Basic ship building tips
Ship building help
Short overview:
(credits to dudiobugtron for making this)
More detailed explanations:
Paths:
When you make a path you want it to be as long as possible and only split at the end, by that you can put many items in one row and they can all interact with each other.
The same goes with custom chunks, you can try to move the chunks and then calculate which layout you came up with is the best by counting the items spots in the ship and the crosses you use (read multipliers and loop for the numbers)
There is also a list of the Meta Paths for classic ships.
and for custom ships, you can always ask in the discord server.
Items:
Priority:
When you want to fill your path with items you don't prioritize direct dmg first. You want to first put additive dmg, extra projectiles, piercing and aoe etc. But keep in mind there that you want to fill cross loops with them so put the there. Then if you have free space you can fill it with Damage Crits and maybe other useful items depending on the situation.
Add Damage:
Before you multiply your dmg you want to first have additive dmg like the +1 items or the legendary +100.
+1 damage item will be only good if you have less than 3 "total damage" which you can see in the stats menu. Because each new +1 will give deminishing effects.
Skills of course change that. With the first skill each +1 dmg item will now to +6.
The 2nd skill however. Different from what you might think will NOT make it a +26.
What it does:
When you have the 2nd skill for the +1 damage, you want to fill all the empty spots (in other words, spots outside the path of projectiles) on the ship with it. This is because when skilled, it adds +20 damage per +1 damage item on the ship, not just on the path. And use 2-4 of them at the start of your build depending on the build and base damage.
What it does:
When you have the 2nd skill for the +1 damage, you want to fill all the empty spots (in other words, spots outside the path of projectiles) on the ship with it. This is because when skilled, it adds +20 damage per +1 damage item on the ship, not just on the path. And use 2-4 of them at the start of your build depending on the build and base damage.
Damage crits:
Multipliers are very important, they are the items you stack the most in your build.
Which multipliers to use and when:
Early game:
Since you don't use as many items in a build then, it is ok to use the arithmetic mean(what most would call the normal average) as the factor to compare the average dmg increase of a crit item.
Item tier-list:
1. The 0111111111234 crit, on average is a x1.38 increase per item. Yes even with the chance to delete a projectile this is good.
2. The x2 crit, on average is a x1.33 increase.
3. The left and right curves (with skills) are a x1.3 increase. Their very low price makes them a good replacement for better items if they are too expensive.
4. The x10 crit is overall not really good, even though it has a theoretic average of x1.36. The chance is just too low. Don't use this
4. The x10 crit is overall not really good, even though it has a theoretic average of x1.36. The chance is just too low. Don't use this
After tier 650 when you can get both skills for the x2 crit it becomes the new best crit item with an average of x1.44 (geometric mean).
Later game:
Since you now stack more items in one ship (mostly custom ship) you want to use the geometric mean
Item tier-list:
1. Thermite, Charged and Plutonium Round are the best with an average of x1.54 (very good since they don't need skills, but very expensive)
2. The x2 crit, on average is a x1.44 increase. (skilled)
These two are your priority
Combinations of the skilled x10 and skilled charger can work well too
The math:
2.6^0.45=~1.537
2.5^0.4=~1.443
1.3*100%=1.3
18/13=~1.385
2*0.33+1*0.67=1.33
10*0.04+1*0.96=1.36
Pierce
Pierce is a very good addition to your build. And also interacts very well with aoe.
When you pierce a block you also hit the block after it.
I recommend using 2 of the regular pierce except for waves, why 2? Considering you already use aoe: the first one will give 2.5x and the second one will give around 1.6x which is better than any Damage Crit.
The remaining damage of hp pierce depends on the damage, I recommend only using one of it, this is only good for waves and low tiers. It will pierce through as long as its damage will run out, the damage formula is [Damage / (Items×(Items+1))].
AoE
The more aoe you use the bigger the area of effect gets. But it does not increase enough to make the use of many AoE worth it.
