For example, theory 1 - immune field most likely is teraformed for each 300kT.

However, it may be that 250kT are extracted from the 1000kT packet, then remaining 750 are divided to 200kT, not to 300 - theory 2. You perfectly get the same max 4% of teraforming (with 4% somewhat more rare than could be if 1000kT is divided to 250kT), and it explains well why 100kT packet does not do any teraforming at all.

With results I currently have both above theories give more or less matching chance values to the test results I made at the moment.

There may be more "theories" about how it works

If you got one, let me know, so I'll check that too during tests.
This wil ltake some time though...

[Updated on: Tue, 21 February 2006 09:51]

Report message to a moderator

Report message to a moderator

Tomasoid wrote on Tue, 21 February 2006 16:05

There may be more "theories" about how it works If you got one, let me know, so I'll check that too during tests. This wil ltake some time though...

The scientists who actually make experiments are more valued than the ones who make theories.

Your packets were 1000+1000+1000 = 40*(25+25+25) ?
and 100+100+100 = 4*(25+25+25) ?

Maybe you can improve your experimental data by genning statistics for other sizes of packets too?
Theories are hard to make on so few data.

Report message to a moderator

Report message to a moderator

Hi!

Kotk wrote on Tue, 21 February 2006 20:48

Tomasoid wrote on Tue, 21 February 2006 16:05 There may be more "theories" about how it works If you got one, let me know, so I'll check that too during tests. This wil ltake some time though... The scientists who actually make experiments are more valued than the ones who make theories.

I do experiments. However, if you have some theories ahead of experiments, you have some goals for experiments, and these goals help to organize experiments in a better way - you will know what trials have more priority to decrease number of available theories in general first of all. This will improve your experiments a lot - you will throw out a lot of other "obvious" tests that you can tell results for with 90% or more reliability.

Quote:

Your packets were 1000+1000+1000 = 40*(25+25+25) ? and 100+100+100 = 4*(25+25+25) ? Maybe you can improve your experimental data by genning statistics for other sizes of packets too? Theories are hard to make on so few data.

That is what I'm going to do. I'll also send packets with decay to try to get even more different figures. What I will do next is try to calculate exact chunk size for permaforming calculation, and minimum packet size required for immune field teraforming, as well as chunk size for it. It appears that a packet size that gives the best probability of 1% permaforming and never gives 2% is the closest one to the chunk size. The packet with worst 2% permaforming chances is also very close. (This is also one of the "theories", right? ) My testing environment I made allows me to make 500 hits/hour or even quicker, if sending small packets, though it will still take some time, ald larger packets testing is much slower because takes time to look how much it is teraformed (1% or 3 or more).

Report message to a moderator

Report message to a moderator

Hi!

I updated the Excel file with test results and a lot of conclusions could be already made. See Results tab in Excel file.

Who do no want to download 80k Excel file and analyze all the info, below is short summary of results:

- PP permaforming works as 5% chances per each 100kT

- teraforming of immune fields works as 75% chances per each 300kT, 100kT always subtracted from first 300kT chunk (include chances correction - chances are 200/300*0.75=0.5 for first 300kT). 100kT packet would not do immune field teraforming at all.

- permaforming of immune field IS RELATED to teraforming, however, teraforming of immune field may be 1% less than permaforming, no more. I tried to colonize the planet and see it is not the inaccuracy of the terra/permaforming report - current hab value actually stays untouched while permaforming is made.
It means that when permaforming happens, less than 300kT packet may make 2% immune field teraforming, thugh it is soooo rare that I cannot confirm it. If you manage to have 3% of permaforming by 250kT packet on immune field, and have only 1% or less of teraforming by that, let me know

- for uninhabited planets, chances are calculated for the total packet size that arrives to the planet, not for the packet that actually impacts (not for 67% of packet). Next weekend I'll do more tests, this time for inhabited planets. Will try defences and mass drivers to check if all ok there as well

Report message to a moderator

Maybe some 150kT or 250kT packet tests can better show how the not full 100kT remainder affects result?

1000kT should also cause 5 clicks perma once in 16411 tests if the formulas are true. No one is probably going to make tens of thousands tests with 1000kT to check such 5% permaform likelihood?

Report message to a moderator

Hi!

Kotk wrote on Mon, 27 February 2006 20:11

Maybe some 150kT or 250kT packet tests can better show how the not full 100kT remainder affects result?

