This week I’m going to take some more time to update you on the progress of the Demand Matrix and expand on what we’ve gone over in the past.
I’ve re-written and expanded the source data that feeds into the matrix and laid the ground work for expanding it to be able to cover Extended, Legacy and Vintage. I’m a ways away from having all the code together to calculate the demand trends in Legacy and Vintage, but Extended should be added before too long. I’m now using three primary time frames for referencing and establishing demand of the cards in standard. Since there is an easily accessible collection of data summing up the last month, week, and 2-15 days as groups, I’m crunching those and comparing the results to establish some bite sized results.
I’m very happy to reveal that the hypotheses seem to be bearing out. For the first time since I’ve been recording and watching these figures dance around a new leader has been crowned as the most “in-demand” card. As you can see above, Primeval Titan has upset Jace, the Mind Scupltor for the most purchased card (by overall cumulative sales volume). It’s also interesting to note that Baneslayer Angel has appeared to be making a comeback and I would guess this is because of the recently falling price of said Angel. All my Magic brethren out there who are finally completing play sets can thank the more interesting cardboard in the Titan cycle as well as Fauna Shaman for the bargain. Also new to the Demand Matrix this week are the columns for “DI Change” and “Cards of Note.”
The values in DI Change (Demand Index Change) represent the difference between the most stable of the calculations, those boiling down a month’s worth of data, and the most volatile, figures only from the last full day. There are a number of pretty interesting Index movements in addition to the cards listed above. Obstinate Baloth has jumped up 14 spots, Verdant Catacombs up 7 spots while Arid Mesa is down 7, and quite surprisingly to me, little baby Jace Beleren is up 7 spots in rank. I’ll leave the reasons for this up to the strategists, but the individual reasons could be anything from coincidental purchases to every player in Texas deciding on a U/W Control build with original recipe Jace’s for extra crispy Jace removal.
Also new on the chart with week is the Cards of Note column along with its defining value. I’ve rather arbitrarily chosen to indicate the top 6 cards according to the largest numerical jump in Demand Index values from the weekly results versus the daily results. I would use the monthly figures instead of the weekly figures but they tend to create some false positives on newer cards since they haven’t been around for a full month. This week the cards leaping up and demanding some attention are Baneslayer Angel, Vengevine, Gideon Jura, Grave Titan, Fauna Shaman, and Obstinate Baloth in that order. Some of you might wonder why Vengevine and Gideon Jura are both down in rank for the week as well as some of the cards that are showing the most overall individual increases in the Index. This is because they’ve gone down in relation to some other cards rather meteoric rises, but still have notable increases compared to their own records and we’ll no doubt see more apparent contradictions. Once I am able to solely rely on my own data (I’m learning Python!) I’ll be able to specifically present some cards that are falling in demand as well, hopefully as a precursor to a price drop, but I’m sure that you can agree with me that this is a place where we don’t want false positives.
I’m also continuing to build in some more calculations to give us a bit of a heads up about what cards I should possibly be paying special attention to outside of the most valuable cards. There’s a lot more to come with the Demand Matrix.
When we last looked at the Demand Matrix I mentioned a few other aspects we still could still explore, one of which is a practical implementation for this information. I envision that once I have the calculations, formation of the data and organization of the data fully automated, with all the details disclosed for independent auditing, we’ll have a new system that we can really rely on. In my grandest delusions I can see where an online store who is particularly fond of us or the system might tack on the Index number to some of the more valuable cards. For example, perhaps instead of charging $37.99 for a Baneslayer Angel and $84.99 for a Jace, the Mind Sculptor, CoolStuffInc.com might instead charge $37.83 and $84.90 to indicate those cards Demand Indexes of 83 and 90. As I fall farther into my delusions I see how some stores might even donate their sales data to the cause in order to make the Indexes more accurate. I could be dreaming a bit too big on that one, but I really hope that it becomes a tool that people other than myself use. I also really hope that you can see the same potential in it that I do.
And lastly for this week, in the spirit of giving you information you can use in real life I leave you with the Set EV spreadsheet and the following request. If there is something you’d like explained or other data you’d like to see, please sound off in the comments or drop me an email.
Only one question:
when did packs _actually_ get valued at 3.95/ea?
Anyone doing EV analysis should be using about half that number…
I’ve found your analysis very interesting, but it brings memories from my economics class and raises some questions that can be very important in the process of creating a model.
