Talsyrius Posted July 14, 2012 Share Posted July 14, 2012 (edited) I have seen a lot of people ask about this, and a lot of theories about how to deal with it, which is why I would like to clarify it once and for all. In my first attempt at this, as the nice gentleman below pointed out, I was monstrously wrong, but after scratching my head a bit, and poking the internet and his links, I think I got it, or I will at least give it another go. It is easiest to compare every stat gain to power, because that is a static gain regardless of how much you already have. So lets start! To be able to handle the diminishing returns on some stats, you will need to find the derivatives of said stat at the amount you have, for that I suggest you use this site to enter the code for corresponding stat, posted below. http://www.numberempire.com/derivatives.phpThen you will have to input how much you currently have in that stat as the value of x, and you will be presented with a decimal number, that I will refer to as "stat gain" that you will have to enter in the formulas below. A hot tip is to enable "combat log to file" at your in game options, and then upload the file to http://swtor.askmrrobot.com/combatlog To be able to evaluate where your damage comes from, and which effects apply to the different skills. POWERPower adds a static 0.23 bonus damage, regardless of how much you have. MAIN STAT (strength, willpower, cunning, aim)This adds a static 0.2 bonus damage, but it also adds to your critical hit chance, which usually makes it superior to power. To get the comparative value, do the following:0.2+(stat gain)*0,01*bonus damage*critical multiplierTo get the stat gain, use 30 * ( 1 - ( 1 - ( 0.01 / 0.3 ) )^( ( x/ 50 ) / 2.5 ) ) at the site posted above*note that you can add an additional 5% to this (multiply by 1.05) if you have the consular/inquisitor buff**note that this isn't accurate if you have some skills or procs that temporarily increase your critical hit chance for certain abilities to 100%, because whenever that proc is on, critical hit chance adds 0 damage CRITICAL RATINGThis work almost like your main stat, to get the comparative value, do the following:(stat gain)*0,01*bonus damage*critical multiplierTo get the stat gain, use 30*(1-(1-(0.01/0.3))^((x/50)/0.45))at the site posted above*note that this isn't accurate if you have some skills or procs that temporarily increase your critical hit chance for certain abilities to 100%, because whenever that proc is on, critical hit chance adds 0 damage SURGE RATINGAgain, very similar way to calculate:(stat gain)*0,01*critical hit chance*bonus damageTo get the stat gain, use 30*(1-(1-(0.01/0.3))^((x/50)/0.11))at the site posted above*note that this isn't accurate if you have some skills or procs that temporarily increase your critical hit chance for certain abilities to 100%, because whenever that proc is on, you have to remove the critical hit chance component of the formula, making surge more useful. ACCURACYUp to 105% accuracy on main hand (every class have at least 5% defense chance by default), for white damage attacks:(stat gain)*0,01*(min weapon damage+max weapon damage)/2To get the stat gain, use 30 * ( 1 - ( 1 - ( 0.01 / 0.3 ) )^( ( x/ 50 ) / 0.55 ) )at the site posted above*This however isn't entirely true, against someone with 5% defense chance, the real formula would be (stat gain)*0.01*0,95)*(min weapon damage+max weapon damage)/2 up untill 100% accuracy, and at 100%-105%:(stat gain*0.01*(0,95+(hit chance-1)))*(min weapon damage+max weapon damage)/2 EXPERTISENow this is a bit tricky, especially the defense part, because you don't have anything like your melee or bonus damage to compare it to. For the defense, I would suggest you first use the bonus damage, then play a few games with combat log on and find out how much damage you take compared to how much damage you do (or check scoreboard) and use that as reference. For example, if you have 800 force bonus damage, you do on average 300k damage in warzones, and you take on average 150k damage, the "gain" in damage numbers, would be half of your bonus damage, meaning you should use 400 as the multiplier (I will refer to this as "damage taken"). PvP damage increase:(stat gain)*0,01*bonus damageTo get the stat gain, use 50 * ( 1 - ( 1 - ( 0.01 / 0.5 ) )^( ( (x) / 50 ) / 0.8 ) )at the site posted abovePvP damage reduction(stat gain)*0,01*("damage taken")To get the stat gain, use 100-100/(1+x/100)at the site posted above. The value for x in this case is NOT how much expertise you have, but rather how much PvP damage boost you currently have. I would like to get any comments on however this is right, because I think (hope) it might be this time:D Edited July 29, 2012 by Talsyrius Link to comment Share on other sites More sharing options...
Talsyrius Posted July 21, 2012 Author Share Posted July 21, 2012 Bump Link to comment Share on other sites More sharing options...
