Tracking Legacy growth - SCG Legacy Opens

[Koby]

Let me preface this thread by stating that I believe Legacy to be growing in popularity and attendance at large tournaments. This thread aims to build a correct model to track the growth of Legacy tournaments, specifically SCG Legacy Opens, since that is the largest driver in the last few years of growth.

We're going to begin with only 2012 tournament data, and add back years as we gather more data. Calendar years might not be the most logical categorization since there is more logical breakpoints in relation with both seasons and set releases. Let's start with just an annual model, then examine the set release periods afterwards.

I am not a statistician, so I'm not exactly sure how to transform the discrete values into a form better suited to study natural growth (logarithmic) or periodic growth. I know enough to know that I'm not confident in building the model. If you're more familiar with the subject, I'd be delighted to learn a thing or two.

Let's start with some data:

Link to Google Docs Spreadsheet

Mark Sun suggested I break the tournaments into a more regional approach. N/E/W/S seems the most appropriate, but we can fine-tune this at a later date. I've also calculated some values that will help us with the study.

Mean = 199
STDEV = 63.3

Based on linear growth, we're looking at 12.8% growth in tournaments. However, this is not exactly accurate and I realize it.

I appreciate any help in building this model. Thanks!

[Esper3k]

Are you planning on tracking data from GPs as well?

Some of those data points I think may even be lowered (I remembered there being something going on in NO when there was that really small SCG there) due to outside factors?

[Koby]

Are you planning on tracking data from GPs as well?

Some of those data points I think may even be lowered (I remembered there being something going on in NO when there was that really small SCG there) due to outside factors?

We can definitely track GP's as well. However, as they are so infrequent, and sizes so large, I don't think we'll get a good picture of what's going on.

I definitely plan on identifying outliers in the model. Attendance <100 is a dead give away, but also Opens that coincide with Invitationals can be removed since it shows a huge turnout. I plan on showing both cases, as a sanity check.

I think the first step will be to normalize the data. That will require generating Gaussian distribution of the data points. I want to also take the 80% data (lopping off both the lower 10% and highest 10%). This will even out the data from the outliers like NOLA (91) and Atlanta Invitiational (350).

[Tammit67]

A regional partitioning is probably best: there are too many factors to consider when comparing events near New York, and ones in Iowa. It would be best to break it down by city per year (although I realize thats not so hot). We could group cities deemed 'within acceptable travel distance' from each other (Someone going to SCG Baltimore, MD can often attend SCG Washington DC for instance), and measure the average of events in that area. Use a one sided hypothesis test and find the p-value. That way we can ascertain whether or not the apparent growth is due to simple variance in attendance or more players getting engrossed in legacy.

I do similar trend analysis for a living, hope this puts you on the right track.

[Megadeus]

NOLA if I remember correctly had a hurricane warning or something. Im sure it would have had a good 60-100 more people if not for that

[Koby]

We don't need to spend too much thought into why outliers are outliers. If the number is so out of the normal range it will be culled regardless. The idea is to track the long-term growth rate, and perhaps correlate it to locales.

My guess is that a 100-mi radius is a good place to start picking the locales. That's about the longest driving distance people can stomach in the morning and still make it to the event at 10 AM. There may be people who travel farther to reach the tournaments, but I don't think that's the typical norm. My hypothesis is that Legacy is growing within the local metagames that the Legacy Open taps into, rather than the grinders who will go out of their way to play in these events.

EDIT: Matt, thanks for those ideas. I'm going to incorporate them into the model once I get the statistical portion hammered out.

[Kuma]

Why not do something like create a new variable that is a 1 for the chronologically first tournament, 2 for the second, 3 for the third and so on? Then, see if that new variable is positively correlated with tournament attendance (attendance tends to increase when the new variable increases, i.e. tournament attendance is up over time.) Most statistical software will even give you a p-value to determine if the correlation is statistically significant. If you're worried about region, you could do the same thing with each region separately. The problem is, this test isn't going to be very powerful if there are lots of locations with only one tournament, which appears to be the case.

I might do this tomorrow if I feel sufficiently motivated.

[Zakman86]

NOLA if I remember correctly had a hurricane warning or something. Im sure it would have had a good 60-100 more people if not for that

The New Orleans Star City weekend was the same weekend as VoodooFest. Hotels were super difficult to get and likely prevented a lot of people from attending.

[Finn]

I know that I am out of my element here. But back in college when I wanted to track data much like this, I used a model that compared data year after year adjusted for data the previous week. This should allow for predictions of attendance as well.

