Mark Barbour (@18sktrs, patreon.com/18skaters)
Hello readers of Apples and Ginos! My name is Mark and I’m excited to say that I’ve joined the A&G content team. I’ll provide regular content when the NHL season gets going, but I’m introducing myself now with this (longer) piece about best ball fantasy hockey drafts.
What Is Best Ball?
If you’re not familiar with best ball, it’s a format where you fill your roster before the season starts and then you live or die with it for the full season. No further action is required – your highest scoring players are automatically selected for each scoring period and their points count toward your place in the standings.
Nate here at Apples and Ginos has been running best ball drafts on Fantrax this summer and that’s when I became more interested in them. The entry fee is only $5 (USD) so maybe you’ll want to take your shot after reading this article. Anyone can enter through the Apples and Ginos Discord.
How Do You Win At Best Ball?
I had entered a couple of best ball fantasy hockey drafts in the past but I wasn’t confident that I had figured out the subtle differences for this format of fantasy hockey. After finishing a few of the Apples and Ginos drafts I decided that I wanted to make better (data-driven) decisions when selecting my skaters, so I created a model to help me do that. My model has given me a couple of insights into how to win at best ball and I’ll share those insights here. They won’t guarantee wins, but they should help your team be more competitive.
My Best Ball Model
First, I’ll explain a few things about my best ball model so that you have some context for what comes next.
The model can do a few interesting things: 1) it can compute my team’s total expected fantasy points for the full NHL season; 2) it can tell me the percentage of potential points that my team is expected to “capture” (i.e., points that will not be lost to skaters sitting on the bench); and 3) it can recommend skaters during the draft if I tell it who is currently on my team and who is still available. How does my model recommend skaters? It adds each available skater to my team (one at a time) and then computes the increased expected point total for my team. In other words, it shows the marginal increase in total expected points for each of the available skaters. This means I can see which skaters would add the most expected points overall, as well as which skaters would add the most expected points relative to other skaters who play at the same position.
Here are some caveats about my model: 1) it assumes that all my skaters will play in every game, which will not happen in reality; 2) it assumes that each skater will score his average number of expected fantasy points in each game, and this also will not happen in reality (more on how to mitigate this issue below); and 3) it ignores goalies. Another important assumption is that the NHL schedule will not get scrambled again this season. If that happens … well … good luck.
One last thing: it is important to keep in mind that the discussion that follows is relevant to best ball formats that are similar to the Apples and Ginos leagues, with daily scoring periods and skaters eligible to fill multiple roster spots at each position.
Insight #1: Draft To Maximize Games Played
This might seem obvious, but it’s probably more important than you think. You want your team to consist of skaters who have uncorrelated schedules (essentially, setting yourself up for good daily “streaming” during the season). This means that you should try to avoid selecting skaters from the same team, especially if they play at the same position. Let’s run through an example to illustrate this point.
Assume that I’ve drafted a team with the following skaters playing at RW: 1) William Nylander (TOR); 2) Pavel Buchnevich (STL); 3) Alex Tuch (BUF); and 4) Brendan Gallagher (MTL). Based on the other skaters I have on this team (not listed here) my model tells me that the total expected points for the season is 5899, which represents a capture rate of 77.9% (recall that this is the percentage of potential points that are not lost to skaters sitting on the bench). What happens if I swap Brendan Gallagher for Vladimir Tarasenko (STL)? This seems like a nice upgrade, right? Afterall, Gallagher’s expected fantasy points per game is 4.09 while Tarasenko’s is 5.36 (based on my fantasy point projections, which I will use here to illustrate my point).
It sounds good, but the problem with adding Tarasenko is that I already have Buchnevich at RW and he’s a teammate of Tarasenko. While adding Tarasenko would slightly increase the total expected points for the season to 5914 (from 5899), this move would strand a lot of points on the bench as the capture rate would dip to 77.0% (from 77.9%).
