PDA

View Full Version : Quick close combat probability



AndrewGPaul
10-09-2009, 15:41
I was bored, and had a copy of Excel ...

Marine vs 'stealer:
Marine wins 225/1296 times (17.4%)
Draw 216/1296 times (16.7%)
Genestealer wins 855/1296 times (66%)

Against the Broodlord, the numbers are:
Marine wins 72/1296 times (5.6%)
Draw 81/1296 times (6.3%)
Genestealer wins 1143/1296 times (88.2%)

That's for a basic Marine, not Claudio, Lorenzo, Gideon or Calistarius.

GrogsnotPowwabomba
10-09-2009, 15:47
I was bored, and had a copy of Excel ...

Marine vs 'stealer:
Marine wins 225/1296 times (17.4%)
Draw 216/1296 times (16.7%)
Genestealer wins 855/1296 times (66%)

Against the Broodlord, the numbers are:
Marine wins 72/1296 times (5.6%)
Draw 81/1296 times (6.3%)
Genestealer wins 1143/1296 times (88.2%)

That's for a basic Marine, not Claudio, Lorenzo, Gideon or Calistarius.

And this is why its better to SHOOT THEM IN THE FACE! :p

Patriarch
10-09-2009, 17:31
I challenge you to do Lorenzo, on Guard, against the Broodlord.

GrogsnotPowwabomba
10-09-2009, 17:52
Yeah, if possible all the "special" marines and their chances against Genestealers would be nice. I'm particularly interested in Gideon and Claudio, since they can't shoot.

AndrewGPaul
10-09-2009, 20:35
I challenge you to do Lorenzo, on Guard, against the Broodlord.

...
No. :p

wilsongrahams
11-09-2009, 09:35
It's easy really. It's just 16.7% to get any required roll. Just double that for 5+ as opposed to a 6, divide the result for each extra dice. Just have to think about whether the odds go up or down each time.

WildAnimal
11-09-2009, 10:41
The Emperor says: SHOOTS THE ALIENS BUGGERS!

wilsongrahams
11-09-2009, 12:53
You just made Brothe Claudio cry...

athamas
11-09-2009, 13:41
It's easy really. It's just 16.7% to get any required roll. Just double that for 5+ as opposed to a 6, divide the result for each extra dice. Just have to think about whether the odds go up or down each time.

the problem is rerolls...
and then factoring in rerolls when you have a +1/2
and then the fact that the marine will only reroll of he is currently loosing...

AndrewGPaul
11-09-2009, 13:58
With re-rolls;
100% chance of getting 1 or more (obviously!)
97% chance of getting 2 or more (1/6 chance of failing; chance of failing twice in a row is therefore 1/6 x 1/6)
89% chance of 3+
75% chance of 4+
56% chance of 5+ and
31% chance of 6.

(yes, they don't add up to 1; those percentages are cumulative)

altwaev
11-09-2009, 14:23
Doing a few quick tests in excel...

WITHOUT guard

Claudio vs Genestealer:
54% Claudio Wins
22% Draw
24% Stealer Wins

Claudio vs Broodlord:
19% Claudio Wins
16% Draw
65% Broodlord Wins

wilsongrahams
11-09-2009, 14:46
54% really? He hasn't survived a combat for me yet! I think I'll get the dice out and do a probability test with 100 combats and see how they end up. Reality is always different ot probability anyway but it would be good to see how close it gets.

altwaev
11-09-2009, 15:07
New Test:

chance with guard on is in ( ). Assumes reroll if at least one die is less than 3, and that he'd otherwise lose.

Tested with 500 simulated combats.

Claudio vs Genestealer
51% (60%) Claudio Wins
23% (26%) Draw

Claudio vs Broodlord
18% (23%) Claudio Wins
13% (16%) Draw

EDIT: The guard statistics are incorrect - should be a little better.

altwaev
11-09-2009, 16:06
This one got a little complicated, so i'm not totally sure if it's correct:

Assumes Parrying before trying Guard (is that correct?) and only using guard if marine rolled a 1 or 2

Chance with guard is in ( )

Lorenzo vs Genestealer
41% (50%) Lorenzo Wins
25% (29%) Draw

Lorenzo vs Broodlord
8% (12%) Lorenzo Wins
11% (13%) Draw

Edit: the guard statistics are incorrect - should be a little better.

athamas
11-09-2009, 17:51
This one got a little complicated, so i'm not totally sure if it's correct:
and only using guard if marine rolled a 1 or 2



this does not work.. :(

the stealer could roll 3 1's so no reroll on a 2... simularily, if the marie loses he will always reroll...so he could reroll on a 5...

AFnord
11-09-2009, 17:51
Guard can be represented by simply rolling one more die. You will always reroll when you don't win the combat, so basically you have a 2 dice chance of a success.

altwaev
11-09-2009, 18:30
this does not work.. :(
the stealer could roll 3 1's so no reroll on a 2...

