I decided for shits n giggled to score my barbarians personal gear, cause i never seem to find anything remotely close to what he is wearing. I found that my bracers that i am wearing that i purchased for 15 million, only scored ~24000. Now while i admit they are a far cry from the best bracers.. i would think the purchase cost of 15 million would warrant them scoring a little bit better and over my minimum score of 25000. Stats: Str 70 Vit 70 Int 45 single res 40 all res 70 armor +250 crit chance 6% (Rounded the stats a little, just to be cautious ) Now the trouble i am having is i cant work out a way to score them better without opening the flood gates for a lot of junk. As you all know they score so low because of the low primary stats not triggering any of the possible scores.. but as a collective they all add up to something worth selling. I was trying to work out a way that if a item has lots of stats that a class is looking for (say 4 or more), the minimum % for each stat could be a little more lenient. But it does seem a little more trouble than its worth. Possible Examples i.e Barbarian rimary Stats:Str/Vit/All res/Crit%/Crit dmg/Armor Demon Hunter Dex/Vit/Crit%/Crit Dmg/AttackSpeed My other idea was maybe if the items primary are under the threshold, they can still get a score, but its reduced by a further X%, or possible they only get the reduced % score if a stats such as crit/crit dmg/attack speed is present. Anyone got any easier ideas to score that sort of item better without scoring worthless items to high? Regards Jarrod
The scores don't relate to auction house values. 24000 for a pair of bracers is a good score The default armor pickup scores are much lower than that. As long as you don't try to equate "Score = AH price... the higher the price, the higher the score", you'll probably be fine - scores simply are not designed in that way, because it'd be frankly impossible to relate the two (if it *WAS* possible, then I'd be able to make the plugin suggest prices! ). If you find valuable items being scored too low (and 24000 is not too low!), let me know the stats etc. and I can work out special circumstances that need to be taken into consideration, similar to what you mentioned.
Yeh im aware of that and thank you for taking the time to read the thread and reply. I am mainly just looking for help to "remove the grey" between items worth keeping and items that should be vendored, your scoring does a excellent job at sorting out the majority of the items. All i was wanting to do is work towards improving the scores, by figuring out if a item scores fairly low when its worth alot of money, why did it score low and how could it be scored better? And if a item scored well that isnt worth much, why did it score so well and how can the score be reduced? What makes it so hard is cause whats worth a lot of money is always changing I might try and tinker with a way to apply a final multiplier that "increases" the score if the item contains certain combinations that people are looking for and "decreases" the score if they do not. I'll let you know how i go, im not too good with this language, i was much better with autoit. Regards Jarrod
There's some great examples already in place that do exactly that - they analyze all the stats based on an item type *AFTER* each individual stat has already been scored, and can then apply final multipliers (up or down) - eg for quivers and other off-handers, or items with the combo crit hit/crit damage/attack speed etc. eg search for the comment "// Quivers" (exactly that nothing else) - you'll see in that area of code (both above and below) some "special post-stat-scoring" that gives final negatives/positives after each individual stat has already been scored. You can also see that individual stats themselves can impose a "global modifier" on all following stats in the main scoring code (find all occurrances of the variable "iGlobalMultiplier" and look through you'll see special stats at certain values can add to that, and that gives a FINAL score-boost to the total score after all the individual stats have been processed). There's a lot of ways you can achieve tweaking scores - be it in the individual stat-scoring loop, or in the final bonus calcs at the end. Finally, to help you out MASSIVELY - find this variable; "bFullAnalysis" - and set it's value to true, instead of false. Now whenever Trinity scores an item (eg from hitting "test scores" in the config) - you'll get a big detailed output telling you how the scores were calculated, so you can see which stats were giving most of the score, which stats were being underscored, etc. - but be sure to remember to turn it off when running the bot normally, as otherwise you'll be spammed like hell with these outputs each town-run. I try and only have 5 or 6 items in my backpack at a time that I am comparing the AH values with when tweaking and running test-scores with "fullanalysis" on. I'm usually pretty good/quick at tweaking scoring and removing junk etc. but I just don't have the time to spend hours comparing every "probably junk" item in my stash to equivalent Ah values and shifting scores up or down. If you make some good changes, be sure to pass them on to me and I can include them in a public release - but always check items that SHOULD be scored highly haven't accidentally been "nerfed" by changes first
Thanks for that Gile, thats a huge step in the right direction, far out you have so many lines of code for just scoring, that "bFullAnalysis" will make it way easier to understand whats going on, was sitting there with my calculator today for about 30 mins trying to manually calculate the score. Yeh i dont know how you manage to spend quite so much time on it as it is and manage to reply to so many people. I'll have a play with it when i get bored at night and see how i go. Is there any sort of "item worth" your considering when you classify an item as "not junk", like over a million?
I generally make sure that items go for a MINIMUM of like 500k before I try and make the stats of that item give it a lower-than-default score (and always have to be careful not to remove the scores of good items in the process). But 1 mil is probably a safer benchmark to start basing stuff on, these days.
Hey Guile, i thought id start with just the 1 handed weapons and work out a way to modify them by transferring the equations that the plugin uses to calculate weapon score into excel, that way i could more easily edit the parameters and quickly test a large array of weapons in excel, making sure i dont mess up the scores of good items/bad items. I can work out the calculations for all the stats, but from there im a bit lost to the final modifiers. This one weapon has high dex/str and scores : 12556/12293 for them, just under the bonus threshold Score 16000 for a socket and 39024 for dps. Thats all easy to follow, but then it somehow gets to 69731.89 total score pre special reductions then 80192 final score But it says the the global modifier (GM) = 1. I have been through all the code on scoring and cant find the section for score reduction due to low dps, found the 2 hand n range weapon sections Also not sure if it loses score due too it having no vitality, as for alot of that section im pretty sure it excludes weapons Lastly how much bonus score is added due to exceding the bonus threshold for sockets Regards Jarrod
