War Room High Scores & Strategy

Talk about Advance Wars DS. Debate, laugh, cry, argue about everything to do with our favourite tag game here.
User avatar
DxDyDzD
Rank: Blue Bomber
Location: War Room

Re: War Room High Scores & Strategy

Post by DxDyDzD » Mon Feb 15, 2010 12:55 pm

Edit: I was wrong about how the RNG works, but here are my thoughts based on how it does:
According to the author of the AW2 TAS, there are only 2 ways to abuse the RNG: by delaying when selecting an enemy to attack, and moving diagonally. I don't know how this will work out for the AI, but the main point is that because there are so few ways to manipulate it, you may not be able to advance the RNG enough times to make both attacks end in enemy destruction.
Here is an example: Suppose I need to get an RN of 254 or 255 to succeed, corresponding to 1 in 128 odds. If my first success is with 255, I must get a 254 next. I made a linear congruential generator with the modulus 256, multiplier 25 and increment 7. I seeded it with 184, so the first RN produced was 255. I checked and with this RNG, 254 appears as the 178th RN if you let 255 be the first. 255 appears again after 256 RNs. Even if my first success is with 254 (seed with 143), my next success comes 79 RNs later. You can't possibly advance so many RNs ahead as far as I know, so you will not succeed in destroying both Infantry.

tl;dr : AWDS is not truly random, so the 1 in 10000 odds are not accurate, it may even be 0 probability to succeed.
Of course, this is just an example and the AWDS RNG may not use the same modulus/multiplier/increment as I did. I feel that if you want to try something crazy like finishing Fortress Isle in 2 days, you should do a bit of hacking research first, like this:
1) Find out the memory location of the random number. It's already been done for AW2.
2) Observe enough iterations of the RN. One full period (256/65336, depending on the number of digits you think the RN uses up) is enough.
3) (optional) Derive the formula for the RNG. Pokemon Emerald uses a LCG and it's already been cracked, so why not AWDS.
4) Search through the random numbers produced to see if your task is possible or not.
You should also find out if there are other ways to manipulate the RNG. Maybe it's been changed from AW2.

With your tests on the emulator, the different results can be attributed to human error. It's not likely that you can press the buttons perfectly to within 1 frame of accuracy, so maybe it takes 34 frames to end your turn the first time, then when you reload, you take 36 frames instead, so the RNGs will be different. Save states save every single detail, so you should be getting the same results if you do everything exactly the same.
Or it might be possible that Desmume doesn't save the DS's internal clock, so when you load a save state, it grabs the clock reading from your PC's internal clock. Then, if the RNG is based on the internal clock, you'll get different results. I don't think this is very possible, as the RNG is usually seeded as soon as the game boots up or you soft reset.
Also, a bit of statistical nitpicking, you can only multiply the probabilities if the outcomes are independent. If my hunch is correct they're not so the probability of success may be 0.

edit: Pivot Isle Jake 6/300, I haven't got the all-CO strat yet. I needed a +10 boost on plains to OHKO a 4HP T-Copter with a 9HP Recon.
My strat was to build 2 Recons every turn, then block bases like a pro. On D3, block the east base. On D4, move the Recon on the east base to block the airport (otherwise a B-Copter will be built), then block the east base with another Recon. The Infantry created on D5 needs to stay on the base it was built from to attack your Recon. I had just enough Recons needed to win (7), -STL- said he had 2 Recon attacks left so I'm doing something wrong.
I could do better by abusing luck so that my Recon has 9HP after the counterattack on enemy D2, so I can do more damage to the T-Copter. Still not sure how Sami and Colin can do this unless you're blessed with lots of luck.

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Tue Feb 16, 2010 12:10 am

Ok, once again, the RNG on Enemy Day works different (are not linear) from those on ur day, i didn´t made a "human error", i made a save state directly before the Enemy attacked and got different luckoutcomes from reloading, besides that the hard point on FI is not the killing it´s the getting attacked. The kill rate on Hawke is 44,2% on Andy 21,36%. We didn´t made enough/serious tests to know the real value of the chance of getting attacked, neither would it help us at all.
The frustrating things about FI are that so rarely something is happening and even if u had 9999 failed attemps it doesn´t increase ur chance in the 10000th try

On Pivot fidel posted enough info to find the strat i think. it was pretty straightforward and required not much luck, not even with Sami, Colin can build more and has even an easy strat with killing Enemy Tank. My tank killing strat required so much luck with normal COs >_>

PS: U can leave those "-" away, just say STL, i had some probs while registering, that´s why they are there.