Use one circle aoe for missions and early tiers. Use 2 circle 1 rectangle if for a bigger aoe e.g. if you want to clear
AoE damage reduction is simplified so instead of it being 5 6% damage reduction ([94/100]^5) it would be simplified to 6%×5 so 30%, however, aoe is just too good it is very worth it however don't use too many. AoE damage reductions apply at the ejector so it doesn't matter where you put it
AoE damage reduction is simplified so instead of it being 5 6% damage reduction ([94/100]^5) it would be simplified to 6%×5 so 30%, however, aoe is just too good it is very worth it however don't use too many. AoE damage reductions apply at the ejector so it doesn't matter where you put it
Note: the formula for aoe reduction is separate for each AoE type, so consider using different types if you get the same area
(simplified)
Charger
The charger should be used at the end of the path and with a legendary, however only use the charger if there are at least 10 damage multipliers. This item will help with consistency. (Note: there are other ways to use a charger)
For how this item really works, it will copy 20% of the projectiles damage (does not reduce) and it will add it to its "damage pool" and for passing projectile it will add 10% of the damage pool to the following projectiles basically smoothing the damage spike out
Add Projectile
This will add 1 projectile. In tiers or waves use 2 or 3 of them and add them at the start. Why at the start? to raise consistency. Why 2/3 or not more or less?
2nd add projectile will give a buff of 1.5x which is good
3rd add projectile will give a buff of 1.33x which isn't very good mathematically it's still ok as consistency but I don't really recommend using this.
Since the Combiner legendary will always put out a projectile with the ORIGINAL tag (read what that means down below) it is good to also put +1 projectile items after the combiner.
And the legendary cloner will basically double the rate of fire. Cloner and +1 projectiles order does not matter in the ship if there are no other parts between them.
The ejector creates a projectile with the tag ORIGINAL
A cloner creates a COPY for each COPY and an ORIGINAL for every ORIGINAL. With this, it doubles every projectile added with +1 projectile before cloner and after it
+1 projectile adds a COPY for each ORIGINAL
Copied projectile will COPY every ATTRIBUTES but it CANT become an ORIGINAL projectile itself. This means all damage which is the current ATTRIBUTES will get COPIED.
Crosses:
Crosses are very unique items in this game, and complicated to use.
You can place Crosses in the sections where a path crosses itself, though that only helps if it is the same path crossing there twice and also going into an output so you need to have a kind of a loop. See the example picture:
.png "Example for Damage Cross placement")
.png "Example for Damage Cross placement")
In such places, it can be very helpful, but you need to be careful so that you don't lose too many item spots for a cross since it is at max x2 (or x3 if skilled) damage. Additionally, the cross needs to be rotated correctly, as seen in the picture, the path needs to cross through the side first.
How the cross then works is it will save the damage the projectile had when it passed the cross the first time. And adds it onto the projectile again when passed the 2nd time. This will result in x2(x3) damage in the best case.
How the cross then works is it will save the damage the projectile had when it passed the cross the first time. And adds it onto the projectile again when passed the 2nd time. This will result in x2(x3) damage in the best case.
It is important that you don't increase the damage of the projectile in the "loop". For any damage-increasing items inside the loop, while they will indeed add damage to the projectile, that damage won't get doubled, because the damage-to-be-doubled is "remembered" when the projectile enters the loop (passes the Cross) and is applied to the projectile once it exits the loop (passes the Cross again). It is better to put such damage-increasing items before or after the Cross' loop.
So fill the loop/cross with: aoe, pierce, add projectile, trident (if needed), bounce, etc. that don't directly increase the damage
Two examples:
Optimal:
no dmg increasing items in the "loop"
Damage is 10, enters cross, 10 gets saved, exits cross, adds 10. Dmg is now 20 so it is doubled by the cross. Then you can put a x2 crit to make it 40.
Suboptimal:
you have a x2 crit in the "loop"
Suboptimal:
you have a x2 crit in the "loop"
Damage is 10, enters cross, 10 gets saved, x2 crit makes it 20, exits cross, adds 10. Dmg is now 30 and lower than 40.
(Made by Skill_issue and MasterNO)
Your math seems wrong on multipliers, since you mix multiply and power operations.
ReplyDeleteThe best unskilled is 18/13 = 1.38, with the highest expected value.
But (0.96+10*0.04) = 1.36
And (2*1/3+1*2/3) = 1.33
Are close behind it in expected value, not sure what you were calculating but 10x is 2nd best.
Nope math is not wrong.
