I was recently sent a spreadsheet containing information on every NLL game ever played, from 1987 to 2011. For each game, it contains the date/time, home and away teams, final score, attendance, and what type of game (i.e. regular season, division finals, championship, All-Star game, etc.). For a numbers geek like me, this is heaven. I immediately imported the data into a SQL Anywhere database and started to write stored procedures and web services to aggregate and analyze the data.

Thanks very much to NLL Chatter reader Dennis for sending this to me, and props for all the work it took putting it together!

Over the next little while (probably more in the off season), I will be posting articles summarizing and analyzing certain aspects of this data. For example: which team has the best record in Friday night games? What about Saturday away games when they also had a Friday home game the night before? What about Saturday afternoons on artificial turf against a lefty starter when the roof is closed? Oh wait, sorry, wrong sport.

If you have particular requests or interests, please leave a comment here or send me an email and I’ll see what I can do.

Just to whet your appetite for the types of information I can get from this data, here are some random pieces. Note again that these numbers do not include the current (2012) season.

Now, if you have a spreadsheet containing every goal from every game, who scored it, who got assists, what quarter it was scored in, and whether it was a power play; shorthanded; empty-net; or penalty shot, and you feel like sending it to me, that would be perfect.

P.S. Yes, I stole the title of this article from John Hodgman’s second book.

I looked over the 2012 NLL attendance stats yesterday and decided that the Bandits were the biggest draw in the league. I explained why and described the method I used to come up with that conclusion. I also said that my conclusion was wrong. Well, kind of. As it turns out, my conclusion was right but (a) the method I used was slightly flawed, and (b) there are really too few games to make a definite conclusion. Let me explain. No, there is too much. Let me sum up.

I will describe the flaw in my reasoning by way of an example using made up numbers. Let’s use the same example as before – Toronto playing in Colorado and Buffalo, and Rochester playing in Minnesota and Washington, and we want to find out which team (of Toronto and Rochester) had the biggest impact on attendance. Let’s say for simplicity that each game’s attendance is 500 higher than the average attendance for that arena. Using the previously described method, the average draw for both Toronto and Rochester would be 500, and so we’d conclude that they have roughly the same impact on attendance. This is false.

The reason that it’s false is this: bringing in an extra 500 people to a game in Colorado is not as big a deal as bringing in an extra 500 people to a game in Washington. Colorado’s average home attendance is 14626, so an extra 500 people would increase attendance by 3.42%. But Washington’s average is only 4017, so an extra 500 people means a 12.45% increase. Rochester’s presence in Washington had four times the effect on attendance than Toronto’s presence in Colorado. Once again, we need to factor out the average attendance by using the draw percentage rather than the draw. In this case, the draw percentage for Toronto was 3.42% in Colorado and 3.26% in Buffalo while the draw percentage for Rochester was 12.45% in Washington and 6.38% in Minnesota. Toronto’s average draw percentage was 3.35% while Rochester’s was 9.42%. While our previous calculation showed no difference in the draw numbers, in actual fact Rochester had about three times the impact on attendance than Toronto did.

Let’s add the draw percentage to our previous table. Note also that the attendance for the Minnesota @ Washington game from Feb. 24 was not included in the game sheet until today. This affects Washington’s home average attendance, and therefore the draw numbers for teams that have played there (Calgary, Edmonton, and Minnesota). I’ve updated the numbers from the previous article.

Despite the flaw in my logic, the ordering remains the same, and therefore the conclusion remains the same: the Bandits are the winners. Or so it looks.

As I briefly mentioned before, the biggest problem with this whole scheme is sample size. Buffalo has the highest draw percentage, but this is only based on three games. One of those games was in Rochester, which is close enough to Buffalo that many of the fans in the Rochester arena actually drove from Buffalo. Indeed, the draw percentage for that game alone is 38.3%, while the draw percentages for the other Buffalo away games were 5.0 and –2.6 in Philadelphia. That one game skewed Buffalo’s draw percentage to the top of the heap, while Colorado has positive draw percentages (10.0%, 15.0%, 15.2%, and 3.4%) in all four of their away games, so you could also argue that Colorado is really the biggest draw.

If I were to take the attendance numbers for the past five years and calculate the draw percentage, then perhaps the numbers would be meaningful but half a season’s worth of data is just not enough.