User avatar
DxDyDzD
Rank: Blue Bomber
Location: War Room

Re: War Room High Scores & Strategy

Post by DxDyDzD » Tue Feb 16, 2010 1:29 am

Yeah, I just tried Pivot Isle Grit vs Colin in Versus on an emulator and made a save state on D2. I then pressed A, up, A, ending my turn and the AI captured a random city (SW or NW) each time. I guess you're right.
I know that a run of bad luck doesn't mean you will soon have good luck. In sports it's known as the hot-hand fallacy. I've heard stories of people pumping coins into slot machines based on this fallacy. :geshorts: shakes head

Also, I made the D2D for Pivot Isle, I've only tried it with Javier, but Hawke, Von Bolt, Max and Jess should work fine as well. Grimm (unlikely though) may have problems following my strat. Colin and Sami might need more work. I think Grit can 7/300 this map, just block all production facilities with your Recons on D6 and rout next turn. In fact, that might have been swifty's 7//300 strategy. Kanbei...hopeless, he will never perfect this map.
minor tidbit: When I turned visuals on, because Eagle was wearing his alternate outfit, I noticed that when I attacked the T-Copter with my Recon, the most the health bar dropped was till the same level as his spectacles.

Code: Select all

Day 1:
Build a Recon in the south base, and the base 2S of your HQ.
Day 2:
Move the Recon on the south base 1N 7E.
Move the Recon on the other base 7S 1E.
Build a Recon in the south base, and the base 2S of your HQ.
Save.
Eagle Day 2:
His Infantry must start capturing the southwest city.
Day 3:
Move the eastmost Recon 1N 6E and attack the Infantry.
Move the Recon on the south base 1N 7E.
Move the Recon on the other base 7S 1E.
Move the last Recon 1S 6E.
Build a Recon in the south base, and the base 2S of your HQ.
Day 4:
Move the west Recon of the pair 1S 7E and destroy the Infantry.
Move the east Recon of the pair 1E.
Move the Recon on the south base 1N 7E.
Move the Recon on the other base 7S 1E.
Move the last Recon 1S 6E.
Build a Recon in the south base.
Save, then move the damaged Recon 4E and attack the T-Copter (x-7).
Save.
Eagle Day 4:
His Recon (on a city) must attack your Recon (on a shoal) (7-8).
He must build an Infantry, not a Mech.
Day 5:
Move the Recon on the east base 3N.
Save, then move the 7HP Recon 2N 4E and attack the T-Copter (7-4).
Save, then move the east Recon of the trio 6E and attack the enemy Recon (9-6 or 9-5).
Move the west Recon of the trio 1N 7E and attack the enemy Recon (10-2). Turn visuals on to estimate if you have taken little counterattack damage.
Save, then move the middle Recon of the trio 1N 7E and attack the capturing Infantry (10-5). Turn visuals on to estimate if you have taken little counterattack damage.
Move the Recon on the south base 1N 7E.
Move the last Recon 1S 6E.
Save.
Eagle Day 5:
Both the Infantry on the northwest city and the west base must attack your Recons without moving (x-6).
Day 6:
Save.
Move the east Recon of the pair on the bridge 8E and destroy the Infantry.
Destroy the T-Copter with the Recon on the east base and the Recon 1W of the west base.
Destroy the Recon with the Recon 2W of the west base.
Destroy the capturing Infantry with both 9HP Recons.
Destroy the last Infantry with the 10HP Recon.
Aside: what other scenarios of insane luck have you encountered in this game? I recall the old Pivot Isle strat for Flak was something like 1 in 600+ odds and the Risky Vale strat has bad odds as well. Anyway Map 3 of Money Survival is tedious, just like Fortress Isle (but M3 has no luck involved), so if you're feeling depressed, just think "I had the patience to charge powers for 999+ days, I can do this!"
Last edited by DxDyDzD on Sat Mar 13, 2010 10:15 pm, edited 1 time in total.

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Tue Feb 16, 2010 3:59 am

My Bankers 1,0 strat took me the most time (4 days for the double suicide, about 20-25 playhours), but Fidel found a strat that worked in 1-2 hours...
On the other side Fidel´s Lost Basin 2,0 strat with 2 Money, Star and APC Boost took him 2 weeks, while using 3 Money and Boost my strat takes only about 30 min-1 hour, and even allowed 1,5 imperfect. ( 3 Money equal Colin, while 3 Money are slightly better (80% compared to 79,1%))
So yeah Risky Vale, 1,0 is a joke, 1,5/2,5 aren´t too hard , only 2,0 is the tough one.
So if we talk about hardest nessecary strat it´s clearly Risky Vale 2,0 now.
Twin 2,0 was hard but not insane. Pivot 2,0 was very annoying because the 6 (iirc) visual is messed up. And we always need to do high luck to kill Base Tank with just Recons.I remember all full HP/9HP Recon attacks added up to exactly 100%, ur room came only from 1-2 low HP Recons, i had 3-4 Recon HP in latest strat i think.

Btw fidel could u remind me our wanted skills for FI 2,0 2 days pls?

Yeah i had patience for M3, but i knew when i play my 2000-3000 days i have my double TAG. I also had the patience for Bankers 1,0 for 4 days , but here the attacks came more often, so there was at least something happening.

fcastro

Re: War Room High Scores & Strategy

Post by fcastro » Tue Feb 16, 2010 4:01 am

Mmm it has nothing to compare with money map 3 <_<... The point is that u can start going crazy, I even sometimes kept on playing FI in my dreams <_<...

