How does the enhancement work?
There had been some speculation as to whether the enhancement was a) a fixed percentage and/or
Even at this early stage I feel pretty confident about answering these two points. Duelist offers a fixed % increase in your attack and damage stats which is always in effect during PvP. I will outline my reasons for this below, but this diagram broadly shows the effect as it would be seen on the damage stat. N.B. Since multiplication is commutative you could apply the variance first, but it makes for a less simple diagram.

How effective is the enhancement?
So far I have only tested the effect of the enhancement on damage, so conclusions here are only *assumed* to hold true for attack as well. Many thanks to wingyann and MaximusGR for being my test dummies to this point.
I have so far tested out 5 different levels of Duelist and hope to include a further 5 more shortly. Doubling the number of data points for each level would also be nice - and I need player help to do that, so PM me in game if you're in my GvG range!
For each level I compared the damage delivered to a nearly naked opponent (~50 armour) over 20 combats and compared the Minimum, Mean and Maximum values to those expected without duelist. The results are listed below and shown in the following graph.
100% Duelist
Expected Mean: 5887
Expected Minimum: 5593
Expected Maximum: 6182
My observed results were:
Mean: 6424
Minimum: 6185
Maximum: 6759
That equates to a 9.1% damage increase (on average) over 20 combats.
80% Duelist
Expected Mean: 5160
Expected Minimum: 4902
Expected Maximum: 5418
My observed results were:
Mean: 5617
Minimum: 5310
Maximum: 5846
That equates to an 8.9% damage increase (on average) over 20 combats.
60% Duelist
Expected Mean: 4973
Expected Minimum: 4725
Expected Maximum: 5222
My observed results were:
Mean: 5289
Minimum: 5018
Maximum: 5530
That equates to an 6.3% damage increase (on average) over 20 combats.
50% Duelist
Expected Mean: 4809
Expected Minimum: 4569
Expected Maximum: 5049
My observed results were:
Mean: 5096
Minimum: 4831
Maximum: 5281
That equates to an 6.0% damage increase (on average) over 20 combats.
30% Duelist
Expected Mean: 4622
Expected Minimum: 4391
Expected Maximum: 4854
My observed results were:
Mean: 4741
Minimum: 4519
Maximum: 4988
That equates to an 2.6% damage increase (on average) over 20 combats.
In all cases it's evident that there was never a point where duelist 'failed to activate' (or at least I was extremely lucky when it did), and that there is still a typical +-5% distribution on the final damage values. This is what gives me confidence in the example diagram above.
Here is the graph showing the summarised results:

The error bars are calculated from the 5% variance in the damage after duelist is taken into account, based on the averaging over 20 combats. They currently work out as +-1% or so, which I'd love to reduce. I've constrained the best fit line to pass through the origin for obvious reasons (no duelist = no bonus
You can see clearly from the graph that despite a bit of a spread in individual test samples the trend is towards the following HCS Style Buff Description:

Like I stated earlier, I've only tested damage values so far, so I'm only assuming that the same equation applies to attack as well. Based on previous HCS coding examples I would be surprised if it didn't, but I would like to test it out.
We can dig a little deeper into our results by checking them against the hypothesis presented in that tongue in cheek buff description.
How likely is it that we got the observed results if that hypothesis were true?
Well, so far we only have 5 data points (100, 80, 60, 50 and 30%).
Running a Pearson's Chi-Square test on just these data points gives a P-value of 0.97, meaning that for this hypothesis we have an excellent fit to the data!
Where do we go from here?
First of all I'd like to get at least another 20 combats for each Duelist value, and fill in the gaps at 10, 20, 40, 70 and 90%. That should nail down the damage boost and reduce the error on each sample by around 30%. I'd also like to test out a couple of different aspects of the attack/defence balance, as I've had some interesting PMs from some top players that I think are worth checking out.
As always, comments and criticism are welcome and I hope you enjoyed the third part of Mythbusters! I'll be making the data for all 3 of my enquiries available online shortly if anyone wants to have a look at the raw numbers.
Myth: Busted
RJEM
EDIT: I have now tested this enhancement for attack as well. The boost is the same as for damage - a permanent 0.1% per 1% Duelist enhancement as far as I can tell.
HOWEVER, ATTACK DOES NOT INTERACT WITH DEFENCE IN THE SAME WAY DAMAGE INTERACTS WITH ARMOUR. PLEASE SEE MYTHBUSTERS PART 4 FOR FURTHER INFORMATION.