DeleteYours isn't too, but the approach is wrong, you calculate the arithmetic mean, and we use the geometric mean. The geometric means is better in this case (except at the random crit item where we don't use it because it is not possible to calculate, or atleast not easily) because the arithmetic mean is strongly influenced by extreme results that's also why the x10 seems so good. But it isn't ;)
You have to correctly calculate the basic probability calculation before you use "geometric mean" for multiple item calculations. I guess you better check with you maths teacher
DeleteAn 'average' is a way of talking about what things are like 'in general'. There are lots of different averages or 'measures of central tendency' you can use, and the best one to use depends on the situation. For example, if you are working out the average number of arms that people have, the mode is the average that most people would use (although the median would also get the same result). If you are looking at average house prices, or incomes, or something like that, then it makes sense to use the median, as it is less influenced by extreme or weird results than other averages can be.
DeleteIt's up to you to choose a good average for the situation. In a lottery with 3 equally likely outcomes, the median and mode are both bad to use as they will not give you a good indication of the 'expected' result.
It can be confusing for a lot of school students, as a number of people use the word average refer to a specific average - the (arithmetic) mean. 'Means' tend to be ways of working out the average by combining all the results in some way. The most common one, the artihmetic mean, is where you add up all the numbers, and divide the result by how many numbers there are. Because many people learn that this is 'the average', it can be hard for them to see that it is not always the best average or 'measure of central tendency' to use. The reason it is bad to use when combining lots of RNG multipliers like in MGB, is because it is highly influenced by extreme results; on almost all runs with a long chain of multipliers, you will get 'below average' damage if you use the arithmetic mean to work out the average.
Yes this is a copy paste, I don't even know how matching it is here. But long text = I am right. If you actually want to discuss about this topic go the discord server and let the people explain.
DeleteThe first mistake of you is that, long text don't have anything to do with your correctness. The second mistake is, this is not statistic, not a problem of central tendency. This is probability, as you said "expected" result.
DeleteMake it simple, for a case, 50% chance to have double damage, then we can expect one item will give us average of 150% damage, two items will give us 225% damage.
Will you get this results with your what so called "geometric mean"? I'm afraid your geometric mean is different from the mathematical "geometric mean", which indicates the central tendency of a set of numbers by using the product of their values, opposed to the sum of their values in arithmetic mean.
When you look at that 50% for x2. Averagely you could say that is x1.5, but this is taking the Arithmetic mean. When you take the geometric mean it is x1.41. The Arithmetic mean is only good to use to see what an infinite amount of projectiles through one item would be on average. When you have multiple items chained though, and also not infinite projectiles the geometric mean comes closer to the expected value of each item.
DeleteI made a calculation sheet to compare x2 multiplier vs x10 multiplier.
DeleteLike anon, I used to have the wrong perception as well before creating the sheet, thinking x10 is better.
Hopefully these numbers will help illustrate the case to others.
https://docs.google.com/spreadsheets/d/1zTW4GD7wcJzcCQg0WA2gjYn72sfdrCGWJSszAik1IqI/edit?usp=sharing
tl;dr
In fewer chain / early game (<10 slot), x10 might be better,
but in long chain / late game (>20 slot), x2 is the clear winner.
Op, you should really rethink your math. Using a geometric mean here is completely wrong. A geometric mean is a tool to understand a set of data, here we are trying to calculate something that's very specific: the expected damage over a set of bullets. There is only one way to calculate it, and it involves the arithmetic mean.
DeleteThink of it this way: over a large enough amount of bullets, the rng will have output all the possible numbers for activation in an equal distribution.
Suppose the rng can only output numbers from 1 to 100. Since over a large amount of bullets the rng will have output those in the same amount, it's statistically the same as sayin that each number gets chosen once.
So, instead of an rng you have a counter that goes from 1 to 100.
Over the first 4 of those the 10x box gets activated. Over the first 40 the 2x box gets activated.
So the expected damage for the 10x box is four activation for a total of 40 damage, plus 96 non activation for 96 additional damage. in 100 shots it would output 136 damage.
for the 2x box, you have 33 activations for 66 damage plus 60 for the non activations. For a total of 126 damage every 100 bullets.
