Jump to content

Mathemagics


Myyrä

Recommended Posts

I have been doing some Malifaux computer simulations again after a while and thought you guys might be interested in the results as well. I did a quick study about how different attack stats and damage tracks affect the model's ability to kill stuff. The results can be seen in the table below.

attacks.png?dl=1

The numbers in the table tell how many AP a model with attack value found on the left and damage track found on the top will need against a model with 6 wounds, Df 5 and no other significant abilities. To actually compare different models you need to divide the number of attacks the models can make per turn with the reference number on the table.

If the image is too difficult to read or you want the numbers for some other use, the same data can be found as an excel sheet or csv.

What do you guys think, is the ability to kill Witchling Stalkers a good proxy for model's offensive capability or should the target be something different? Should I use an average of multiple different targets or something?

At this point calculating new targets takes less than a minute, so I can do other targets as well if someone is interested. What targets would you like to see?

Some notes: The commas are decimal commas, because I use Finnish language OS. The models do not cheat fate. The numbers found in the table are each an average number AP spent on 10 000 kills in computer simulated malifaux combat.

  • Like 5
  • Thanks 1
Link to comment
Share on other sites

*reads down the columns, as a power looping player ought to*

...huh. Usually we don’t really bother looping in the sort of low damage tracks which get the best results from an increase in attack value. But then, those high damage track models are often being looped for defense (and sometimes positive flips) instead.

Speaking of which, is there an attack value 3 in Malifaux at all? I think there’s...one totem? Or is this chart doing double duty for some Other Side mechanics I don’t know?

5ss models aren’t the baseline defense in any army I meet, but my meta skews elite. Two- or three-beatstick armies are common. Or wait, the Ten Thunders does bring a bevy in the 5-6ss range.

Edited by Gnomezilla
I forgot a word
Link to comment
Share on other sites

8 minutes ago, Gnomezilla said:

*reads down the columns, as a power looping player ought to*

...huh. Usually we don’t really bother looping in the sort of low damage tracks which get the best results from an increase in attack value. But then, those high damage track models are often being looped for defense (and sometimes positive flips) instead.

Speaking of which, is there an attack value 3 in Malifaux at all? I think there’s...one totem? Or is this chart doing double duty for some Other Side mechanics I don’t know?

5ss models aren’t the baseline defense in any army I meet, but my meta skews elite. Two- or three-beatstick armies are common. Or wait, the Ten Thunders does bring a bevy in the 5-6ss range.

You can use attack value 3 as a proxy for model with attack value 5 attacking something with Df 7, because all that matters really is the difference between the stats.

Also, the relative increase from power looping doesn't really depend on the damage track. You get about 12% reduction in AP spent when increasing AV from 5 to 6 regardless of your damage track. Thus you should usually loop something with a better damage track, because the absolute increase in damage done is larger. Damage done is basically the inverse of the AP spent.

  • Like 1
Link to comment
Share on other sites

nothing too surprising in those results, though it does show that sometimes a good damage track is better than high minimum damage.

I think a 7 wound model would be interesting because then the min 3 can’t get there in 2 without a moderate.

also the effect of hard to kill. 

Link to comment
Share on other sites

5 hours ago, JoeCL said:

nothing too surprising in those results, though it does show that sometimes a good damage track is better than high minimum damage.

I think a 7 wound model would be interesting because then the min 3 can’t get there in 2 without a moderate.

also the effect of hard to kill. 

Here's some 7 wound models for you with varying defensive abilities.

7wounds.png?dl=1

Also excel sheet if someone wants it.

  • Like 4
Link to comment
Share on other sites

3 hours ago, Clement said:

So based on this, we could compare Df4 HtW vs. plain Df5 by just comparing the values in the first and 3rd charts, so long as you compare the av3 row off the "plain Df 5" to av4 of the htw table, right?

Yeah.

3 hours ago, Clement said:

Informative, I didn't think HtW did that little.

It's the worst defensive ability.

  • Like 1
Link to comment
Share on other sites

15 hours ago, Clement said:

So based on this, we could compare Df4 HtW vs. plain Df5 by just comparing the values in the first and 3rd charts, so long as you compare the av3 row off the "plain Df 5" to av4 of the htw table, right?

