How do you predict the outcome of multiple combats?
Moderators: Forum Moderators, Developers
How do you predict the outcome of multiple combats?
The attack dialog shows the expected hitpoints histogram for the results of one combat, but what do you do to predict the result of two or more combats?
For example, suppose there's an Orcish Warlord and you have three Elvish Marksmen. How likely is it that the three Marksmen will bring the Warlord to 5HP or less in one turn? I know how to compute the answer, but what do you do during play when you need to estimate the probability of success of a particular attack plan?
For example, suppose there's an Orcish Warlord and you have three Elvish Marksmen. How likely is it that the three Marksmen will bring the Warlord to 5HP or less in one turn? I know how to compute the answer, but what do you do during play when you need to estimate the probability of success of a particular attack plan?
Re: How do you simulate multiple combats?
Are you asking about how players work it out in their head? That's what I understand from the final sentence and which forum this is in, but I didn't expect that from the subject line and preceding text. If I wasn't also interested in development topics, I'd have assumed this was a topic to skip.
An Orcish Warlord implies a boss, so would get more thought than the typical enemy units, and with different aims. I'd probably have already killed his recruits, and be trying to give the XP to a specific unit. So first I look at the histogram and CTK with the unit that I want to get the XP, and then the chance that other units have to weaken the boss without killing. It depends also on the chance of my units surviving if the boss doesn't die this turn.
For a strong nonboss enemy:
... when I had similar units to attack (such as three Marksmen), I'd look at the histogram for one attacker, and either just rely on instinct for working out what the histogram for 3 attackers is, or just calculate (total number of strikes) vs (hits needed).
... when I had very different units but none with slows (such as 2 Pikemen and 1 Mage), I'd get a very rough idea of how much damage the other attackers could do, just by looking looking at the amount of damage that the long bars on the histograms have and adding up each unit's likeliest damage.
... when I had very different units including one with slows (such as a Shaman, Pikeman and Mage), the first question would be whether my shaman could survive missing every time (adjusted by how valuable she is, how many XP she has), and a very rough estimate of the other attacks as in the (2 Pikemen, 1 Mage) situation. If the shaman is still part of the attack plan then she always goes first for slowing, I see what the RNG rolled, and then I look with more detail at what damage the other units could do. If the shaman hit then I'd probably continue with the attack even if the CTK is zero this turn.
An Orcish Warlord implies a boss, so would get more thought than the typical enemy units, and with different aims. I'd probably have already killed his recruits, and be trying to give the XP to a specific unit. So first I look at the histogram and CTK with the unit that I want to get the XP, and then the chance that other units have to weaken the boss without killing. It depends also on the chance of my units surviving if the boss doesn't die this turn.
For a strong nonboss enemy:
... when I had similar units to attack (such as three Marksmen), I'd look at the histogram for one attacker, and either just rely on instinct for working out what the histogram for 3 attackers is, or just calculate (total number of strikes) vs (hits needed).
... when I had very different units but none with slows (such as 2 Pikemen and 1 Mage), I'd get a very rough idea of how much damage the other attackers could do, just by looking looking at the amount of damage that the long bars on the histograms have and adding up each unit's likeliest damage.
... when I had very different units including one with slows (such as a Shaman, Pikeman and Mage), the first question would be whether my shaman could survive missing every time (adjusted by how valuable she is, how many XP she has), and a very rough estimate of the other attacks as in the (2 Pikemen, 1 Mage) situation. If the shaman is still part of the attack plan then she always goes first for slowing, I see what the RNG rolled, and then I look with more detail at what damage the other units could do. If the shaman hit then I'd probably continue with the attack even if the CTK is zero this turn.
Re: How do you simulate multiple combats?
Yes, that's what I meant. I was curious how other players deal with having to work out in their heads the likely outcome of each particular attack tactic (especially when there are only so many hex sides open, and there are some weak units that need to be given XP, and so on...) Thanks for explaining your approach
 StandYourGround
 Posts: 256
 Joined: May 13th, 2009, 2:16 am
 Location: On a blue ball spinning through space at incomprehensible speed
Re: How do you predict the outcome of multiple combats?
The most important thing for me is intel. When attacking, I want to know what's going on behind my enemy's front lines. What kind of reinforcements are standing by, and how many of them can get to the front in one turn? What's going on in the other fronts? Can I count enough units to add up the amount of gold they should have? Also, where are his scouts, and how much of my army do they know about? These are all important for deciding whether to strike or reallocate units to firm up for defense. And when it comes to defense, I ask all the same questions about myself. Can I lure my opponent into overextending and overwhelm them with my reinforcements, or should I pull back and wait for a better time, etc.
As for your original question, when it comes to the math, I look at what I have and what I can see my opponent has, and ask if I can succeed if only about 40% of my attacks land. If no, I regroup. If maybe, I prod the weaker fronts with scouts to see if I can draw away some of his reinforcements. If yes, I attack. If I feel confident I can do some damage with 40% hits, I'll probably do better than 40% and really strike a blow. And this may not be even my all out attack, I go through this process when deciding whether to call a bluff and "take the bait." If I can wipe out his front line bait and bring enough force to make him think twice about retaliating, that's a small victory worth taking, and possibly the beginning of a successful incursion.
As for your original question, when it comes to the math, I look at what I have and what I can see my opponent has, and ask if I can succeed if only about 40% of my attacks land. If no, I regroup. If maybe, I prod the weaker fronts with scouts to see if I can draw away some of his reinforcements. If yes, I attack. If I feel confident I can do some damage with 40% hits, I'll probably do better than 40% and really strike a blow. And this may not be even my all out attack, I go through this process when deciding whether to call a bluff and "take the bait." If I can wipe out his front line bait and bring enough force to make him think twice about retaliating, that's a small victory worth taking, and possibly the beginning of a successful incursion.
I will now resume lurking silently.
Re: How do you predict the outcome of multiple combats?
For just the numbers, I use expected value. (then use experience for the quirks of wesnoth that go beyond numbers)
A bowman on a saurian in forest at day does 18 total damage, and hits 40% of the time.
The average damage will be 18 times 0.4, or 7.2 damage. Of course we don't hit fractions of a number in wesnoth, that makes it interesting.
You can do this for multiple units, which takes you beyond what the GUI can do.
I never really find the probability, I just compare what I need with what is expected and eyeball it.
A bowman on a saurian in forest at day does 18 total damage, and hits 40% of the time.
The average damage will be 18 times 0.4, or 7.2 damage. Of course we don't hit fractions of a number in wesnoth, that makes it interesting.
You can do this for multiple units, which takes you beyond what the GUI can do.
I never really find the probability, I just compare what I need with what is expected and eyeball it.

 Posts: 2
 Joined: May 4th, 2019, 3:28 pm
Re: How do you predict the outcome of multiple combats?
just do the mental math and account for as many results as you can
it's not that hard
it's not that hard