I'll make a post about lucky strats in perfect scores thread, maybe after STL gets FI. The question is to know how we can count it, u'll note that Spann with Javier is counted (its 3 times min luck, 1 to 1000) as I made something similar with Hachi, but FI NB/2.0 arent counted and not Risky Duo for Koal, Sasha and Sonja too. Dont worry STL, I dont forget them, (and neither Pipeline, but how the hell did u find 2 kills at same day) they need some play, did u take a look at it DxDyDzD?
Also something just comes to my mind, the first visual is 97 right? So in Spann with Javier we could save after attacking first inf to 5 hp, and the strat would then only need 1 to 100 luck. ^^ (if we ignore the other things ofc).

Kanbei in Pivot wouldnt be hopeless if inf would attack his Recon. My hope was that inf would get a chance to attack if Recon previously received a high counter when attacking city inf, I tested a bit but failed, would u bother test too DxDyDz?

For Risky Vale, the hardest strat was Jugger 2.0 as he couldnt charge his 7 stars SCOP, so we have only COP to suicide inf. Kanbei's Inf has an attack of 14, reduced to 12 thx COP's def so we keep 9 hp at 90%. Then to make him 100 damage, we need 100/0.8 so 125 damage, but as we have 9 hp we'll need to do 139 damage (125/0.9=138,88...). As we COP Recon is 77, so there's still missing 62. Jugger has lucky skill so the six good cases are (64,0), (64,1), (64,2), (63,0), (63,1), (62,0), and there are 1600 possibilities... So finally as there were 90% to keep 9 hp, we are around 1 to 296 chances, ofc only when Kanbei attacks and there are also two little suicides asked (but happening most of the time)... Took me like 2 days iirc.

PS: FI 2.0 skills were +12 def for keeping 10 hp Tank from 2 inf attacks, luck, and 2 money skills for building 2 Tanks iirc. Did we ever math Flak? XD

Edit: Sry Flak obviously cant follow that strat, but could he get another?

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Tue Feb 16, 2010 4:13 am

But at Javier Spann u can save between or not? Anyway the Visuals are 97-100,94-96 and 91-93.
I think we can count them in if we have 100% conformation from the math and saw all enemy moves. The Risky Duo strats aren´t 100% checked.
Instead of Pipeline u mean Silo Nation i guess, full HP Inf killed 6HP Inf SE, he went there by foot, the enemy 6HP Inf took a counter from my Cap Inf that i droped on Day 4 on Mountain, u ahve to keep 7HP after first attack so u get attacked by 2 Inf. the 2nd kill is NE, Enemy Inf attacks my Recon b4 Tank, so Tank must be late, then injured Enemy Inf attacks my Inf on Day 5 and we reattack him and kill him this way.
About Kanbei Pivot, i´m very sure it will never happen, AI will only attack Kanbei´s recon if it can make 2HP dmg. So the Ai must have Lucky Skill/Lucky CO or 134+ Att for Infs, i tested that srsly.
Thx for 2,0 info, i will check it out, math the exact number for kiling and if it´s even possible (if the AI attacks)

fcastro

Re: War Room High Scores & Strategy

Post by fcastro » Tue Feb 16, 2010 5:32 am

Pivot: "AI will only attack if he can make 2 damages". Thats exactly what I thought, if Kanbei as a low 10 enemy can hit to 8...

Silo (not Pipeline lol): Mmm cant we kill the Recon? XD Btw how do u play with him? I let him on city, hit him to 5 hp and had my 4 infs around him down to 1 hp, but I guess we could hit him for more...

Spann: At D3 Tank attacks inf, must receive min counter. Then at D3 enemy he gets min attacked by Tank and at D4 reattacks an inf and must receive min counter again.. So yeah u can save but it will always be blind as nothing is shown :-/...

Fortress 2.0: do u seriously want to count it for Jugger? This strat looks like such a joke XD

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Tue Feb 16, 2010 6:07 am

Ai counts also partial dmg it makes (not sure about the partial HP of a Unit)
Proof: 130/80 Grimm is not too rarely attacked on Pivot, while 136/100 Sonja is never attacked in FI and her bad luck could make her min dmg to 91% just like Grimm´s.

On Silo i appeal Recon next to my City, i let the Inf which got attacked by Recon and 1 Inf that attacks him on Day 5 repair on Day 6, all my Inf have 0 or 1 HP and the Recon had 4HP while i think i can hit it to 3 HP with more luck involved.

On FI 2,0 i would prefer to find a way to outrule this ridicilious strat honestly...

fcastro

Re: War Room High Scores & Strategy

Post by fcastro » Tue Feb 16, 2010 6:15 am

Take also in consideration that SCOPing Nell is never ever attacked while she can theoretically have 0 luck.
Anyway the question is if AI is interested into doing exactly 2.0 hp damage or in hitting Recon to 8 hp. Just like u, I think Kanbei's recon will never be attacked but that it deserves a last check.

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Tue Feb 16, 2010 6:39 am

SCOPing Nell is never attacked? On which test do u base that? on Risky Vale 2,0 she finishes with SCOP suicides too or not?