Informative, I didn't think HtW did that little.

I find that most of the time I only expect to do Weak damage because I'm not on a cheatable flip anyway. Hard to wound just makes that more likely. Where hard to wound shines is in situations where your opponent would want to cheat damage, and this chart doesn't allow for cheating. So its always going to be at its worst in this sort of situation. 

  • Like 1
Link to comment
Share on other sites

Hard to wound is mostly protection against someone focusing and oneshotting you in my book. It does decrease the likelyhood of getting a fluke moderate or severe on normal hits which go from :-fate to :-fate:-fate but I think the focus protection is a bigger deal, especially on models with 4-5 wounds who live in dangerous oneshot territory usually.

Link to comment
Share on other sites

3 hours ago, Ludvig said:

Hard to wound is mostly protection against someone focusing and oneshotting you in my book. It does decrease the likelyhood of getting a fluke moderate or severe on normal hits which go from :-fate to :-fate:-fate but I think the focus protection is a bigger deal, especially on models with 4-5 wounds who live in dangerous oneshot territory usually.

Models with 4-5 wounds tend to also live in the dangerous 2-shot territory from two weak hits.

Hard to Wound offers the greatest benefit when it's coupled with armor, because moderate and severe hits contribute more to the expected damage.

  • Like 1
Link to comment
Share on other sites

For me, Hard to Wound is a protection against Blasts. As you've pointed out, it gets better when coupled with damage reduction.

I merely glanced at the data, but I noticed some funny results here and there: i.e. 6wds 5Df chart, av7: compare 4/5/6 vs. 4/5/7. Could you please add standard deviation value? :)

Link to comment
Share on other sites

2 hours ago, Pikciwok said:

For me, Hard to Wound is a protection against Blasts. As you've pointed out, it gets better when coupled with damage reduction.

I merely glanced at the data, but I noticed some funny results here and there: i.e. 6wds 5Df chart, av7: compare 4/5/6 vs. 4/5/7. Could you please add standard deviation value? :)

deviations.png?dl=1

Here's standard deviations from a single run. It didn't use the same random seed as the original, but it should give you an idea of the scale.

Also, for those who don't know it, these numbers don't really tell anything about the accuracy of the simulation method. That's a different thing.

  • Like 2
Link to comment
Share on other sites

My mind goes crazy with this stuff and I wish I had the time to pick up what you did and run with it.

Stuff I would want to do but doubt anyone has the time to do:

programming a 6 card hand in for either player and assigning success priorities to attacker/defender

Where the priorities are high:

Game is on the line, will cheat highest card as long as is logical to do so

Medium:

Attacker will cheat moderate card if losing, defender will cheat moderate card if losing or to put attacker on neg flips

Low: 

No cheating 

Allow the user to mix and match priorities to see success rate.

Of course, it would have to be programmed so that cards in hand don't appear in the respective deck.

It would also need to be programmed that the player would always cheat the lowest card that gets the best result.

Then make the hand size variable

Also making it variable when the deck shuffles an when the players get new hands . If they had bad cards in hand they don't get shuffled in.

Then I want to run simulations with more than two models and use the same hand to govern success. Like 2 on 1, or two separate one on one battles

:)

Link to comment
Share on other sites

I have been considering adding cheating using some simple decision rules like:

Cheat in a severe if you can kill a model, deal at least expected 4 damage more or take expected 4 damage less or you can save your model from death.

Cheat in a moderate if you can deal or prevent 2 damage.

Cheat in a weak if you can get anything useful out from it.

The reason I haven't already (besides laziness) is that a system like that would have rather limited usefulness. It would be somewhat difficult to account for how much of the utility came from cheating and how much came from the model's own capabilities. Stronger models would appear much stronger, when the system would cheat with them more often, but you might still be better off hiring two 5ss models instead of that 10ss model that has better attacks. I guess it would be most useful in comparing different defensive capabilities.

  • Like 3
Link to comment
Share on other sites

There would be tremendous usefulness but it could only come about with a more complex simulation. I agree investing all resources into 1 model doesn't simulate real games.

