5 replies to this topic

[kingofaces]

This is obviously a tough idea to figure out just by watching shops alone, but maybe someone who works on the item database would have access to information on this.

Basically what I'm wondering is what determines how often certain items restock. For instance, the Brightvale Armoury has a relatively low number of high rarity items (r98, r99), while something like the Toy Shop has a large number of potential high rarity items that can restock. The question is, are you more likely to see a high rarity item at the Toy Shop than the Armoury as more high rarity items are aviailable, or is there an equal chance of seeing a specific rarity item regardless of how many items of that rarity are possible?

As an example:

Brightvale Armour - 8 r99s
Toy - 41 r99s

There are two potential ways that whatever program Neo uses to determine what restocks:

1. a. Restock occurs
b. Rarity of item to restock first selected with higher rarities less probable (in this case r99 occurs)
c. Item randomly chosen from the pool of all r99 items for that shop
or

2. a. Restock occurs
b. Item randomly selected from pool of ALL possible items in the shop with higher rarity items having a reduced likelihood of being selected

So with the two shops you have an equal chance of seeing an r99 under option 1. However, under option 2 it's more likely you'll see an r99 because there is a larger pool of r99s to choose from. There are other factors that could come into play such as how often the shop if full of items, but I would think all the necessary data is being gathered whenever people are using Abrosia. It shouldn't be very hard to tease apart the data either since it's pretty basic statistics. Has anyone looked at this yet? I think I'd be a nifty little stats project to look at that I could tease a lot of information out of it there is any data available. All that's needed is a log of what item restocked in a few different shops for a period of time (the longer the better, preferably at least a few days to a week) Then you'd just compare the occurrence of a specific rarity restocking over different shops.

So if anyone has looked at this already, it'd be cool to see how things might be working out. If not, I'd be interested in getting a hold of someone that works with the item database to see just how feasible it might be to check this out.

[kingofaces]

This is also just an open question to see if anyone has noticed a general trend at all? I mainly restock at shops with similar numbers of high rarity items like I mentioned above, but it doesn't SEEM like I'm seeing a difference in the rate of items of a specific rarity restocking. That's just going sight though since I restock semi-legit. Anyone noticed a similar trend? If option 2 held true I'd expect to see a a lot more Leaded Elemental Vials out there, but it's tough to know for sure since the actual numbers aren't so easy to get on Neo.

[Waser Lave]

Even with the Item Database it's difficult to tell anything because we don't have complete data on when every single item restocks in every shop.

[kingofaces]

True, but you only need a sample, not a full record of when every item restocks (and that's the beauty of statistics ). Say you had a time period of a month. All you'd need is the total number restocks observed, new items seen (name, rarity) at each observed restock (maybe rarity > 94 to keep it manageable). and also repeat for a few different shops. To make the comparison between option 1 and 2, you just need the number of different restocked items seen with their rarity and the number of restocks seen over your time period (in this example a month). You'd want the shops to be checked for a restock at the same rate though, and they should both have similar traffic so an item doesn't sit forever in one shop and not the other to lessen the chances of one not item being seen, but missed in the other. Ideally, neither shop should be full of items. Basically if you're going to have error, make it the same error across all shops. Here you'd just be seeing if there is a difference between restock rates of a rarity, not what the actual is.

That was one comparison. Depending on the results of that you could find out the rate a specific rarity. In that case though you'd need a relatively high refresh rate to make sure you're seeing an item before someone else buys it. I typically have a 2 second refresh rate and rarely miss seeing an item at that rate, so you should be able to get a good approximation of the actual rate a rarity restocks in each shop with 2-3 seconds. The amount of traffic in the shop wouldn't be as important. Basically you're just repeating what we did above, except this time you need that high refresh rate.

It's understandable though if there is something here that isn't quite kept track of as needed when the item database information is collected. I almost always have something I wish I kept track of after the fact when collecting data for a project. Who knows though, maybe it actually is feasible if someone wanted to put time into checking it out.

Edited by kingofaces, 13 December 2011 - 10:38 AM.

[Waser Lave]

True, but you only need a sample, not a full record of when every item restocks (and that's the beauty of statistics ). Say you had a time period of a month. All you'd need is the total number restocks observed, new items seen (name, rarity) at each observed restock (maybe rarity > 94 to keep it manageable). and also repeat for a few different shops.

My point was that we don't have a full sample for any shop because we can't know when every restock happens or all of the items which stock in each restock. We rely on individual users refreshing at the right time and we usually only have perhaps 150-200 users at any time spread out over several shops so there are limitations which make it difficult to do this kind of analysis with any significant level of confidence in the results.

[KleinBucket]

Sorry to bump a dead topic, but I'm also very interested in some more rigorousness in our restock knowledge.

 

Showing that restock rate is not just a function of rarity is pretty simple and requires very little data. We just need restock intervals of a small sample of R99s between several shops. Then ANOVA away (or similar)!

Then we could exhaustively try other possibilities:

Restock rate proportional to rarity * shop restock freq

Rate ∝ rarity (or range of rarities?) weighted by quantity

Rate ∝ quantity weighted by rarity

 

It already seems clear that rate ∝ JUST rarity is out. Almost all the morphing potions in Magic Shop are R97+, and apparently those appear at least every 5 restocks, while the R90+ in Food Shop appear once in a blue moon.

 

Also Clothes Shop definitely doesn't restock 170x faster/more than Pharmacy, so there must be some criteria besides rarity.

 

EDIT: I wonder if we can naively assume rarity is related to logscale:

R98s restock twice as often as R99s

But then R89 would be 1000x as often already...that can't be right. But you get the idea.

Edited by KleinBucket, 19 April 2014 - 02:09 PM.