Ok here are the values for FI 2,0

to keep 10HP with both Tanks is a chance of 3%, Hawke kill is 20%, Andy kill is 1%
so the value for killing all assuming all 4 attacked which alone is surely harder then FI 1,0 is 3,6 to 1 billion
or 1 to 277.777.777,77 etc, 1 to 277 millions
so let´s say offhand the attacks would happen 1 to 20 we would be at freaking 1 to 20 trilllion lol, every human in the world would need to play it about 3000 times to have it once lol.

PS: fidel, could u pls resend me our MSN speakings about GG which u sent me already? I lost them when i formatted my PC.

fcastro

Re: War Room High Scores & Strategy

Post by fcastro » Sat Feb 20, 2010 5:58 am

Well I cleared the "Spann program"; it was so much quicker than expected! 0_o And not in the expected order...

1) Colin got an easy 5/300 thanks a D1 Recon. He's even able to end on his turn!
2) Then I got back to Hachi and found him a normal strat using Recon too.
3) After that I took Andy and found how to always get Recon moving very west: just let him 3 hp and appeal with an inf. I wrote wrong things about Andy's strat, in fact D1 Tank attacks city inf and gets reattacked to 9, then the same Tank attacks the Recon (who can thus keep 9 hp). He's attacked from mountain then for a 8-7, and healed by COP later. 1 hp inf also heals to 3 for hitting mountain inf to 5 hp, which is required for getting him suiciding. 5/300 done.
4) Then it was Lash's turn, she was very easy as she didnt need both 10 hp tanks, I believed she was insane because sometimes I saw enemy's infs refusing to suicide at the last day but it wasnt the case anymore.
5) And finally Hawke, this time it was pure luck, the strat is supposed to be hard but I got it so quickly...

So only Javier, Drake, Kanbei and Grit are remaining. I might not count Javier anymore as it wouldnt change the ranking. XD

On the contrary Ridge 2.0 had very bad news. I tested a bit in VS mode the 8 hp plain infs against plain recon (thx Drake's COP). They seem to never attack 100 Recons, while they attack Colin's Recon well. I might test that more but it sounds hopeless. So I tried to play Flak/Nell for real but they look so boring, apparently Jugger's success was very lucky. Nell asks 10 hp plain inf suiciding in plain recon. Thus: 10% to keep 10 hp, 35 luck required (around 41%), 90% to keep 9 hp, 48 luck required (20%). So when inf attacks (not seen yet except in VS test) she kills only at 22%... =(
Flak's strat is played with 9 hp inf, already got the attack twice but the inf survived... 30% to keep 10 hp, 22 luck required (it's 190+9x20 to 1000 ie 37%), 70% to keep 9 hp, 33 luck required (around 15%). So finally he has 22% just like Nell <_<... But Jugger has 22 luck at 38% (300+9x25=525 to 1375) and 33 luck at 18% so he's not that better... 24%...

Oh and I wrote something very wrong about Pivot 2.5, ofc getting a mech is very bad. But it made me wondering, could it be possible to charge Eagle's COP against mech in 5 days? That would give 1.5 or at least easier 2.0... I'll test later.
Last edited by fcastro on Sat Feb 27, 2010 4:35 am, edited 1 time in total.

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Sat Feb 20, 2010 6:55 am

Very nice Spann achievements :)
But i think Javier must be counted as u are sure it´s possible.
Now Drake with Tank suicide? =O

On Ridge 2,0 why needs Nell a 10HP Inf and Flak a 9HP one, only cause Nell has no bad luck?
Also ur math seems to be a bit off or are u really NOT using the luck skill?

On Pivot Mech idea sound very interessting, but yeah offhand it´s hard to say if we can charge.

I also want to see Final Battle 1,0 srsly checked, the closing Base idea hasn´t been really tried yet, and there´s a very little maybe for 2,0 improvement, well i would it have rather suspected then Pivot 2,0 improvement.

And FI, i can´t even stand it for 5 min, it auto-bores me so much, are u planning to ever try it again or should we drop it for good?

fcastro

Re: War Room High Scores & Strategy

Post by fcastro » Sat Feb 20, 2010 7:45 am

Never saw a Tank suiciding in Tank, only in inf/recon...
For Ridge 2.0 9 hp inf wont attack Nell, and 10 hp inf would be to hardly killed by Flak I guess (not checked yet). Money and Move skills are necessary, so no luck skill... But do u confirm the calculs?
The prob of Rachel is that we want Rocket staying on his base, apparently it requires (at d3) to hit base Rocket to 6 hp, and that asks full hp Recon for Rachel (so only 1 inf is allowed to attack him at D2 instead of two, that will makes prob for COP charging). On the contrary Kindle is very good as the attack is from HQ (in order to get attacked by mech and inf, so all bases are closed). She also keeps 10 hp recon but for a different goal, she kills Rocket before charging thanks her incredible HQ's boost, all the others COs would need COP to kill Rocket...
For Final 2.0 it sounds interesting, can u sum up ur strats for NB? I have no idea of them... Is it a good idea to block east base with a D1 recon? Maybe it can also be made west? Also I'd like to use Sensei's COP for routing =)

For FI it's really droppable, but maybe u should try a bit more :p. I hate that map so much... Also did we check if Jugger has a human strat for 3 days?