For example if you could simulate 3+ different interactions all demanding resources from the same hand and run that simulation 1000 + times, you can start to give some real insight into which crews are better, what are the best combos, etc. Who is worth investing resources into, which models are compatible , what are the most efficient combos.

But - again - the time a person would need to program that could be extensive. 

Link to comment
Share on other sites

11 hours ago, JoeCL said:

There would be tremendous usefulness but it could only come about with a more complex simulation. I agree investing all resources into 1 model doesn't simulate real games.

For example if you could simulate 3+ different interactions all demanding resources from the same hand and run that simulation 1000 + times, you can start to give some real insight into which crews are better, what are the best combos, etc. Who is worth investing resources into, which models are compatible , what are the most efficient combos.

But - again - the time a person would need to program that could be extensive. 

The interactions between 2 models are pretty complex. just deciding if you want to walk away and drop a scheme marker, or try and kill and drop a marker, or just try and kill all depending on going first and second is already pretty complex (Going first and given the choice of Focus or Defensive stance or a second attack where 2 weaks doesn't kill but weak and moderate does, which is best for keeping you alive if severe is enough to kill the enemy but you only have 1 severe card in hand  ) . You would be testing your rules as much if not more than you are testing the models abilities. 

Link to comment
Share on other sites

I did some really basic stuff this time, because I think most people might not be aware of this and have not seen the previous thread where I posted more or less the same info (which is no wonder considering even I don't remember when that was).

So, here you have the hit probabilities as a function of the difference between attack and defense values (atk-def):

hitprob.png?dl=1

and also as a figure, because why not

hitprob_fig.png?dl=1

As you can see, having a :+fate is more or less equal to having a 2 points higher attacks stat and :-fate equal to having a 2 points lower attack stat.

One would easily think that hitting something with Df 9 would be impossible when you have Ml 5 and :-fate but it's only about equal to having to throw a 6 on a normal 6-sided die.

 

  • Like 2
  • Thanks 2
Link to comment
Share on other sites

4 minutes ago, Myyrä said:

I did some really basic stuff this time, because I think most people might not be aware of this and have not seen the previous thread where I posted more or less the same info (which is no wonder considering even I don't remember when that was).

So, here you have the hit probabilities as a function of the difference between attack and defense values (atk-def):

As you can see, having a :+fate is more or less equal to having a 2 points higher attacks stat and :-fate equal to having a 2 points lower attack stat.

One would easily think that hitting something with Df 9 would be impossible when you have Ml 5 and :-fate but it's only about equal to having to throw a 6 on a normal 6-sided die.

If you've seen my dice you'd know that that was impossible...

Thats a really useful table because people do seem hung up on the stats are the key, and things with a low attack stat are worthless. 

  • Like 2
Link to comment
Share on other sites

1 minute ago, Adran said:

If you've seen my dice you'd know that that was impossible...

Yeah, I try to never get into a situation where I have to roll a 6. But I guess we all know at least one guy who always seems to succeed at those.

1 minute ago, Adran said:

Thats a really useful table because people do seem hung up on the stats are the key, and things with a low attack stat are worthless. 

High stats are a lot more valuable when cheating, though. That makes me avoid costly models with low stats, but I don't mind having low stats on my cheaper models, who don't get cards from my hand anyways.

  • Like 2
Link to comment
Share on other sites

The last analysis was in early 2015: the model pool has changed a lot since then. Who could have envisioned a wokou raider which is impossible to cheat against when attacking its Df? Also the forum change in temperament, in software breaking old links, in the vigorous suppression of ‘thread necromancy’ so old ideas die...

13 minutes ago, Adran said:

If you've seen my dice you'd know that that was impossible...

Thats a really useful table because people do seem hung up on the stats are the key, and things with a low attack stat are worthless. 

I know the stats superiority thing is a fib I tell myself to make myself feel better about power loop. -_-

Link to comment
Share on other sites

1 minute ago, Gnomezilla said:

I know the stats superiority thing is a fib I tell myself to make myself feel better about power loop. -_-

You could probably improve your power loop game by not focusing too much on getting the best stats possible. Getting most stats to 6 by looping emissary is quite enough.

  • Like 1
Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...

Important Information