I have a database containing information on every game (including every goal) in the NLL this year. I’ve been looking at goal totals, assist totals, and weird combinations like “which quarter has seen the most power play goals scored?” (Answer: the 4th overall, though Minnesota and Philly like the 2nd better). I then started to look at the attendance numbers to figure out which were the teams that fans in other cities wanted to see.

Background (you can skip this section if you’re not interested in where this database came from):

As part of my work writing the Money Ballers columns on ILIndoor.com, I wrote up a program (a python script, if you care) to download and read the game sheets for each game from NLL.com and calculate the Money Ballers points for each player. This was a lot faster and easier than the error-prone manual way I did it for the first week or two. I then realized that I could modify my program slightly to record a bunch of information about each game in a database. I’ve been working for Sybase, writing software for a SQL database engine, for almost fifteen years, so I kind of know what I’m doing there.

I changed my script to store information on each game (home/away teams, winner/loser, final score, attendance) and each goal (scorer, first and second assist, what quarter, what time within the quarter, and whether it was power play, shorthanded, game-tying, go-ahead, game-winning, empty net, or penalty shot). I will hopefully be adding other stats like shots on goal, loose balls, penalties, etc. in the future. This is all stored in a Sybase SQL Anywhere database, and I have stored procedures written in SQL to serve up web pages displaying and aggregating these stats in various ways. And no, these web pages are not available on the internet, only on my laptop. Sorry.

My page on attendance originally contained the following numbers for each team:

• Total attendance (Colorado is the only team over 100,000 this season)
• Average attendance (Colorado leading, 12,614)
• Home average (Buffalo leading, 15,318)
• # home games
• Away average (Colorado leading, 10,602)
• # away games

The away average seemed like the best way to figure out who the biggest draw was, but I quickly realized that this number was heavily skewed by the cities that a particular team had visited. For example, say you wanted to decide which would increase attendance more for any given team: hosting Toronto or hosting Rochester. If each has played two away games but Toronto has played in Colorado and Buffalo, while Rochester has played in Washington and Minnesota, then you can’t compare the averages. To compare them would be meaningless because Colorado and Buffalo get far higher attendance numbers than Washington or Minnesota regardless of who’s visiting. This means that the fact that Toronto’s away average is in the 15,000 area while Rochester’s is maybe 6,000 doesn’t tell us anything. We need a way to take away the differences in average attendance.

To to this, I took the attendance for a particular game and subtracted the average home attendance for the home team. For example, Calgary’s average home attendance so far this season is 8,122. When Colorado played in Calgary, the attendance for that game was 9,341.This means that all other things being equal, Colorado increased the attendance by 1,219. I call this difference the draw, so Colorado’s draw for that game was 1,219. If the attendance for a particular game is less than the average for that arena, then the draw for that game is negative. Doing this for all of Colorado’s away games and taking the average gives you the average draw. For Colorado this number is 920. This means that regardless of what arena you are talking about, having Colorado in town increases the attendance by about 920 people.

This scheme factors out the huge differences in attendance averages at the various arenas, making them more easily comparable. But they’re still not going to be 100% accurate, for a number of reasons. First off, we’re talking about a very small sample size – only a handful of games per team. Secondly, attendance can be affected by many other factors, including which night of the week the game is, the time of the game, the weather in the city at the time, how the team is doing (if the team is losing, some people may not bother going to the game regardless of who the visitor is), how the visiting team is doing (more people might have come out to see Washington last year than this year) and even how other sports teams in the city are doing (a Mammoth game in Denver on Super Bowl weekend will not draw the same if the Broncos are playing than it would if they are not).

There is no way to adjust the numbers to account for all of these factors, but here are the numbers for the draw for each team.

We can see that the Buffalo Bandits are the biggest draw in the league, bumping attendance by an average of 961 people per game. Colorado is next at 920. The Roughnecks and the Rock don’t make much of a difference, while having the Swarm in town reduces attendance by 935 per game. Again, because of the small sample sizes it’s hard to draw meaningful conclusions, so perhaps we’ll wait until the end of the season and see what the numbers look like then. I plan on expanding my database to include games and stats from previous seasons, though it looks like I won’t be able to get the detailed goal information since the game sheets don’t contain them. The attendance is there, so I may be able to get data for the past 2-3 (or even more!) seasons.

This article is the first in a short series about attendance. Next time: why my conclusion above is wrong and how to fix it.

Who’s the most dominant player in the NLL? Well, that’s a pretty vague question – what do we mean by “dominant”? You can’t really argue with fact that Wayne Gretzky, Bobby Orr, Tiger Woods, Roger Federer, in their primes, were dominant. Not only was it a given that when you looked at whatever statistics they excelled in that they would be near the top, but you could assume that unless something really unusual happened, that they would be at the top. When I was in hockey pools during high school, we decided that nobody was allowed to choose Wayne Gretzky because whoever got him had an unfair advantage. That’s dominance. But it doesn’t always mean the absolute best player; Dennis Rodman was never the best player in the NBA, but he dominated the rebound category for years.

This is not the dictionary definition, but for purposes of this article, I’m going to describe a player as being dominant if they are the best at something, and significantly better than second place. A player could be dominant in one aspect of the game (face-offs, for example) but not as much in another aspect (goal-scoring). But even if it’s a difficult question to conclusively answer, it’s fun to think about. Let’s look at a few candidates from the past and present.

Gary Gait won six MVP awards, including five in a row, and led the league in scoring seven times. You could certainly make an argument for him as the best offensive player of all time, but as soon as you mention Gary, you have to mention his brother Paul as well as John Tavares. As good as Gary was, he wasn’t significantly better than Paul or JT so “dominant”, using my definition above, doesn’t really apply. Perhaps you could put all three together and say that they were collectively dominant. After all, one of these three won the MVP award every year from 1994 until 2003, and one or more of them led the league in points from 1991 until 2004.

Dallas Eliuk and Bob Watson are the top two goaltenders in NLL history, and few people would argue with that. The argument over which of those two was the best will likely never end. But that fact alone likely makes it impossible to say that either was dominant. They don’t really fall into the same category as the Gaits and Tavares, in that both were always among the best in the league in any given year, but it wasn’t a foregone conclusion that they would be #1 and #2 in some order every year.

Brodie Merrill is one of the top loose balls guys in the league now, having led the league in 2007 and again last year. He’s also won the Transition Player of the Year award twice, and Transition Player of the week nine times in five years. Those are pretty impressive numbers, and you could make an argument for him being the best transition player in the league. Jeff Shattler won the award last year, but that was a little weird since Shattler played offense far more than transition. Is Merrill really that much better than Shattler, Paul Rabil, or Geoff Snider? No, but when it comes down to Transition Player of the Year voting, you can safely assume that Merrill will be right up there. I may not put him in the “dominant” category given my definition above, but he’s pretty close.

Geoff Snider probably comes as close as any active player because he’s so good in a couple of different categories. He’s led the league in loose balls in three of the past four years, he set records with the highest face-off percentages in league history (75% in 2007 and 73.8% in 2008), and he’s one of only two players to ever reach 100 penalty minutes in a season (Rory Smith is the other). If you’ve ever seen this guy in the face-off circle, he’s absolutely unbelievable. Sure he had the second-highest percentage in the league last year, but his brother Bob only beat him by about one percentage point, while he beat third place by almost nine.

So who’s the most dominant player in any category in NLL history? Even given the vague criteria above, there is really only one choice.

Jim Veltman led the league in loose balls for the first fourteen years of his career, from 1992 to 2006 (minus 1997, when he didn’t play). He has been retired for three full seasons now, and is still by far the all-time leader in that statistic. Nobody else is even close. Veltman scooped 2417 loose balls over 194 games, an average of 12.5 per game. In 25 playoff games, he averaged 13.2 per game. The next closest on the all-time list is John Tavares, with an even 2000 in 58 more games. Tavares’ average is only 7.9 per game, so he’d have to play 53 more games, more than three seasons, at that pace to match Veltman. This means that if he continues to play, Tavares will set the record during the fifth game of the 2015 season at the age of 46.

Both Snider and Merrill have higher loose-ball-per-game averages than Veltman, Merrill with 12.8 and Snider with an amazing 14.1. But they will have to play for six more full seasons before they reach Veltman’s record – and that’s assuming they keep up their current pace and play in all 16 games every year. Interestingly, at their current paces, they’d reach the record at almost the same time – Merrill in 94 more games, Snider in 97. If you’re keeping track, that’s sometime in the 2018 season.