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Sat Feb 20, 2010 8:13 am

About the calcus, ur Flak calcu is a bit wrong, i assume the Inf has 9,0 HP so u need 23 luck/35 luck that makes 35%/12% and a total success rate of 18,9%, Jugger is off in the same way and should be around 19% too.

For Final Battle all the info i have is in this post: http://www.cyberscoreforums.com/index.p ... #msg150929
As i said, i didn´t made a strat with closing a Base yet.

We should make 2 rules:
Winning in 2 days is not allowed
Destroying an army without making a single attack is not allowed

But the prob would still be that we could f.e. attack Andy with a Recon and with Jugger´s bad luck Inf could keep 7 and suicide with 9... so maybe we should say kill on our turn?
Btw there´s another map where we can destroy an army in 2 days (Treble Ridge).

fcastro

Re: War Room High Scores & Strategy

Post by fcastro » Sat Feb 20, 2010 8:54 am

I would rather ban insane luck strats, if one day we find a human 2 days strat ofc it'd be counted. So lets say from 1 to 1000 it's banned. And if FI NB is made the limit would be 1 to 10000...
Also for Flak I first counted like u but then noticed that (77+22)x0.91=90... And yes its a 9.0 repaired inf.

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Sat Feb 20, 2010 9:14 am

oh right i was wrong, sry

Pivot 1,5 FI style strat

Always 2 Recon and moving along Road

Day 3
Attack Inf from N get 9-6 or 9-5

Enemy Day 3
Inf attacks Recon 7-5

Day 4
Western 10 HP Recon kills City Inf
7 HP Recon attack T-Cop to 6 from Base
Eastern Recon only 1E to have no Tank

Enemy Day 4
This is the lulz one:

Recon attacks ur east Road Recon, u keep 7 and deal max luck (41%)
Mech is built

Day 5
Suicide west 7HP Recon in Mech, attack Mech with other 7HP afterwards from City, Mech 3, Recon 5
Kill Recon with Recon, max luck again
Then attack both 10 HP City Infs with 10HP Recon each, i think one 9-5 is enough to charge here
COP, Cleanup

fcastro

Re: War Room High Scores & Strategy

Post by fcastro » Sat Feb 20, 2010 8:35 pm

Wow insane... When do u drop FI?

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Sat Feb 20, 2010 9:13 pm

I don´t know, that wasn´t what i expected or at least hoped to read. When i drop it i don´t think i ever have the patience for the kills again, besides this strat doesn´t look much easier then FI. I rather hoped u say something like: Nice Pivot 1,5 ideas, i put in mine and now the strat is human...

fcastro

Re: War Room High Scores & Strategy

Post by fcastro » Sat Feb 20, 2010 10:12 pm

I wont check now, but probably ideas will come while playing. Btw, I got an idea for a human Spann with Javier, but Drake looks still impossible... Also, did u count ur time of playing in FI?

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Sat Feb 20, 2010 10:22 pm

no i didn´t count it but i would say less then 5 hrs

Well i can only play on Emu now, the way to make it easier is clear, the Mech should suicide, the prob could be that the weak Recon is in Mech range and he would need to attack a 9/10 over it, haven´t tested it and i must find a way charging that way as Mech also must be hit to 2 and Recon takes less counter.

I must take my statement over Enemy Day RNG back! It is time controlled like our turn RNG, the reason why we got different luckoutcome from same savestate depends only on the timing in which we press the Day X window away...
So FI might be impossible, but we can´t know if it is, that´s really sad, i hoped it is different =/

User avatar
DxDyDzD
Rank: Blue Bomber
Location: War Room

Re: War Room High Scores & Strategy

Post by DxDyDzD » Mon Feb 22, 2010 9:10 pm

-STL- wrote:I must take my statement over Enemy Day RNG back! It is time controlled like our turn RNG, the reason why we got different luckoutcome from same savestate depends only on the timing in which we press the Day X window away...
But on FI there are 2 enemies so theoretically you can control the RNG (since there are 2 Day 2 windows, one for Andy and one for Hawke). At least, on an emulator with slowdown.
If you don't like the idea of a 2 day win then you can probably list FI as a tool-assisted run. Or "theoretically possible but impractical".

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Mon Feb 22, 2010 10:16 pm

Well the point is rather that we are playing for the "perfection" of War Room, we are just not sastified if we don´t push the game to its last limits.
So it´s really frustrating that we have the FI "strat" but luck stops us, if we ban those ones we still can reach the perfection, kinda...

