Yes that is exactly it, I've played around with different spell haste values myself in the spreadsheet and it does at certain points reduce your dps unless you tweak the "wait" value for each spellhaste number. If I did that I always got an increase. So for it to be accurate you would need a macro or something that also finds the wait value (and possibly also mindflay interrupt value) for each haste number that results in the highest dps.

I agree that without tweaking the 'wait' values the complete picture of the gains from haste will not be realized but because of the nature of the graphs that are being produced by fully stepping out haste from 1-x in single point incriments the tweaking need not be done to attach a meaningful weight to haste as a stat.

All of the graphs that I have produced with the macro so far have all demonstrated the exact same behaviour which is, namely, a step wise linear function. Adjusting the wait values in the sheet would ultimately produce a completely linear graph or at least an aproximation of one given the amount of variance in the sim itself. That is to say the slope of the line is what matters, not the nominal DPS number that is being produced by the sim.

That said adding in logic to the macro that creates the values to adjust the wait times for maximum DPS is completely possible but as long as it is just guessing and checking it would make the function more time consuming.

Wrong. Tweaking wait is *absolutely* required to produce meaningful estimates of haste's value as a stat. A 0.1 second variance in wait can produce a variance of +/-2-3% in DPS, which is quite significant.

The weight of the stat is a function of its relative effect on DPS. You don't need to tweak the wait stat in order to determine that value because the graph is stepwise linear. All stepping out the wait values is going to do is smooth out the graph so it is completely linear instead of piecewise linear. The distinction between the two is completely academic because the slope is the weight of the stat and for the linear pieces of either graph it will be the same.

It does depend on what you're using the sheet for though. I should amend my argument by saying I would only be interested in using the sheet to assign approximate stat weights that can be used to select gear. If you are actually trying to created a DPS number to match with in game then, yes, the estimate of your DPS will require tweaking of the wait stat to remain accurate.

I think you're missing the point, I only care about the slope. The slope is the same whether you adjust the wait time or not, once you fully step out the data point to 1 haste rating or less. Unless you've already plotted a graph at that level of fidelity and observed something different in which case I'd be interested in seeing the data set. A change in the slope of the line because of changing the wait values would indicate that adjusting your spell casting will make the value of haste increase the more haste you have, which is contrary to what I've read in other threads and my own research.

The slope is not the same when you step the data out to 1 haste rating. I've done it. I've even provided code for other people to do it (sorry, not going to try to do simulation in Excel when I have MATLAB available). Haste rating's value is NOT LINEAR ACROSS THE RANGE. It is piecewise linear in small chunks, yes, but that implies slope variation. Adding wait timing simply changes the nature of the piecewise linearity - it doesn't suddenly make it linear across the range.

Adjusting your cast timing will, in fact, change the value of haste at a given point. Since the required value varies, and the graph itself remains piecewise linear whether you adjust or not, the slope change is meaningful given wait time variation.

Then your mathlab numbers are different than the simulation in the excel sheet because that is very much not the behaviour of the simulation in the spread sheet. In the excel sheet if the wait times are adjusted to optimal values given a specific haste rating and the DPS points at each value of haste are graph it does make it linear across the range. This makes sense because piecewise linear implies changing intercepts, not changing slopes.

What you are saying right now is that each marginal point of haste is worth a different amount, given a fixed set of other stats, depending on how much haste you have. The simulation in the sheet indicates that each marginal point of haste is worth the same amount, given a fixed set of other stats, reguardless of how much haste you have as long as you adjust your cast timing optimally.

A curve can be piecewise linear with a constant slope and a changing intercept. This is what I observed building a curve out of the data produced by varying the haste value and running the sim for DPS. Piecewise linear doesn't require different slopes because a piecewise linear equation where the disjointed segments that all have the same slopes but different intercepts will have a constant derivative across the entire range. The slope from 0 to 250-300 is a completely different value than the instantaneous slope at any given point.

Haste adding different amounts of DPS depending on how much you already is true on live, now that I've thought about it, because of the GCD lock. Assuming 2.4 goes live as is this will no longer be the case. My mistake. The sheet is built with the GCD lock built in so I maintain, based on the graphs that the sheet is producing for me with its simulation, that in the sheet changing the wait values still only changes the intercepts on the various pieces in the graph and not the instantaneous slope at any given point. Again, this is merely what I'm observing from the output of the sheet and if it's incorrect the sheet should be fixed.

I'm not going to fix the sheet. Excel for programmatic simulation is just a bad tool.

Attached: simulation using 2.4 GCD, in increments of 1 haste, timestep of 0.01 seconds, 15 minute fights, across 250 haste at 10 different values of wait. While the shapes are, overall, *similar*, they are not identical. Essentially, as wait increases, peaks/troughs are flattened out.

Note that further smoothing of the graph can be seen as wait is boosted. However, waiting longer isn't always optimal - sometimes it does lower your DPS, while making the graph across haste closer to monotonically increasing. Hence my stipulation that wait is not something you can simply set and have be valid across the range of hastes.

I have a noob question, I'm starting a new alliance shadow priest, for raiding end game... I was wondering which one would be better to roll, draenei or human. I pretty much hate the way draeneis look but I cant overlook the mp5 bonus and the spell hit. But at the same time, humans have 10% spirit and 10% rep gains(which anyone whose raided end game knows is a lot)

Obviously the draenei bring more to the table, but is it worth being horribly ugly too?

They really did buff star shards? I haven't played alliance since pre-bc, and I remember that being a trash racial. I am probably going to end up going human anyways, just because its easier once we hit hyjal/bt and all the other rep things. I'm steering clear of a healer, and I want a decent dmg class, and you can never have enough shadow priests, I figure that will be my ending decision.

Sorry about the sig, but I am too lazy today to go in and fix it.