How does db and pb work in a melee fight?

[Alterith]

How does pb and db work in a straight up melee fight without bashing involved? Some people say that the db will roll against the ob of the opponent first and if it fails, then the pb will independently and separately roll against the on of the opponent (without the db stacking).

Is this true? Or does db contribute to some stacking effect in deflecting or parrying a hit?

[Mhaliah]

This is a hard one to test. The multiple variable ob(attacker), dodge(defender), and parry(defender) would need a very rigorous testing system to isolate and control individual variables. And without knowing the individual roll values for the three independent variables it would need a large data set to look for statistically significant difference between expected values for "ob v dodge, if fail v parry function" compaired to "ob v SUM(dodge + parry)". If someone has the exact way the coding specifies these rolls based on the displayed value I could calculate the expected values for these systems and if the difference is large it would need relatively less data to see which, if it is small it would take larger data sets. If the stacking is not a straight sum it become much much harder to determine.


"the exact way the coding specifies these rolls" means; is this a 1d(displayed value) roll or ydx where the max roll (or some other statistical value) is the displayed value?

[Raeza]

I've noticed that sometimes it says you "dodge" the attack, and other times it says that you "parry" or "deflect" the attack. If these correspond to a failed roll against db and a failed roll against pb, respectively, then I would think you might be able to exploit these messages as a way to test this?

It seems like you could first verify whether or not these messages do indeed correspond to which defense stat caused the attack to fail (e.g. by comparing a defender with high db and low pb, and then vice versa).

Then, if the messages do seem to indicate this, you could use that to test whether the two stats seems to be independent (one roll, followed by the other), or interacting in some way.

Just a thought...

[Draz]

Yes you can test this, also with sitting defence etc.

Easiest way that comes to mind is to vary an extremely high dodge setup, with a very high pb setup, with combinations of both

What I found anecdotally is that either on its own is not fantastic, but dodge above certain threshholds seems to really help pb

[Mhaliah]

The thing brought up by Draz and Raeza are interesting, but not enough for me to make any mathematical conclusion.

I've noticed that sometimes it says you "dodge" the attack, and other times it says that you "parry" or "deflect" the attack. If these correspond to a failed roll against db and a failed roll against pb, respectively, then I would think you might be able to exploit these messages as a way to test this?

Yes, it would rely on the statistical analysis of the ob v (defense). When I say "(defense)" I am referring to all the possible ways that db and pb can be combined into total defense. To do this we would have to make assumptions about what the numerical value of the displayed "bonus" mean (i.e. it is a 1d(ob) with displayed ob being the max roll or is it 6d(ob/6) with displayed ob being the max). Based on the magnitude of the displayed values I tend to think assuming these values are the "max" value is a reasonable assumption.

It seems like you could first verify whether or not these messages do indeed correspond to which defense stat caused the attack to fail (e.g. by comparing a defender with high db and low pb, and then vice versa).

We need to determine how the displayed value is calculated and what kind of roll it is first. The easiest for this is an IMM to tell us. Outside of that we need to do various ob vs only various db and various ob vs only pb with large data sets to determine which roll pattern is most closely followed.

Then, if the messages do seem to indicate this, you could use that to test whether the two stats seems to be independent (one roll, followed by the other), or interacting in some way.

Just a thought...

Once we know the roll patterns for the melee any stacking would be easily verified.

For example
"one roll, followed by the other"

1d(ob=3) vs 1d(db=1) vs 1d(pb=2)

successfully defend 50% ("dodge" 67% of successful defense, "parry/deflect" 33%)

"stacked straight sum defense"

1d(ob=3) vs 1d(db+pb=3)

successfully defend 83% ("dodge" 40% of successful defense, "parry/deflect" 60%)

What I found anecdotally is that either on its own is not fantastic, but dodge above certain threshholds seems to really help pb

This would seem to indicate some stacking. Once the roll are setup even some stacking should have enough of a difference to easily tell if defense is skewed from the one roll, followed by the other.

[Raeza]

Ah, you are talking about completely working out the underlying system for combat. I had a much more modest goal in mind, which was just answering the following questions:

1. Are db and pb independent, or do they interact?
2. Are there 'threshold' levels of db/pb, or is their effect continuous?

I think these questions could be answered without requiring a ton of testing.

Step 1: Do "you dodge" and "you parry/deflect" correspond to a roll against db and pb respectively?

To test: collect one set of data with an attacker with ob of 100 and a defender with db of 100 and as little pb as possible. Then collect another set of data with the same attacker, but with a set that gives db near 0 and pb of 100.

--> If the wording does correspond to db/pb, then virtually all of the misses in case 1 should be "you dodge" and virtually all in case 2 should be "you parry/deflect".

Step 2: Do db and pb interact, or do they seem independent?

To test: calculate the percentage of misses in each of the two cases tested above (100 db 0 pb and 0 db 100 pb, vs. 100 ob in both cases). Let us assume that the percentage of hits is 50% in both cases for the sake of simplicity. Then, collect a third set of data where the defender now has 100 db AND 100 pb.

--> If these are two independent rolls, we would expect 50% "you dodge", 25% "you parry/deflect" and 25% hits, if db is rolled first. (The opposite if pb is rolled first.) If they interact in some way, then we would expect some other distribution. For example, if there was 50% "you dodge" and 40% "you parry/deflect", and 10% hits), then we would assume that there is some sort of stacking effect, where dodge adds some additional bonus to parry and they are not totally independent.

Step 3: Are there thresholds, or is it continuous? (Maybe focus on db here, since this is the one where people often say there is a threshold.)

To test: Use an attacker with a constant ob (say, 160). Some trial and error might be needed here to get the right ob to get sufficient hits at the different db levels. Vary the db of the defender from 145, 135, 125, 115 and collect data at each level. Keep pb constant and low, to reduce its influence.

--> If db is just continuous, then there should be a regular decrease in the probability of hits as db is increased. If there is some sort of threshold, then the data should look different. For instance, there might be a regular decrease in hits until you get to 145 db, where there is suddenly a much larger drop in the probability of a hit.

What do you think? This would not allow us to completely reverse engineer the underlying combat system, but it might allow us to answer some fundamental questions about it without requiring exorbitant amounts of testing.

[Mhaliah]

IF all you wanted to know is if there is any stacking at all that is relatively simple. Given we assume displayed value = max roll and that the xdy setup is identical for each of the bonus rolls, and that stacking is a rated sum not a separate roll.

ob = pb db=0 (test1)
ob = db pb=0 (test2)
ob = 2xpb = 2xdb= pb+db (test3)

The first two must be the same total successful defense rate. If they don't our assumptions are wrong or another variable is not being controlled.
If there is straight stacking test3 should have the same total successful defense as the first two. If there is no stacking the final system should fail to defend at a significantly higher rate. If there is rated stacking (i.e. x%maxdb + x%maxpb) that is not just a straight stack we could tell, but we may not be able to tell if there if a difference in the stacking rate (%) between db and pb only the combined average. This may be able to be determined if the dodge vs parry vs deflect is meaningful.

We would need at least 100 melee rounds for each but i would suggest closer to 1000 and we would need to exclude the initial attack when the target is not engaged. It will likely be impossible to get 0 values for either pb or db. Keep it as low as possible and if the lowest db is x then when you do test2 make the pb = x as well. Once again as close as possible.

If you want to collect the data in a log let me know.

Step 3 from your post is huge. Basically rinse and repeat data collection for all the various levels.