Btw thx DxDyDzD ur note about the Mech on Pivot gave the idea to improve 1,5, just like Little Lion/Rabbit and ofc ur Mega allCO strat gave us the latest improvement^^ Keep em coming :)

that DxDyDzD

Re: War Room High Scores & Strategy

Post by that DxDyDzD » Tue Feb 23, 2010 8:58 pm

I can suggest a lot of stuff but I'm not skilled enough to put them into practice :(.
I had an idea for Pivot Isle 2.5: I used a Recon to block the airport to stop a B-Copter from being built. If I build a Day 2 Anti-Air (cannot be Day 1 because I need to use a Recon to lure an Infantry away from the neutral airport), would I be able to stop a unit from being built in the airport when I don't have a Recon on top? Maybe this will allow Grit to perfect Pivot. If it doesn't, oh well... :roll:

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Tue Feb 23, 2010 9:30 pm

An Day 2 AA cost 2 Recons and can´t attack anything on Day 4, the reason why i dont even start thinking of Grit in Pivot 2,5 is that he can´t 2HKO City Inf with full HP Recons, so he can´t open the way on Day 3 and the D1 Recon can´t attack from/close the Base.
Kanbei would be very possible if Eagle´s Inf would attack the Recon. As i said i had 2 spare attacks with Drake so Kanbei´s missing 2nd D1 Recon which would have 2 attacks is no prob at all.

Edit: Our theories about Money in AI decision hold up! So i thought about Kanbei in Pivot, the best Eagle can do is 16% which means 640 G but 768G in Kanbei´s case Kanbei´s 8HP Recon deals 67% with 0 Luck, 670G. 768 are way more then 670 G, so why does the AI never attack?
Sadly AI ignores that Kanbei´s units cost more. As 640 < 670, the AI will never ever attack Kanbei´s Recon, which means Pivot 2,5 is impossible with Kanbei.
Then i made VS mode test, Sami´s Inf vs. Grit´s Recon Road-Road, in theory Sami deals 96% of the time more dmg, 1% equal and 3% Grit deals more, in pratice
i got 3 of 105 not-attacks ofc only counting the 1st wave as every Inf thinks twice if he attacks or not.
Then to test if the AI really ignores different Unit price i tested Sami´s Inf against Sami´s Recon and vs. Colin´s Recon, both have the same firepower, but Colin has 80% price, so we should see way less attacks on him if Ai would factor it in, i got 14 of 105 non-attacks and 17 of 105 non-attacks on Sami.
That proves AI counts Colin, Hachi and Kanbei´s Units at 100% price for this estimation.
So the only open question is if the AI attacks if damage given/damage taken is equal, but i´m too lazy to make more tests now.

Fidel u know what that means =O
NO MORE VS MAP TESTS, we simply take a calculator and can see if and how often an attack happens!

Edit 2: Well inserting this at FI says Hawke´s attack rate should be at 631 to 7500, this clearly isn´t the case. So the issue isn´t solved even though the former tests worked so good :(

fcastro

Re: War Room High Scores & Strategy

Post by fcastro » Wed Feb 24, 2010 6:13 am

Also according to what u said, Sonja/ Kanbei should be able to get FI XD (btw could be interesting to check if u get attacked when the partner isnt Kanbei)... And there's also the prob of Nell...
So maybe AI counts differently when too much luck is involving? Like he's thinking: "ok i count him making only X damage but as its too random he could make X+x damage so lets not attacking". Maybe the "x" depends of average luck and can be inducted from some test... Or u have other ideas?

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Wed Feb 24, 2010 7:47 am

Yeah, some things still doesn´t add up, but there´s nothing "too random" for an AI and it doesn´t think, just some formulas.
We know the Ai estimates a real luck and not an average luck as Jugger is attacked besides high attack where a normal CO would not get attacked.
Also i/we assume that partial HP are counted in, Max/jess are hardly attacked in Inf-Recon duel while Grimm is not too rarely attacked, the best Inf can do is 2HP dmg, but it could be aswell just that 2Hp on Grimm are more common.
The Inf-Inf duel is changed, AI Infs just refuse attacking ur Inf, if they would take about 2,5 times the damage they would deal, if we capture a prop they attack whatever as we see on Money Map 10 f.e. (Full HP AI Infs suicide in neutral Base capturing Inf while doing no damage/Kanbei´s SCOP)

fcastro

Post by fcastro » Wed Feb 24, 2010 7:58 am

? Ofc AI doesnt think, its just a way of talking... So here is my first hypothesis for the "x factor": it's just ur additional luck; like if u have luck skill and 2 stars partner its +20. Bad luck would ignored. I made some calculs and if they're correct with this factor Sonja Kanbei are never attacked in FI and Jugger Koal are rarely. Would u bother do some checkings?

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Wed Feb 24, 2010 8:32 am

i didn´t rlly unterstand what u mean, but it can´t be only about addinational luck as we then would see no randomness about 90% of the time when no addinational luck is involved, pls explain ur theory more clearly.

fcastro

Re: War Room High Scores & Strategy

Post by fcastro » Thu Feb 25, 2010 5:06 am

Well I made no tests so its just theoretic.
Assuming what u said was correct, in FI Hawke's inf will attack 136 Sonja's recon only if his counter is lower than 22x40= 880G. (I take only the case of max damage 22% to make it clearer). Recon has only 8 hp so that means he must make less than 88/0.8= 110. And Recon is 95%. So clearly, inf should attack Recon. But he doesnt.
So my theory is that AI will always add a (constant) number, ie he will count 90+x (in the case of -5 luck counted, more generally its X+x if he counts X counter), and so here the result will be higher than 110 (x>20). The 1st hypothesis for "x", if we assume that its value is 0 for a normal CO, is that its the added luck, in our example it's 25 (skill 10 and 15 of Kanbei). Eventually bad luck could be minored, this is not sure yet. Ofc u could guess the value of x with some tests.
Is it more clear?
Now if we count for Jugger in FI, we know that he is attacked, so as his min damage is 95-15=80, we know that his "x" is lower than 30. So either only skill/partner luck is counted (+20 and not +39), either it's 39-15=24.
What do u think about it?