Sorry i should have wrote that he rerolls if he rolled a 1 or 2 AND he would otherwise lose. IIRC that's the way i set it up.


simularily, if the marie loses he will always reroll...so he could reroll on a 5...
I think i did indeed overlook this. guard statistics should probably be a little better then.

Scriboergosum
11-09-2009, 21:41
Guard can be represented by simply rolling one more die. You will always reroll when you don't win the combat, so basically you have a 2 dice chance of a success.

That's not always going to be true. If the marine player gets a draw on a high roll, say both he and the genestealer rolled a five, then he might not reroll, even though he didn't win. He might be satisfied with the draw, at least he didn't die.
A gambler might reroll in the hopes of getting a six, whereas a more careful player will accept the draw.

If you simply roll the extra dice from the start, you give the marine player the advantage of knowing what the reroll is going to be, and hence whether he wants to use it. The human facter in making this decision effectively makes it impossible to accurately figure out the probability when using guard. There's choice involved, and when a draw comes up that choice is not a no-brainer, as it obviously is when the marine loses.

Malakian
15-09-2009, 11:24
How do you count those probabilities for a normal terminator against a stealer?

Milgram
15-09-2009, 12:38
if you only reroll for guard if you actually lose as a marine, then the probabilities with guard are the following:

W: 27.5%
D: 25.2%
L: 47.3%

the whole thing my differ by 1/216, but thats it.

Znail
15-09-2009, 14:08
That's not always going to be true. If the marine player gets a draw on a high roll, say both he and the genestealer rolled a five, then he might not reroll, even though he didn't win. He might be satisfied with the draw, at least he didn't die.
A gambler might reroll in the hopes of getting a six, whereas a more careful player will accept the draw.

If you simply roll the extra dice from the start, you give the marine player the advantage of knowing what the reroll is going to be, and hence whether he wants to use it. The human facter in making this decision effectively makes it impossible to accurately figure out the probability when using guard. There's choice involved, and when a draw comes up that choice is not a no-brainer, as it obviously is when the marine loses.

For Claudio so is it actualy true as you can always reroll the lowest dice without risk as your highest dice is left as is. So Claudio effectively rolls 3 dice when on guard all the time.

[EDIT] Used this to calculate the actual probabilities for Claudio against a Genestealer.


Doing a few quick tests in excel...

WITHOUT guard

Claudio vs Genestealer:
54% Claudio Wins
22% Draw
24% Stealer Wins

Claudio vs Broodlord:
19% Claudio Wins
16% Draw
65% Broodlord Wins


New Test:

chance with guard on is in ( ). Assumes reroll if at least one die is less than 3, and that he'd otherwise lose.

Tested with 500 simulated combats.

Claudio vs Genestealer
51% (60%) Claudio Wins
23% (26%) Draw

Claudio vs Broodlord
18% (23%) Claudio Wins
13% (16%) Draw

EDIT: The guard statistics are incorrect - should be a little better.
500 simulations are to low to get accurate statistics for something like this.

Using probability instead of statistics gets you:

Claudio vs Genestealer (Guard)
~52.8% (~64.8%) Claudio Wins
~19.2% (~18.5%) Draw
~27.9%(~16.7%) Genestealer wins

Or Odds for Claudio winning contra losing (draws are ignored as they can then attack again):
1.9:1 (3.9:1)

So you can aproximate that Claudio can take on twize as many genestealers before dying if he is using Guard instead of attacking himself.

Now for some extra bonus!
Old rules for dual Lightning Claws (2D6+2) vs Genestealer:
~72.1% LC win
~13.8% Draw
~14.1% Stealer win
Odds 5.1:1 in favor of the old LC.

Old LC was better even compared to the new ones using Guard! I am somewhat underwhelmed by the current LC's and consider them a downgrade compared with the basic Stormbolter equiped Terminators that also got improved overwatch now.

You could also use the same numbers for a Sergeant with LC's in this edition (except it lacks the Guard numbers). Hmm, I guess its interesting to calculate the Th/Sh numbers to see if one Assault kit is better then the other. Should be easy to do, but will do it later.

eriochrome
15-09-2009, 15:31
I have a whole series on Space Hulk Probabilities coming out on my blog:

http://twilight40k.blogspot.com

so far I have done

Shooting a Steal with storm bolter plus how far can a stealer expect to move against overwatch without jam clearing.

Shooting a stealer with assault cannon and how far the stealer can expect to get with explosions possible vs overwatch.

Shooting the broodlord with both guns on overwatch.

CC Powerfist vs Stealer with and without guard and where the cut off should be for rerolling in ties.

Coming up is Lightening Claws with and without guard(always reroll ties), and thunderhammer with and without guard and discussion for cut off for rerolling ties.

Still Working on Librarian to optimize use of psychic points in combat with guard.

