Automating the RMAH

[Markco]

Hey all, today i'd like to discuss automating searching with third party tools on the rmah.

Now before you start screaming: "Witch, burn him! He's breaking the spirit of the game!" consider the fact that there is already a gold tracking tool on THIS very website!

Now,to the matter at hand. Our biggest obstacle with developing an auction house searching tool is the variety of stats available for the same item.

The same mace could have 3,4,5 strength and 2,3,4 vitality. How do you distinguish between them without crushing your database?

I came up with assigning cost per stat and using the best item stats possible as 100% item value. Less stats take away from the percentage. An item with half the possible stats is worth half the value.

What do you think? Any suggestions?

I have a post on my blog as well.

[Pheal]

You obviously are misinformed, might want to check your sources when you say it's automated and it's doing "auction house searching". If taking screenshots considered as a third party tool usage ingame, you need to be checked OR read the TOS/EULA, which you never did (or did you now that you're banned from the beta for going against it?).

I'll try to help you, so you can risk getting banned on retail with your try at an automated tool:

Please learn basic economics and basic database programming (gotta start somewhere), "I came up with assigning cost per stat and using the best item stats possible as 100% item value." It obviously doesn't work like that, might want to read up on offer and demand. You'll see it's not linear at all and you can't assign a top value to an item / stat anyways (unless you want to create a fake perceived value and fool people with it, but that's mean, is it?). "The same mace could have 3,4,5 strength and 2,3,4 vitality. How do you distinguish between them without crushing your database?" How do you store these values? That should help you answer your question. Look up data integrity constraints and validation rules with primary keys and foreign keys.

[Azzure]

Google themselves probably couldn't program something like the thing you are suggesting to give positive returns. WAY too many variables, and the sheer nature of the item system will make it completely wrong.

You would also need to be an expert in how the item system works. Did you know that certain affixes cannot co-exist with each other, and that every affix has a different % chance to appear on certain ilvls? Meaning that there is a 30% higher chance that Strength XII can appear on a ilvl 50+ item than Strength XI? How would that affect S&D? Times this by a million and that's how complicated this would be

Consider this: There is in excess of a quadrillion item combinations in Diablo 3. Trying to map that out on a database to a logical valuation is impossible, and will fry your brain before you even begin.

[jwfcp]

Its murky, theyre actively prohibited here, though its ambiguous if theyre saying no in a "at all" or "for profit" capacity.
http://us.blizzard.com/en-us/company...legal-faq.html

You are right that valuation is going to be really tricky, especially as flavor of the month builds and patches ebb and flow. The focus is really defined by how much clutter needs to be cut through to get a clear picture. See, due to the values sliding on logarithmic scale the closer you get to perfect, with the valuation of off stats varying very widely from player to player (from no value added to exponentially perfect) its just not something that can be prechewed for the user.

Range filters and multiple layers of sorting would be the way I approach it- say Im selling a helm with near perfect strength, so I choose the slot and set the upper end of the str filter to perfect and the lower end to a bit below what I have. I sort by price and maybe notice that all the ones with crit are selling for crazy amounts of money, so I set the filter to how much crit I have (zero) and I also tighten the range to a point shy of perfect str since I notice the perfs are exponentially more valuable than what I have. I also might want to tighten up the date to the last few weeks, since I cant jump in my time machine and go back 6 months when anything that dropped from nightmare was crazy valuable(at the same time if I have something so crazy rare that it only drops once a year, I would want to see records of similar sales as well, no matter how far back it happened). From there I might have burnt away enough clutter to notice that people seem to be paying a crazy amount if it has a decent amount of stam+goldfind on it and I just happen to have a decent amount of both of those etc.

You might be able to get away with putting alot of these values into a tiny int, but many might need a small int for some values, or if stat inflation rears its ugly head. small money should be able to handle prices over money etc. Looots of indexes, multifield indexes, a table to track what people query for (maybe you want a way to have them send this to you? its the creme of the crop, the info to gear indexes to the program run most efficiently in future builds and info to corner the market).