Do we have access to previous years?

[Ziveeman]

Here is the data from the 2011 Open Series:

1-9: Kansas City - 193
1-16: San Jose - 151
2-6: Indianapolis - 267
2-27: Washington DC - 290
3-6: Edison, NJ - 296
3-13: Memphis, TN - 137
3-20: Dallas/Fort Worth - 157
3-27: Los Angeles - 216
4-3: Atlanta - 210
4-24: Boston - 252
5-1: Charlotte - 211
5-15: Orlando - 146
5-22: Louisville - 203
6-5: Indianapolis - 334 (Invi)
6-13: Denver - 117
7-17: Cincinnati - 228
7-24: Seattle - 184
7-31: Pittsburgh - 198
8-14: Richmond - 164
8-21: Boston - 267
9-11: Atlanta - 249
10-2: Indianapolis - 288
10-9: Nashville - 169
10-23: Baltimore - 238
10-30: Kansas City - 156
11-6: Las Vegas - 167
12-4: St. Louis - 276
12-11: Charlotte - 325 (Invi)

As a note, this is the era when the Players Club incentivized people to travel to these events (Alex Bertonici, Gerry Thompson, etc), so the numbers may not be representative of the metagame in the area. I have no idea how to calculate that into the data (or if it's relevant at all), but I figure I'd mention it.

[Ziveeman]

2010 Open Series Data - A lot of it is missing because they didn't post standings:

12-13-09: St. Louis - 128
1-3: Los Angeles - 142
1-10: Dallas/Fort Worth - 117
2-28: Richmond - ???
3-12: Indianapolis - ??? Coverage mentions nearly 300
3-26: Orlando - ??? Coverage mentions 120
4-30: Altanta - 200
6-4: Philadelphia - ???
6-11: Seattle - ???
6-25: St. Louis - ???
8-20: Denver - ???
8-27: Minneapolis - ~170
9-17: Baltimore - ???
10-15: Nashville - ~200
10-29: Charlotte - ???
11-5: Boston - ???
12-5: Richmond - 179 (Invi)

[nedleeds]

The New Orleans Star City weekend was the same weekend as VoodooFest. Hotels were super difficult to get and likely prevented a lot of people from attending.

The upcoming Atlanta one is on Super Bowl Sunday. <sigh>

[Zakman86]

The upcoming Atlanta one is on Super Bowl Sunday. <sigh>

It's not like the Falcons didn't choke away the Championship Game or anything...

[.dk]

2010 Open Series Data - A lot of it is missing because they didn't post standings:

12-13-09: St. Louis - 128
1-3: Los Angeles - 142
1-10: Dallas/Fort Worth - 117
2-28: Richmond - ???
3-12: Indianapolis - ??? Coverage mentions nearly 300
3-26: Orlando - ??? Coverage mentions 120
4-30: Altanta - 200
6-4: Philadelphia - ???
6-11: Seattle - ???
6-25: St. Louis - ???
8-20: Denver - ???
8-27: Minneapolis - ~170
9-17: Baltimore - ???
10-15: Nashville - ~200
10-29: Charlotte - ???
11-5: Boston - ???
12-5: Richmond - 179 (Invi)

we had somewhere around 130 in Denver in 2010.

2011 was tiny as there was some other huge tournament going on that weekend (Sunday specifically), but I can't exactly remember what.

[Ziveeman]

Awesome, that reminds me. Maybe anybody who attended these events could check their Planeswalker Points page to get the exact number of players at the event.

[Zakman86]

we had somewhere around 130 in Denver in 2010.

2011 was tiny as there was some other huge tournament going on that weekend (Sunday specifically), but I can't exactly remember what.

PT Nagoya.

[alphastryk]

It's not like the Falcons didn't choke away the Championship Game or anything...

*sigh*

Definitely will be interesting to see what trends come out of this data. We might be able to find out some of the missing attendance numbers from the Planeswalker Points site as long as somebody played in a given open (do they go back that far with full data? I know the old site did)

[dillonkbase]

Doing a linear regression is relativly simple using a few different types of software, the hard part, if you call it that is organizing the data. If someone puts the event dates into one column of excel and the totals in another I can do a simple regression analysis.

I would probably group the data by months to get multiple samples each month, but the statisitcal power of this will still be very low.

[ESG]

Awesome, that reminds me. Maybe anybody who attended these events could check their Planeswalker Points page to get the exact number of players at the event.

Good idea. SCG Seattle drew 191 players in 2010.

[Koby]

I put the data on Google Docs for easy access.