Powersword with guard and parry gets trickey so I have not started the analysis of when to take which reroll yet.

Znail
15-09-2009, 15:41
Coming up is Lightening Claws with and without guard(always reroll ties), and thunderhammer with and without guard and discussion for cut off for rerolling ties.

Still Working on Librarian to optimize use of psychic points in combat with guard.

Powersword with guard and parry gets trickey so I have not started the analysis of when to take which reroll yet.

As I mentioned above so can you always reroll with Lightning Claws and Guard as you only reroll one dice, leaving your highest untouched so no risk of getting lower then you already have.

You use psi-points after seeing the results so there is no need to consider anything else then winning.

Easyest thing with such a choise is to asume no rerolls if tied and see what that gives you. The only reason to reroll a tie would be if the odds of winning is greater when doing so compared to letting the stealer attack again (this also asumes that the Stealer has leftover AP for that). So you would always reroll a 1 as that cant get lower and possibly a 2 as well, althou that isnt sure as this loadout may have decently high odds of winning a rematch.

eriochrome
15-09-2009, 16:10
As I mentioned above so can you always reroll with Lightning Claws and Guard as you only reroll one dice, leaving your highest untouched so no risk of getting lower then you already have.

You use psi-points after seeing the results so there is no need to consider anything else then winning.

Easyest thing with such a choise is to asume no rerolls if tied and see what that gives you. The only reason to reroll a tie would be if the odds of winning is greater when doing so compared to letting the stealer attack again (this also asumes that the Stealer has leftover AP for that). So you would always reroll a 1 as that cant get lower and possibly a 2 as well, althou that isnt sure as this loadout may have decently high odds of winning a rematch.

Lib is interesting since you want to maximize use of psy points so if it was significantly cheaper to buy a tie than a win you might just buy the tie to use up the stealer action point and have him as a screen from further attacks. But it is not without guard a tie is 1.08 psi points average while the win is 1.74 so the straight win is better. On guard it gets really interesting since you might not want to reroll a tight loss and instead buy your way to the win similarly you might want to reroll a tie or buy a win. A two dimensional decision array is required based on what you and the stealer are sitting on.

Patriarch
15-09-2009, 18:38
Hmm, I guess its interesting to calculate the Th/Sh numbers to see if one Assault kit is better then the other. Should be easy to do, but will do it later.

On the Monte Carlo principle (i.e. play loads of games) the old style LCs were far better than TH/SS. A TH has a 1/3 chance of rolling a 1 or 2 with every close combat; even with a parry from the SS the Stealer will be fairly unlucky if his highest roll is a 4. LC marines were pretty reliable roadblocks, even more so than bolter marines on OW with corridors.

The only reason the TH/SS combo works in 3rd (at least compared to the LC)is because a sergeant is carrying it, and he can use guard.

eriochrome
15-09-2009, 18:58
Lightening claw on guard is 64.8, 18.5, 16.7 W/T/L
Thunderhammer on Optimal Guard(Never reroll a tie you could lose) is 71.7, 17.3, 11.0.
Thunderhammer is better.

Znail
15-09-2009, 20:33
Lightening claw on guard is 64.8, 18.5, 16.7 W/T/L
Thunderhammer on Optimal Guard(Never reroll a tie you could lose) is 71.7, 17.3, 11.0.
Thunderhammer is better.

Ah, but you are comparing a Sergeant with Th/Sh to a private with dual LC! Its true that the sergeant is better then Claudio, but if we asume that we are free to equip a sergeant or private with either option, then it looks rather bad for Th/Sh! This is even more so when not using Guard as Th/Sh gets realy bad then or LC's realy good depending on your point of view.

LC sergeant on Guard W/T/L: 83.3% 10.3% 6.4% Odds for the sergeant 13:1!
LC sergeant without Guard W/T/L: 72.1% 13.8% 14.1% odds 5.1:1.

I made an error somewhere with the private Th/Sh so need to redo them later to get the exact numbers, but they look bad. We might have to rethink the options for custom forces as LC is notably better then Th/Sh and equiping a private with Th/Sh is rather bad.

eriochrome
16-09-2009, 00:54
If you are housing custom setup, you might consider making LC's reroll their highest on guard if they choose to use the guard reroll. Would make them more likely to kill and die compared to the thunderhammer which might match better with the fluff behind the equipment.

Znail
16-09-2009, 14:55
Hmm, nah, I think its a better idea to buff the Th/Sh setup if any change is to be made to the rules as right now so is it a realy bad idea to use them for privates as that is so much worse then the basic loadout. Even dual LC is somewhat worse as the Stormbolters are quite nice on overwatch now.

But I think its a better idea if using custom setups to simply factor in any quality diffirence and balance the force used or points cost around the actual quality of the models. Althou that means that Th/Sh terminators would have a worth of less then a basic terminator with Stormbolter.