[Naviaras]

Another issues would be how to collect all that data. Lets take a look at an example, 1 million items with 100 items displayed on screen at one time. That would be 10,000 pages. At a speed of 1 second per page that would take 2m 46s. At 50 items per page it would take 5m 32s. This isn't even taking account of the time it would take to adjust the filters or that in that time x% of the times would have sold and y% more items where added.

Now, I assume that you can't move that fast so you'll have to use a program to collect the information. Congratulation, you used an unauthorized 3rd party program, now your banned.

I'd say if a fairly safe bet that Blizzard will keep an eye out for any account that is pinging the RMAH that fast.

Your best bet for any sort of automation is for Blizzard to release an out of game system to access the RMAH. Until that happens your flirting with a Ban. While Blizzard may soften its stance later I think it it will be quite draconian in the beginning with any unusual 3rd party programs running in the background.

[jwfcp]

Very true, though suppose you set out to corner some niche of the market (say the combination of crit and +dmg), a manual appraisal of the market would be much easier to accomplish, all a tool like this would be doing is making it easier to keep your ducks in a row, I certainly wouldn't turn it down if it was a feature.

[Markco]

The tool is already under construction, pulling data live from beta and making calculations.

What I'm looking for help with is a solution to identifying the value of items with different stats but the same name.

How do you calculate that? That's what we're brainstorming now.

[Markco]

You obviously are misinformed, might want to check your sources when you say it's automated and it's doing "auction house searching". If taking screenshots considered as a third party tool usage ingame, you need to be checked OR read the TOS/EULA, which you never did (or did you now that you're banned from the beta for going against it?).

I'll try to help you, so you can risk getting banned on retail with your try at an automated tool:

Please learn basic economics and basic database programming (gotta start somewhere), "I came up with assigning cost per stat and using the best item stats possible as 100% item value." It obviously doesn't work like that, might want to read up on offer and demand. You'll see it's not linear at all and you can't assign a top value to an item / stat anyways (unless you want to create a fake perceived value and fool people with it, but that's mean, is it?). "The same mace could have 3,4,5 strength and 2,3,4 vitality. How do you distinguish between them without crushing your database?" How do you store these values? That should help you answer your question. Look up data integrity constraints and validation rules with primary keys and foreign keys.

Believe it or not I code for a living. Regardless, storing an item's value is not so simple. A perfect item will sell for far more than a nearly perfect one.

We don't need help with storing, we're working on what's worth storing and telling value of items apart.

If we store every combination of item stat setup per item we'd have no way to store it all.

Yes azzure. We know about affixes. The guy from diablo3inferno is the one doing the coding work. He's probably as good as the blizzard coders.

Thank you all for constructive comments.

[jwfcp]

Thought about this a bit more; To keep it simple, gear it for top quality gear. Pick an arbitrary threshold for calling something "near perfect" (maybe 5%-10%?) and put the threshold for worth storing to at least 1 perfect stat, or 2 near perfect stats. That lets you store and look up the data very efficiently in a bit map with values representing {has perfect, has near_perfect, doesn't have} instead of data. Finally, instead of comparing against the absolute max possible stat for an affix, take the item level into consideration and compare its stats against what is considered perfect for that level. queries would need to take ilvl into account too due to this, and be a real mess at that.

Lets say resistance has tiers at level 18, 24, 30 etc, while crit has tiers at 20, 25, 30 etc. so if you want to know about a lvl 25 item you just found with near perfect in both crit and resistance, it wouldnt be appropriate to compare it to a lvl 24 item which has perfect in the same pair of stats because what is considered perfect crit for the level 24 is vastly inferior to what is considered perfect crit on the 25. You may want to consider going absolute stats afterall, by the time you are in inferno you won't care about the nuances of whats hot in lvl 24 and 25 trash.

[BohemianStalker]

I have a post on my blog as well.

Yes this is point of your every message. To advertise your stupid blog.

You know what? I think you should be banned from this forum.

You juts make things up to get some attention.

AND Your blog is full of wrong info and lies.

Just like the lie that you were banned for making too much gold, you were actually exploiting the bug in beta.(google it people if you want)

And you are filling this decent forum with you wannabe millionare posts.

Go somewhere else, I had enough of your bull****s.