Edit: I have an idea for Sonja's test. If her 131 recon is attacked (91%, min attack of 86), that would mean that her "x' is 20, ie added luck minored by bad luck. That would sound more reasonable, as with this formula Nell is D2D attacked like a CO with luck skill (and not in the other case because as we saw with Jugger, if we dont minore bad luck, we mustnt count "natural" luck). Ofc I could be very wrong but that sounds good no?

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Thu Feb 25, 2010 5:59 am

That sounds not good, rather strange.
From FI we can simply see that the assumement that AI attacks if it makes more dmg in G then it takes is not correct or at least not the full story.
Now for FI, Sonja´s bad luck is not enough (assuming u made more then 200 tests or so), so 91% base dmg is too high to get attacked, we know it´s at least 84% as that´s Max/Jess´s power and we got attacked.

125 Max (87%) is attacked, 128 Max (89%) never. Ai was Andy
Hawke Ai doesn´t attack 128 Max either.
Neither does Sami AI, so if AI really counts partial HP is in question now

Another theory is that the estimated counter from the Recon assumes that the Inf only made Base dmg, so 9HP Recon counter is always taken.

Not even 156 Sami (18% Base) attacks 128 Max! so partial HP are clearly not taken into account, but why does Grimm get attacked is the big question now.

In fact Andy AI doesn´t attack 130 Grimm, does the Pivot AI have a different setting? Or do i just remember bad? I´m sure i got 6/300 with Grimm on Pivot and the requirement is to get ur Recon attacked by Inf.

On the AI setting "Strike" Andy Ai attacks 151 Max!

fcastro

Re: War Room High Scores & Strategy

Post by fcastro » Thu Feb 25, 2010 11:39 pm

That sounds not good, rather strange.
Lol I was just trying to fix your theory XD. I had missunderstood that u had already done that kind of basic checkings like when Recon is attacked when u get a normal CO... So yeah, it wasnt a problem of lucky COs, sorry. Let me know ur next idea.

Btw, if 87% is the requirement for getting Hawke attacking a Recon, we can try to estimate when it happens in FI. 87% are made when Jugger gets -8 luck, so 28 to 750. Now we can guess that Hawke also wants to make 2 hp damage, which happens 3 to 10. So finally he would attack like 1 to 89, which sounds reasonable. But that means killing Hawke is now rather 1 to 200 (0.49%)... Well maybe 1 to 100 if he has a 2nd wave, which is more close to VS tests. There's also the possibility that "2nd wave" occurs in VS mode but not in FI (indeed remember, when I tried to count the attacks during FI play I found they were happening less than in VS mode).
For Andy as he has a plain the requirement is 97%. Jugger gets +2 luck in 153 to 750, and Andy makes 2 hp damage 1 to 5. So he would attack like 1 to 24, and finally kill a little less than 1 to 100 (0.87%). Yes that sounds good now!!!

Anyway, all this calculs are very beautiful but what are the next achievements? You restarted Risky Duo?

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Fri Feb 26, 2010 12:01 am

87* 0,8 is actually 69,6%, just below 7HP dmg, so i have the theory that Inf only attacks if he takes less then 7HP dmg, 70% happens to be the normal base dmg Recon on Inf but i don´t know if it´s linked. I will make more tests later and post them.
I haven´t started Risky, i was about to start Silo but got pissed how much luck it asks like all remaining things, so i take a break, i don´t know for how long.
Will u try Pivot 1,5 on DS? So it´s finally done for real XD

Yes the 70% limit holds up

Colin AI Infs (can only do 19% max) attack 110 Hawke hardly (69%) but 115 Hawke (72%) never.

Open question:

Why does Grimm get attacked in Pivot?
Why does Grit not get 100% attacked as he couldn´t even with full HP and full luck do 70%, 9HP Recon has 59% max dmg!

In war room i don´t think there are 2 waves of attack, we always saw same order of attacks except when there are late Units.

Ok i have a new theory:
The Inf must deal more dmg in G AND take less then 70% counter.
That would explain why the deal more dmg in G theory worked so exact with weak Recons and why Grit is not 100% attacked. so i would say the general formula is
Enemy must do more dmg in G then it takes and the counter mustn´t be higher then the regular Base dmg of a 100% CO 10 HP Unit.

fcastro

Re: War Room High Scores & Strategy

Post by fcastro » Fri Feb 26, 2010 1:36 am

Will u try Pivot 1,5 on DS?
Sure I'm very hurry to test it, but wont play soon.
Did you test what happens with Nell's SCOP/COP? My former idea of adding luck was clearly bad, as COPing Nell gets attacked by inf (my tests were plain vs plain). So I guess SCOPing Nell would be too, and that your very old tests with Nell (in Money Survival iirc) were only with Mistwalker?
Anyway, I'm very interested in finding the exact value of FI's luck (more than in understanding the reason behind). What do you think of my calculs?