Will do, though there is a high probability it is the same as for teraforming - 0.05 * (remainder kT/100kT). It would require more than 500 hits to test reliably. I'll try to do that during next weekend, though I already see that chances to get 2 teraformings by a 138kT packet are so small that you really rarely get it (see maximums tab in Excel file).

Quote:

1000kT should also cause 5 clicks perma once in 16411 tests if the formulas are true. No one is probably going to make tens of thousands tests with 1000kT to check such 5% permaform likelihood?

I'll try to automate tests in a better way - generate turn, dump the planet info into the file from the command line, find specific planet that I used as target, and write down the hab, then goto back to repeat in an endless loop. It will require to program a small utility, though it's not a problem. So it would be fully automated - you can run it and go away for hours.

Forget to tell - I will also try to make a setup and calculate the best approach of planet teraforming - whether multiple small packets or few larger ones are better.

Report message to a moderator

Report message to a moderator

Hi!

mazda wrote on Tue, 28 February 2006 16:57

Slightly strange that it should act like that. This is only for immune fields, correct ?

If you meant 0.75/300kT - correct, this is only for immune field.

Quote:

Tomasoid wrote on Tue, 28 February 2006 08:32 I will also try to make a setup and calculate the best approach of planet teraforming - whether multiple small packets or few larger ones are better. For planets you want to live on then there is no difference. Any terraforming seems to be done on an exact prorata basis for non-immunities. e.g. 10kT has a 5% chance of terra

See other posts in this sub-forum for PP. There was a similar question and some suggestions.

There is no difference from chances point of view - I agree. However, when you try to split packet and see what happens from different points of view, such as chances distibution throughout the scale with the total teraforming % done, you will see quite interesting pictures to compare between one and multiple packets with the same total amount of minerals

It looks like sending 10 150kT packets or so is better than sending one 1500kT packet for maximums and distribution reason. 10x150kT packets may do max 20 teraformings, while one 1500kT - only 15. One 1500kT packet have small chances (less than 0.5) to get 10 or more teraformings, while 10x150kT have much better chances to get the same 10 (more than 0.5) This is because each 150kT packet have 0.625 chances to do at least 1%. I may be wrong - will se in tests.

Also, it seems it is bad idea to send 210kT packets or something like that because losing minerals on these 10kT - these 10kT have very little chances to help. On the otehr side, you cannot always send mixed packets because geramium usually is required elsewhere So you are usually limited to 70, 140, 210, 280 etc. packets, or packets with mostly one type of mineral with few (1-2) mixed joined to it.

Report message to a moderator

Tomasoid wrote on Tue, 28 February 2006 15:50

There is no difference from chances point of view - I agree. However, when you try to split packet and see what happens from different points of view, such as chances distibution throughout the scale with the total teraforming % done, you will see quite interesting pictures to compare between one and multiple packets with the same total amount of minerals It looks like sending 10 150kT packets or so is better than sending one 1500kT packet for maximums and distribution reason. 10x150kT packets may do max 20 teraformings, while one 1500kT - only 15. One 1500kT packet have small chances (less than 0.5) to get 10 or more teraformings, while 10x150kT have much better chances to get the same 10 (more than 0.5) This is because each 150kT packet have 0.625 chances to do at least 1%. I may be wrong - will se in tests.

Yes. There are some "interesting" statistical effects.
But if you intend to do a lot of terraforming then you will get the same net result overall. It will all average out to that 50% per 100kT ratio.

Yes. Splitting into smaller packets increases the chances of getting more terra (at the expense that there is more chance of getting less terra as well, see below). However it won't change the average.

To take your example.
You have more chance of getting 10 terraformings with the former but also more chance of getting ZERO terraformings.
Both 10x150 and 1x1500 have the same average of 7.5 terraformings (do you believe me )

You only really need one example to show yourself this.
Compare 1x100 packet with 2x50 packets.
100kT packet = 50% chance of 0 terra and 50% chance of 1 terra.
2x50kT packets = 9/16 chance of 0 terra (more than 50%), 6/16 chance of 1 terra (less than 50%) and 1/16 chance of 2 terra (certainly more than 0%).
Average for both distributions is 0.5 terraformings.

You have more chance of getting the fringe values, and hence less chance of getting a "middle" value, but the average never changes.