1st The demand and the offer are closely related, How much weight has the offer in the final price? and if the offer is important: how much do we know about the offer for a particular card ? how many copies have been print? how many copies are on sale? wich other factors than rarity can modify the offer for a given card?
2nd Since the price of a card depends (among other things) in how many decks it’s used, can we think about buying cards as buying stocks of certain decks? but in the other side cards are the raw material for decks so can we think about them as commodities?
(i’m a non native speaker so excuse me if something doesn’t makes sense)
@xteban
I’ve never had an economics education, so I don’t know exactly what you mean by “offer,” but if you mean the value that it’s priced at vs what it’s purchased at, I’m not sure that is a relevant value given the malleability of the acquisition prices. I’m not currently including any online store’s retail pricing because I don’t have any information about how many of the cards are actually sold at those prices. Based on the information available on some web sites, I could make some guesses, but they are bound to be inaccurate. The data is all from completed sales only, and since some of the variables you mention aren’t generally available I’m using the quantity sold and the the average price over various points in time to establish the card with the highest cumulative value and compare all the other cards by percentage of that maximum value. I considered altering the weight of certain values, but decided that if I was to do so it would need to be done formulaically to eliminate arbitrary results. There are a few other data points I’m considering adding to the mix, but I haven’t yet started evaluating how to get reliable data and how to add them. If I can make the results more accurate, I will.
for your 2nd question… I could do that if I had enough information about how many of each card are used in some number of decks at some level of tournament or above. However, at this point, even with the cumulative data available, the results would be little more then anecdotal, and I don’t think they would give any clear signals about the actual demand of specific cards. I like the idea though, so I’ll see what I might be able to get out of it.
Chris
Can you explain how the value of packs is calculated? It’s unfair to assume my Haunting Echoes is worth the 50 cents it retails for, as moving a card like that is exceptionally difficult.
@Mike, @ redsai.
I went over how the EV was calculated in my last article, but to recap, I work it up in 2 different ways within my spread sheets. One way kicks out anything worth less than $0.75 and totals the rest based on distribution probability. The other way (which is used here) doesn’t toss out the smaller card values. The reason I’m using the later method is a calculation issue since it’s more difficult to filter out the lower prices with the way the data is summed. If this is a large issue for you guys I can work on changing the way it’s calculated. it really shouldn’t matter though because these numbers should be used for comparison reasons. I could assign arbitrary values to each one and as long as they were proportionally the same, they would still be an indicator of which packs have the higher dollar cards on average. Also, the average cumulative value of a pack as calculated here is pulled from real sales data. No where does it assume any card values. If common card X has never been bought as a single on eBay then it gets a $0 price in these figures. That $0.50 Haunting Echoes you mentioned might be hard to shift but but it has been sold at the calculated price in the last 7 days. Good guess on that by the way, in the last week Haunting Echos has sold 68 copies at an average of $0.049, and 265 copies at $0.55 for the last 30 days.
As for using $3.95 for the price of a pack, that is the suggested retail. I know that they are $3.95 at Walmart, $4.99 at Best Buy, $3.95 at Card Store X, $3.50 at Card Store Y, $2.36 when purchased by the box on the world wide web, and $1.00 when your brother Phil “hooks you up” but since I can’t predict what kind of deal every person is getting, the Manufacturer’s Suggested Retail seemed like a good baseline. I wouldn’t use any other value here just like I won’t lower the M11 EV because someone traded me a Baneslayer Angel at $10 last night. If you would like to sub in your own values there, feel free, but it won’t alter the comparative EV.
Chris
Part of getting these online stores to donate the data may be convincing them that their data won’t be used to make recommendations to their competitors. That said, if you come up with a solid enough engine, you might be able to sign stores up on a subscription (assuming that they don’t have an internal system already). Prove that your method is less expensive and / or more economically viable than their current method, and it’s probably a no brainer
I don’t see any real contact information outside of this spot, but I can’t see all of the images/graphs you posted here. I’m using IE. I can see both the first and last of the google doc sheets but nothing in between. I had this problem with the last article too.
@Paul
Thank you for the information. I use Chrome and hadn’t noticed an issue, but we have our web guy looking into it to get it sorted. Also, I’ll make sure to get the contact info in there.