Tibbel Posted July 22, 2012 Share Posted July 22, 2012 (edited) What you've done here is calculate an average gain per point for each stat, and then use that to linearly extrapolate its value. Allow me to argue that this can be a misleading method to use. To start, I graphed the gain-vs-rating curves for both surge and crit rating. (I'll use those two stats as examples.) Here's a link to the graph. The value-per-rating for each of those stats is seen as a slope, so I also found the running average for each stat as you did above (selecting 400 rating as an example point), and then graphed those as red lines. (I didn't include bonus damage or the other constant multipliers that you had in your formulas.) Lastly, I found the current value-per-rating for both stats at 400 rating, which is represented graphically (in green) as a line tangent to the instantaneous slope at that point. The resulting pairs of lines show that:The running average method overestimates the value-per-rating for both stats.The degree of overestimation is not constant, with surge overestimated more than crit.At these fairly achievable rating levels (400 each of crit and surge), this difference in overestimation is large enough to misleadingly conclude that surge gains more value-per-rating than crit does at 400 rating, when in fact the opposite is true. To accurately calculate the value-per-rating of stats with DRs, you can solve the resulting differential equations (or find a graph that has those DEs already solved). However, even that's not immensely useful because of how many class mechanics interact with certain stat values (e.g. crit procs, ranged/melee crit vs tech/force crit, varying coefficients and/or bonus crit damage on various abilities). To get a true feel or values of each stat, we need either immense amounts of empirical testing (nobody in SWTOR has even come close to doing anything near what would be needed) or a method to simulate game situations based on known mechanics. Over at http://mmo-mechanics.com/swtor/forums can be found several different flavors of simulation, including spreadsheets like the one linked in my signature all the way up to the full-blown combat sim SimulationCraft, in addition to formulas for answering simpler questions. Edited July 22, 2012 by Tibbel Link to comment Share on other sites More sharing options...
Talsyrius Posted July 22, 2012 Author Share Posted July 22, 2012 (edited) Thanks for your response! I should probably have been clearer on this point, I didn't mean that you would use the values you get and extrapolate to get some sort of fixed number, rather I would suggest that you do those calculations every time before you buy a new piece of equipment to see what is actually most needed. I might be rather dumb, but I don't see how that second graph you made with surge crit and cunning on the same sheet could be correct, I thought the percentage gain from your main stat was dependent on how much you have in the other stats. For example, if you have same amount of points in to your main stat, but your power differs greatly, so that you have 10k bonus damage(too much I know) one point to main stat would add just a teeny tiny percentage gain to your attack (ca 0,002%), because the increase is for the most part a fixed numerical value (not the critical hit chance compartment). Reversed, if you have 20 bonus damage, one point in to main stat would give a much higher (ca 1%) more damage. i.e you can't successfully make a generalized comparison for all values if one stat that have a fixed numerical gain to one that have a logarithmic percentile gain because of the way they interact with each other. So my point really is that while solving the differentials might render nice looking graphs, they won't give you an exact comparable result, where as my way gives an exact answer for your current gear level, if you only add one point. I think this is more meaningful when you should try to extrapolate how much you would gain from a new set piece, where the increase for default gear is about 60 more in each stat at the most, which gives you a reasonable chance to decide which stat to prioritize. If nothing else, it is a good way to check if you where right in your assumption. I must apologize for my bad English, it is really cumbersome to not have this kind of discussion in my native language D: Edited July 22, 2012 by Talsyrius Link to comment Share on other sites More sharing options...
Tibbel Posted July 22, 2012 Share Posted July 22, 2012 (edited) Here's a more numerical approach to my argument: 400 crit rating gives us a 13.58% increase to crit chance. 400 surge rating gives us a 27.45% increase to crit damage. Using the running average method, we'd find that at that point crit rating rating has given us an average of: 13.58% / 400 = .0340% crit chance per point Likewise, for surge: 27.45% / 400 = .0686% crit damage per point So this method predicts that adding 1 rating would give us either +.0340% crit or +.0686% crit damage. However, if we use the actual formulas for stat gains as functions of their rating (which are available at http://mmo-mechanics.com/swtor/forums/thread-SWTOR-formula-list, we see that adding 1 rating on top of 400 in fact only gives +.0246% increased crit chance or +.0152% increased crit damage. These values are not only smaller than those predicted by the running average method, but also reversed in position. That's because surge has a stronger DR affecting its rate of increase, i.e. its slope or first-order derivative, which is what the second graph I posted shows. Also, since you mentioned it... The increase in a stat is not dependent on any other stats, only on its own rating. As an example, the crit chance that Imperial Agents gain from cunning is not dependent on how much power we already have, or even on how much crit chance we've already gained from crit rating or from aim -- it's only dependent on how much cunning we already have. How that stat increase affects our overall DPS, however, absolutely does depend on all our stats, as well as our spec, our target, as well as which abilities we use and at what frequencies. For complex calculations like that, there aren't simple all-encompassing formulas for a stat's value. Instead, the best option (in my opinion) is to use a combat simulation like a spreadsheet or SimulationCraft. Edited July 22, 2012 by Tibbel Link to comment Share on other sites More sharing options...
Talsyrius Posted July 23, 2012 Author Share Posted July 23, 2012 (edited) But um... Since the point of this is to get as much damage out of this as possible, how can you say that stats don't affect each other? If I have 10 bonus damage, one point in to critical rating will give us same percentage increase as if we had 1000, but if you should chose between critical rating and power, I would strongly suggest you go with power, because you would gain a lot more damage from that.... (if you have 10 bonus damage). Edit: I am sorry, was in a real hurry last time I read your answer and made the above reply, so didn't really get it then... and when I did I was at my parents house without security key so couldn't log in to reply! Yes I see what you mean now, I should really derive (or however it is called in English) the function for x=stat value! I kinda feel silly not seeing that... although it could have something to do that I failed "mathematic analysis in one variable" or however it translates... But um... the formula for the derivative should be static, and just substitute in x for respective stat, and insert that result in my "very scientific formulas" above? Edited July 24, 2012 by Talsyrius Link to comment Share on other sites More sharing options...
Talsyrius Posted July 26, 2012 Author Share Posted July 26, 2012 Bump on account of correcting the formulas Link to comment Share on other sites More sharing options...
Recommended Posts