---
To the anon before me: The 10x box is so much better that you spreadsheet was able to calculate 40 rows of possible activations of the 2x box, while going in overflow at the 10th row of the 10x box. And you didn't even notice. The central column, the one you use to "prove" that the 2x is better, suddenly goes from extremely in favour ov the 10x to extremely in favor of the 2x, and you didn't ask youself why there is such a sharp discontinuity? You should have expected a continuous function, if there's a discontinuity something is wrong.
-If you want to talk about using geometric mean or not when calculating the multiplier of a crit item, go to the discord server. There are a few people who can explain it
Delete-I am not reading what you write because I know it is probably not all wrong but also know that using the geometric means is neither
-if you want to see how bad the x10 is go in game and test it
-oh also on your calc with the x2 you are missing 7 damage (which doesn't change anything since you used the arithmetic mean anyway)
To the anon who think there's something wrong with the sheet calculation that there's a discontinuity when the 10x supposed to be better.
DeleteYou can try adding more bullet to see that happen (top left cell I think)
But ask yourself this :
Can you really shot that many bullet ?
Fo real, I think most of the levels, only a few thousands is spent..
Check yourself how many expected bullet can activate all 40 box of 10x..
Even if you shot as many bullet as there are stars in the observable unicerse, not even 1 bullet can do that.
In low range bullet number, 2x wins on average, by far.
Welp. didn't believe above anon, but checked with 10^24 as number of bullet (est. Number of stars in universe) not even 1 bullet expected to get buffed beyond 25 and above items in x10 box 😶
DeleteMuch less all 40, wonder how many bullet is needed to see that happen
Wouldn't it all depend on using the "projectile stays alive until all damaged used" module. You can only really take advantage of the massive spikes of damage if you use all of it. With this the mathematical mean should be valid.
DeleteThat is a good idea, but it's not the full point that dmg spikes can not be used. they don't happen. if you want to use that strat with the hp-pierce, it is still more likely to get the dmg spikes with the x2 than the x10
DeleteFor conversation purposes, you can call me anon-p.
DeleteBefore reading this I was a 30x guy because my napkin-math, intuition, and experiments all said it was better especially with chargers. Now I have reevaluated my position and I agree that the preference should change in different multiplier item amounts. BUT...
My transmitter has like 9 multipliers (I am using the pathing from this blog) and 5 chargers. I have approximately 10 firerate, 6 total projectiles, app. 30 seconds of effective shooting per world, 12 projectile exits. That makes 7200 projectiles ejected per run.
The spreadsheed had an overflow mistake so I wanted to run a simulation in Python to set the record straight.
I tried different multiplier amounts to see where the tide will turn. It seems that, somewhere around 60-70's the balance shifts and 2.5x becomes better w/ 7200 projectiles. My code is below so you can replicate & tell me if there is anything wrong with it if you happen to know python. I usually don't comment my code so I welcome feedback on that as well.
It basically simulates results for 41 different multiplier counts, each having a 1000 runs, each run consisting of 7200 projectiles, each projectile going through 60 multipliers. Coin is flipped to see if a multiplier is activated, damage is multiplied or not multiplied, then the same for the next multiplier and so on. Before each new projectile, the damage is reset to 1, which is the initial damage for all projectiles.
x = 1000 #MC count. You can play with this all you want. Lower => Less reliable results. Higher => More compute time required. You could probably get away with 30 tho
p = 7200 #projectile per run
b = 40 #initial item count
import random as r
#function for the 30x multiplier
def thirty(d):
c = r.randint (1,100)
if c <=4:
d *= 30
return d
#d for damage
#c for chance
#function for the 2.5x multiplier
def fivehalf(d): #5 * 0.5 = 2.5
c = r.randint (1,100)
if c <= 40:
d *= 2.5
return d
d30=1
#damage of the projectile that travels through 30x multipliers
d30_total = 0
#total damage of the projectiles that travel through 30x multipliers
d25=1
#damage of the projectile that travels through 2.5x multipliers
d25_total = 0
#total damage of the projectiles that travel through 30x multipliers
for b in range (b,b+40+1): #range of multiplier amounts we want to test
d25_total = 0
d30_total = 0
for i in range (1,x+1): # we are running the simulation 1000 times for each multiplier amount
for i in range (1,p+1): #we are firing 7200 projectiles per each simulation
d30=1
d25=1
for i in range (1,b+1): #finally, we are running the projectiles through the current amount of multipliers
d30 = thirty(d30) #d30 multiplier is working
d25 = fivehalf(d25) #d25 multiplier is working
d30_total +=d30 #damages are summed
d25_total +=d25 #damages are summed
print ("Run: ",b-39) #Says which run it is reporting now
print ("30x's damage = ", d30_total)
print ("2.5x's damage = ",d25_total)
if d30_total > d25_total:
print ("30x is better")
elif d30_total == d25_total:
print ("They are equal")
else:
print ("2.5x is better")
Limitation: This code doesn't look into in how many of the runs do 2.5x fares better, only to the total picture. For that I wrote an uglier piece of code and it seems the tipping point is still around 60-70 with a very sharp turn.