T

Report message to a moderator

mazda wrote on Tue, 28 February 2006 18:30

To take your example. You have more chance of getting 10 terraformings with the former but also more chance of getting ZERO terraformings. Both 10x150 and 1x1500 have the same average of 7.5 terraformings (do you believe me ) You only really need one example to show yourself this. Compare 1x100 packet with 2x50 packets. 100kT packet = 50% chance of 0 terra and 50% chance of 1 terra. 2x50kT packets = 9/16 chance of 0 terra (more than 50%), 6/16 chance of 1 terra (less than 50%) and 1/16 chance of 2 terra (certainly more than 0%). Average for both distributions is 0.5 terraformings.

Small correction: chances of 50kT teraforming 1% are 1/8, not 1/16.

I knew you would say above. Yes, as far as I remember from statistics, when you have trials of series of trials, they just sum up into a one long single serie of trials for particular distribution parts. So it is the same as 20 trials with 10x100kT and 10x50kT.

However, what we are comparing is "what are chances to get 10% or more" for 2 different series: 15x100kT and 10x100kT + 10x50kT. These are, indeed, completely different because distribution is different.

Calculation of distribution for larger packets is rather troublesome because lots of combinations, so I'll just do tests and see.

For smaller packets (skip if you do not like math)...
Let me show it to you on a simpler example.
Let's take 300kT vs 2 150kT.

300kT:
Distribution of chances is following:
0%: 0.125
1%: 0.375
2%: 0.375
3%: 0.125

Average:
3*0.125 + 2*0.375 + 1*0.375 = 1.5

For a 150kT packet, distribution of chances is following:
0%: 0.375
1%: 0.5
2%: 0.125

However, for a 2 trials of each 150kT packet, distribution of chances is following:
0%: only one case when both 150kT do 0%: 0.375*0.375
= 0.140625
1%: 2 cases when either does 1%: 0.5*0.375*2
=0.375
2%: 3 cases: both do 1%, or either does 2%: 0.5*0.5 + 0.375*0.125*2
=0.34375
3%: in 2 cases: one does 1% and another 2% and vise versa: 0.5*0.1

Report message to a moderator

Report message to a moderator

Actually lets dont forget the game flow. Despite throwing 15 separate 70+70kT packets at one planet gives best terraforming & permaforming final result you probably dont want to sit 15 turns to do it. Instead you throw 1050+1050kT if you have and then go and colonize it.

Report message to a moderator

Hi!

Kotk wrote on Wed, 01 March 2006 04:32

Actually lets dont forget the game flow. Despite throwing 15 separate 70+70kT packets at one planet gives best terraforming & permaforming final result you probably dont want to sit 15 turns to do it. Instead you throw 1050+1050kT if you have and then go and colonize it.

Agree. However, at the very game start you usually do not have 1000kT. So you would like to choose whether to wait and then send 1000kT at once, or send smaller packets to teraform just at start of the game.

In addition, later in the game, when you have multiple sites, you can send small packets from 5 or so planets simultaneously.

There is also an approach to send 2 packets/turn, though it costs you both minerals (need to send higher warp packet so it decays) and approximately 70 resources to re-build back higher warp speed mass driver at the same turn to repeat the process (assuming you use MD5 & MD6). Don't use MD7 for this - too costly

[Updated on: Wed, 01 March 2006 01:53]

Report message to a moderator

mazda wrote on Wed, 01 March 2006 01:20

However I'm still not convinced about this :- Tomasoid wrote on Tue, 28 February 2006 18:22 Also, when permaforming takes place, and teraforming is less than permaforming, teraforming is equalized to that. As an example to contradict it, if you do the permaforming and have no terra tech then you will not get any equalization of terraforming.

There is such thing as "limits". You cannot teraform when you already teraformed up to your ideal point, for example There is also an opinion that in certain cases you actually can do teraforming beyond your normal capabilities, though I did not tested it yet (already planned).

Quote:

I'll have a look at the figures for the 300kT packets where you get no terra. I wonder how many tests you need to do to be confident that it follows the equalization theory rather than not. You suggest it will make a difference roughly 1 in 60 tests.

Well, you can see in all "chances" tabs for Gravity and Temp teraforming that it follows the rule wery well - in 2000 hits, no even a single case, though the probability to have permaforming without teraforming is high enough to get it in 2000 hits.

As about immune field - you're right, to confirm "1% difference only" rule, need to do more tests, as well as do this with a bit larger packet (2000 kT or so) which have better probability of 2-3% permaforming without teraforming.

A lot of tests already scheduled. Will see If I can do all that during weekend...

Thanks a lot! Without your comments I certainly could miss some important points

Report message to a moderator

Report message to a moderator