DEX with War

[Gamemako]

I am not aware of any testing involving sufficiently small variations in DEX at a sufficiently high level of confidence as to support your assertion. Care to provide a link to your source(s)?

If you review review Ninjiitsu's data and Hulan's data via Valk, you will find no other reasonable conclusion at this time. No, the data is not good, and the relationship is complex (especially with Hulan's data, which includes both parry and blocking both), but, there is an effect observed with every DEX interval.

So the tool tip about increasing healing received via spells is false? I'm surprised SE would would advertise a false statement on their tool tips.

As Paikis pointed out, it doesn't actually say that. The wording is just a bit poor. However, tooltips are pretty bad in general in ARR. "Clusters are unavailable in the beta version", anyone? And Steady Hand tooltip on the buff itself is wrong, saying 10% instead of 20%.

[Phreak]

As Paikis pointed out, it doesn't actually say that. The wording is just a bit poor. However, tooltips are pretty bad in general in ARR. "Clusters are unavailable in the beta version", anyone? And Steady Hand tooltip on the buff itself is wrong, saying 10% instead of 20%.

For spending 2ish years rebuilding this game 'from the ground up' it's a bit disheartening to see how much seems to be a copy/paste job.

[Ysarel]

If you review review Ninjiitsu's data and Hulan's data via Valk, you will find no other reasonable conclusion at this time. No, the data is not good, and the relationship is complex (especially with Hulan's data, which includes both parry and blocking both), but, there is an effect observed with every DEX interval.

Neither data set nor the combination of the two provides enough data, much less good data, to support your conclusion. What conclusions we can draw from them does not extend to anything remotely close to sub-1% levels of accuracy. If this is all there is then we do not have enough data at this time to determine how exactly the effect of DEX on parry is calculated.

[Gamemako]

Neither data set nor the combination of the two provides enough data, much less good data, to support your conclusion. What conclusions we can draw from them does not extend to anything remotely close to sub-1% levels of accuracy. If this is all there is then we do not have enough data at this time to determine how exactly the effect of DEX on parry is calculated.

You are demanding irrelevant information to irrelevant certainty. You want the absolute value of 1 point when no player will take just one point, nor is there any reason to create such an estimate. I have repeatedly stated that we do not have enough data to piece together any formula. It is actually likely that there is no such point due to co-dependence. However, there is no interval within the data (between 15 and 30 points) which proffers the existence of any tiering effect. For one to exist, the tier would have to be less than 15 points of dexterity and placed conveniently such that an interval is crossed between 211, 216, 226, and 246, but not more than once. This is either an incredibly-fortunate placement of a 14-point interval (213, 227, 241) or it doesn't exist, and Occam's Razor cuts the former to ribbons. It is extremely unlikely that DEX is based on tiers.

[Ysarel]

You are demanding irrelevant information to irrelevant certainty.

No, I'm demanding that you provide adequate support for your assertions. If you are going to claim that every point of DEX is beneficial for a tank, you should be able to back that up. You have thus far failed to do so, instead making claims that have less to do with the data presented than your own wishful thinking and lack of knowledge of statistics.

I doubt anyone here would dispute the notion that parry rate is directly proportional to DEX and the parry stat. But did Ninjiitsu's parsed parry rate not decrease in one test when he increased his DEX by 27 and his parry by 16? Was his parry rate not virtually identical (+0.2%) to the first test in a third test wherein he increased his parry stat by a further 30 points? There are similar instances of results running counter to expectations in Hulan's data, where in several instances lowering block rate without varying DEX or parry also exhibited a decrease in parry rate where one would expect either an increase (if block is checked first; if memory serves someone confirmed this but I'd have to dig it up again) or no change (if parry is checked first). Should we draw some bizarre, convoluted conclusion from this data? Of course not. The cause is nothing more than low sample size.

EDIT: Before you attempt to twist my words around, I take no position as to whether the effects of DEX are tiered or not. I claim only an absence of sufficient information to draw a conclusion regarding this question.

[Gamemako]

If you are going to claim that every point of DEX is beneficial for a tank...

First, let's review what I said:

Adding DEX in every tested case that anyone has done has increased parry rate. The exact formula isn't known, but nobody has managed to find a case where adding DEX did not increase parry rate. That does not mean the addition of a small amount of DEX will be noticeable, however.

Now, it's funny, but I don't see "every point of DEX is beneficial for a tank" anywhere in there. But let's address this anyway:

There are 5 possibilities for how DEX affects parry rate:

1) Adding DEX offers no benefit.
2) Adding DEX offers benefit at a fixed rate
3) Adding DEX offers benefit at specific tiers
4) Adding DEX offers benefit up to a given cap
5) Adding DEX offers benefit dependent upon a third value with no cap

It is clear that (1) is not the case. Current data is not conclusive with regards to (2) -- it is possible, though the discrepancy in parry rate added per point leads me to believe that it is not the case. (3) is what I just addressed -- it is extremely unlikely that there is any tiered effect. For (4), there is not sufficient data to identify any potential caps. (5) is a reasonable candidate.

Now, since (1) and (3) are clearly not the case, then the only way that you would not get value out of DEX is assuming that you have already reached the theoretical cap of (4), which we lack any evidence exists. If it does exist, then at 456 parry, it's greater than 246 dexterity and extremely likely at least 256 (assuming no dependence between parry and DEX benefit as well as assuming that the other data is correct and DEX adds 0.11833% per point, the theoretical cap would be this value).

//EDIT:

I doubt anyone here would dispute the notion that parry rate is directly proportional to DEX and the parry stat. But did Ninjiitsu's parsed parry rate not decrease in one test when he increased his DEX by 27 and his parry by 16? Was his parry rate not virtually identical (+0.2%) to the first test in a third test wherein he increased his parry stat by a further 30 points?

So you're basing all of this on the incomparable outlier? The one which provides an impossible result based on any imaginable formula, from walls to tiers to anything shy of a screw-you-stat? Kay.

There are similar instances of results running counter to expectations in Hulan's data, where in several instances lowering block rate without varying DEX or parry also exhibited a decrease in parry rate where one would expect either an increase (if block is checked first; if memory serves someone confirmed this but I'd have to dig it up again) or no change (if parry is checked first).

Block is checked first. Increases in block rate reduce parry rate, which must be increased to stay relatively constant.