Hello anon-p
Deletethanks for your comment, it's always nice to see when people test things so detailed for themselves
I didn't fully read the code cause I don't know much about python but I think I got the main things. seeing that the x2 becomes better with more items overall is already a good result. But the number is a bit high. I think that is because you tested with so many projectiles and then only took the sum of the dmg. that will result in also very unlikely results to happen more often in one simulation and then these extreme results especially from the x10 effect the sum of the dmg immensely. better use a bit less projectiles or take the geometric mean of the results
Do you know what is nicer? Seeing this blog so thanks a lot to you and skill_issue.
DeleteI was aware of that limitation so I wrote another code for that but didn't want to comment it so didn't post it, sadly that is lost now because I have closed the online interpreter tab for that afterwards.
I briefly said the result of that in the end of my post, tipping point was still in the 60-70 range. (64 I think but I am not sure.) I will write it again, do some commenting, and post back. (I think the spreadsheed's mistake may be that it runs afoul of Berkson's Paradox, that is even though high-numbers-of-items-activated ejection scenarios are near-impossible when singled out they are probably not that impossible in combination. I don't know, I didn't exactly check for that.
I am also planning to write something that will simulate the average, max, min damage of any given single-ejection-and-multipliers-only-path (multi-ejection is harder) but that will take longer probably.
you can write me on discord, I think that is easier to chat here. or join the miner gun builder discord server and write about it there if you want others to see it too.
DeleteMasterNO#6140
https://discord.gg/FzgSsGBA
Doesn't the use of a charger eliminate the need to use the geometric mean? The geometric mean would only be relevant for the very beginning of a level before you get a crit. After each tile has gotten its first crit, wouldn't the charger bring the expected value much closer to the arithmetic mean?
ReplyDeleteAlso, why did you use an arithmetic mean for the 18/13 but not for the others? The 18/13 would have a geometric mean of 1.323, so it wouldn't change your ranking.
Charger math is very hard, and tbh i don't know how it effects it here. but i don't think you as a single person changes what many people thought about already
DeleteThe reason to use the arithmetic mean on the risky crit is 1.that calculating the arithmetic mean is very hard 2. as you said it doesn't effect the ranking 3. you only use it in early game with maximum 4 of this item since it is very expensive and the the arithmetic mean is not the worst in this case.
Charger math doesn't change anything here; the highest, rarest damage values from a multiplier chain still aren't going to come up in practice, so excluding them from the average damage (as geometric mean kinda does) is still valid. Chargers just lift your lowest-damage bullets (most of your bullets are very low-damage) up to 10% of your average shot damage, which in late-game cases can easily be a 100x multiplier or higher. It's very valuable for that, but regardless the geometric mean is still best to use when chaining a dozen or more multipliers, as all late-game ships do.
DeleteYeah good point. Though one example against this would be of you spam charger and the ×10 it works pretty good. And can be even better than the skilled ×2
DeleteHow to use faction items efficiently?
ReplyDeleteuse the 3 rounds instead of any other crit item. and if you see in later tiers use 2 double pierce items at the end of the path. the other items are not good
DeleteIn later game is it ok to use the 10x crit instead of the 2x one since skilled 10x has an average of a 2.16 multiplier? or am I doing my math wrong?
ReplyDeleterather not. 2.16 is very optimistic. as I wrote in the guide using the arithmetic mean is better when having chain's of crit items. so especially in later game the x10 is not so good. cause it is only x1.15
DeleteI use 100 damage skilled, it does 1k base dmg and necessary stuff are in path before I place it, is that okay?
ReplyDelete