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Fri Feb 26, 2010 2:43 am

ur calcus look good but 2 things:

Jugger´s chance to get +2 or lower is 150 to 750, When Andy deals 18 or 19% Jugger has a 21 to 750 chance to deal less then 70% dmg.

So all in all
Hawke´s attack rate is 84 to 7500
kill rate is 3315 to 7500
Andy´s attack rate is 342 to 7500
kill rate 1590 to 7500

Hawke´s A+K rate is ca. 1 to 202
Andy A+K rate is ca. 1 to 103

FI rate is 1 to 20895(,8)

fcastro

Re: War Room High Scores & Strategy

Post by fcastro » Fri Feb 26, 2010 8:44 am

Wow. So much more than what I expected when starting FI... But apparently VS tests were biased by "double wave" effect. :( Btw, how did u get 150 and not 153 with Jugger?

If u're looking for "not-too-boring" challenges, u could play Bankers and Ridge 2.0 a bit, maybe u'll find a crazy idea for new COs (well not too crazy I hope XD).
Btw we can now better estimate how hard is Nell in Ridge 2.0 (plain inf suicide in plain COPing recon).
As min attack is 62% (620G), Inf must do at least 16% damage (640G), that happens only 1 to 5. Then only +2 luck is allowed (as 79x0.9x0.9 is rounded to 63), so we have 1 to 20... So finally 1 to 100 to get attacked! And with a 22% chance of killing rate, that makes 1 to 454... New most insane allowed strat =O
I'll math the stupid robots when knowing how 9 hp inf plays.

Also u mentioned Spann 2.0, there could be some things to do there. And also... FI 2.0 if u still can stand that map.
Good luck!

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Fri Feb 26, 2010 10:52 am

Ur mistake with Jugger was that +1 and +2 doesn´t have 16 and 17 chances but 15 each.

9HP Infs behave like 10HP ones, only they do less dmg to Recon, so less counter is allowed...

From pivot 1,5 5HP Recon can only do more dmg then 9HP Recon if it has it´s full luck and the 7HP zero luck,
5*44 =22% 7*31= 21,7%
So it would say AI attacks too if the damages are equal cause the attack rate was clearly higher then 1 to 250!
so for equal or more dmg the rate is 4 to 250, that sound pretty close to reality, while i think it was a bit more common maybe like 1 to 40 or so
kill rate 10,15 or 20 to 250 depending on 5HP recon partial HP (4,1; 4,2 or 4,3), that makes 1 to 1562/1042/781.
So yeah i hope u find an easier way XP.

I thought u were caring about Spann. It´s ur map. On FI 2,0 do u want to know the number with our new formula? We could also check if there are possible strats with Recon.

fcastro

Re: War Room High Scores & Strategy

Post by fcastro » Sat Feb 27, 2010 2:16 am

Ok. Look also for 3 days Jugger strats requiring less than 1 to 1000.

I math'd the chances for Robots to get Plain Coping Recon attacked by 9 hp Plain Inf, and found 1 to 51 for Flak and 1 to 30 for Jugger... Well it looked to be easier, maybe I made an error, I'll sent u my calculs if u want. So getting Flak looks also very boring with his killing rate of 22%... I checked with 10 hp inf, and found that he would attack 1 to 5 but only be killed 1 to 68 so it's less interesting (but however less annoying, it's a choice to make...).
Btw in same situation against 8 hp inf, Colin with +8 attack would make at min 56% damage and receive at max 560G... Do u think he will be attacked? It is testable in VS thanks Drake's SCOP. If yes Colin might be able to get Ridge 2.0...

So yeah your theory is really excellent, soon VS mode will be useless =O

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Sat Feb 27, 2010 2:54 am

I don´t think there´s a 3 day strat which such a "low" luck.

I can make a test map with Colin against 8 HP Inf if u want, both on Plains i can see from ur calcu. (I first came to 54% min dmg, once again i forgot to factor in lower Terrain def first)

It´s really frustrating that 1 to 1000 became average on this game now =/

Oh and do u remember our final statement over Pay Dirt 1,5 6 day SCOP, were we sure it is impossible?

fcastro

Re: War Room High Scores & Strategy

Post by fcastro » Sat Feb 27, 2010 3:26 am

XD Nah when asking for 1 to 1000 strat I didnt mean as "average" or "low" luck strat but as "allowed", that's all. Imo Ridge 2.0 Flak/Nell strats are already insane, I'd like even to ban 1 to 500 strats.
So u made Colin's test? And u got something new in Pay 1.5?

User avatar
-STL-

Re: War Room High Scores & Strategy

Post by -STL- » Sat Feb 27, 2010 3:34 am

we should only ban FI strats, so strats over 1 to 20k XP
Wasn´t Jugger on Risky Vale 2,0 over 1 to 1000?

I didn´t made Colin test yet, i will do later today, i won´t play the game outside of Emu VS testing though.

On Pay i was just asking abut status, did we srsly check it and agreed that SCOP charging is impossible or did we not srsly check?

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest