The Greatest Angles Ever - #3: Home in the DEL

Those who bet the NHL seriously at all know that the lines are pretty sharp with any large moves mostly due to injury or goalie announcements. Using the same data provided by the league, that everyone else uses and there are a thousand free packages one can use to scrape, you are just not going to get a very big edge. But if you combine qualitative and quantitative analysis of the teams, or have some new or better data, perhaps you can still beat the NHL for a nice winrate.

For example, the shot location data in the NHL dataset is horrible, and everyone knows it is horrible. You can adjust it but it will still be horrible. And even if it was good, it doesn't include the most important variable needed to slope shooting percentages well: where the defense and goalie were on the ice when shots were taken. This makes it hard to determine which goalies are good, or how good defenses are.

What if you created a dataset that was good? The NHL and NFL are doing this now with chips, but it looks the public probably won't get their hands on it. The MLB did this and the public did get it and it does have some betting value. The NBA did this and the public kind of got the data but not really. Regardless, the video is out there to make this dataset as it is; someone just has to do it. The problem from a betting standpoint is that one season is not enough. One would need several seasons, to see if your model based on the new improved dataset is actually better than the old one for the one thing that matters in betting, out-of-sample performance. This means you will have to pay a lot of people to watch games, which would require quality control, since everyone would have their own opinion on where the puck and players are.

Another option would be to develop a machine learning tool to watch a game and simulate a human. But most of the people good enough at machine learning to write something like that are busy doing really horrible stuff to ruin humanity, and you'd have to pay them a lot to pry them away from god's work. Unfortunately, there is the very real possibility that one spends hundreds of thousands or millions of dollars on building said dataset and it turns out adjusted Corsi is still better than whatever adjusted goals metric you made. I'm not sure it would be legal to sell your dataset either. But if I had to guess, you would get something out of it.

Anyway, lines in the second biggest league, the KHL, are also made by a sharp oddsmaker. In fact, of all the things I have ever bet, the KHL was the biggest loser by far ROI-wise. It was incredible how quickly I lost, especially since I remember being very close to the market on almost every game (usually if your model is horrible you will be way off the market) and was only laying -105. Most likely, I bet the dumb side on every fixed game, and it's Russia, so if you told me the KHL was 100% clean I would never believe you, or game with a major injury and saw no value on anything else. I never did very well in the VHL (Russian Division 2) or MHL (Russian Youth League) either although I at least managed to break even in those leagues.

Fortunately, in the early part of the 2010s, multiple sportsbooks began expanding their offerings to include other European hockey leagues including the SHL, the Allsvenskan (Sweden Division 2), the Finland Liiga, and the DEL and DEL2 in Germany. In all these leagues, except Finland where I had something more advanced because there was more data, I used simple shot + goal differential models adjusted for home ice and occasionally basic roster measures and cleaned up. It was a joke how bad these lines were and how many fundamental aspects were mispriced.

Other than the bad lines, the best part of these leagues was that betting didn't include overtime, so you didn't get screwed by coin flips like in the NHL. Typically, the play was to always bet the good shot differential / bad goal differential teams, especially if they had poor recent form, which was always way overvalued. There was also a small bias that underdogs were better in general, but it was not rare to play favorites especially at home. Games went to OT way more often for a given difference in team strengths than the lines reflected, meaning it was always right to take the +0.5 or +1.5 if one liked the underdog, but lay big prices on the "0" money line (meaning you pushed if the game went to OT) if one liked the favorite. Finally, with totals, in some leagues I had basically no model at all. The play was under all high totals and over all low totals, everyone converges to the mean, and this won at a good rate.

This was great, but one league was even better: the German DEL, and the angle was to bet the home team almost every single game. All European hockey leagues have ridiculous home advantages compared to the NHL. In the modern NHL the home team wins about 54% of games excluding OT, but of that excess 4%, about 1% is due to the road team facing more back-to-backs, so it is really only 53%. European leagues are above 56% and if you have ever watched a European hockey game you would understand why. Rather than the corporate experience typical in North America one sees packed loud arenas with loud soccer-style chants the whole game, even for bad teams in the second division, and in some arenas they even allow smoke and flares. The away team basically never gets a power play unless they are down two goals. And the DEL was the most ridiculous of all. In my database of games from 2008 to 2015 the home team won 61.2% of games not drawn, with an advantage in regulation of .612 goals per game (in a victory for symmetry, the number really was .612 for both). The lines instead reflected a home advantage similar to the rest of Europe.

While most of these angles came crashing down due to books adjusting to bettors betting the same stuff over and over, in this angle it was not the books but the league which adjusted. DEL officials eventually figured out that games were rigged to favor home teams too much and adjusted, but over-adjusted to the point that over a two month period, away teams had a few dozen more power plays, and had won more points, than home teams. Around that time I found a PDF from somewhere in German hockey saying they were aware of the disparity between the DEL and other leagues, and although the PDF did not say they were going to do anything about it, apparently they did.

Fortunately, lines had already sort of caught up to where it was possible to bet the road team from time to time, so I did not lose that much. I quit betting the DEL at this point as it was impossible to bet in an educated manner given the uncertainty in such a key part of handicapping, plus they had introduced a new website with various unusual anti-scraper protections and I had already been forced to input box scores by hand.