Now assume that I instead replace Gallagher with Ilya Mikheyev (VAN). Mikheyev’s expected fantasy points per game is 4.07, which is slightly less than Gallagher and much less than Tarasenko. It seems like a downgrade, but Mikheyev’s advantage is that he has an uncorrelated schedule. With Mikheyev at RW my team’s total expected points would rise to 5911, which is not substantially different than the 5914 that was the result when I added Tarasenko. This happened even though Mikheyev is expected to score fewer fantasy points per game. The key here is that the capture rate increased slightly to 78.0% when I added Mikheyev, rather than dropping when I stacked skaters who play on the same team.
The effect of stacking skaters from the same team is less pronounced when the skaters do not play at the same position, but the same problem exists in that situation. I won’t run through another example, but the takeaway is the same: a skater with fewer expected fantasy points can be more valuable if his schedule is not correlated to the rest of your team. This is especially important to keep in mind when adding skaters late in the draft. Those late skaters are like your streamers, and you’ll get maximum benefit from them if they play on nights when your best skaters are out of action.
Assuming you don’t have a model to tell you which skaters are the best fit for your team, you can make an informed guess by using a matrix that shows how many times two teams play on the same night (this is copied from my fantasy point projections which are available to the public at patreon.com/18skaters):
Note that using a matrix like the one above is better than simply relying on how many games a skater will play on “off nights” (an “off night” could be defined in different ways, but a common definition is any day on which no more than half of the teams in the NHL are playing). For example, Anaheim has the most off nights of any team in the NHL, but using the above matrix shows that simply relying on off nights can be a trap if you assume that you can stack skaters from Anaheim and New Jersey.
Insight #2: The Number Of Roster Spots Matters
I discovered something surprising the first time I used my model to recommend a skater. I had just filled every available roster spot and was about to pick my first “extra” skater. There were still plenty of decent forwards on the board, and they were expected to earn more fantasy points than any available defenseman. I assumed my model would tell me to pick a forward, but I included the top three available defensemen anyway. The model recommended all three defensemen over every forward. In fact, there was only one forward who even came close to matching the increase in total expected points that would result from picking any of the three defensemen. Why did this happen?
Admittedly I have not tried to “prove” this, but I have to assume that the advantage the defensemen had was the number of roster spots available to them.
In the Apples and Ginos best ball leagues there are two dedicated roster spots for each forward position, plus two “flex” spots that can be filled by any skater. So, a LW could fill one of four roster spots on any given day. In contrast, there are four dedicated roster spots for defensemen, plus those two flex spots that can be filled by any skater. So, a D could fill one of six roster spots on any given day (two more than the LW). I had been working on the assumption that higher-scoring forwards would typically occupy those flex spots and therefore there likely wasn’t a big advantage for the defensemen. After running the model, I now expect that the defensemen land in those flex spots more often than I would have guessed.
The takeaway here is that the roster restrictions in your best ball league can affect the value of skaters in ways that you might not expect. A skater who can potentially occupy more roster spots could be more valuable than a skater who can occupy fewer roster spots, even if he is expected to collect fewer fantasy points over the course of the season. Emphasizing skaters who have the potential to occupy more roster spots is probably a good idea when you’re drafting your team.
There is a synergy between this insight and the first one described above – if you select a skater with an uncorrelated schedule who can potentially occupy more roster spots you should capture a higher percentage of that skater’s total points (rather than leaving those points stranded on the bench).
Insight #3: Turning “Off Nights” Into “Big Nights”
As I noted near the top of this article, my model assumes that each skater will score his average number of expected fantasy points in each game. This will not happen in reality – skaters will have big nights while other nights will be … errr … small nights. It is impossible to know in advance when the big nights will happen, but it would be ideal if they were to happen on the off nights in the NHL schedule (leaving the skater’s “small nights” for busy days in the NHL schedule when other skaters can contribute). Can we do anything to make this more likely to happen?
We can select skaters who play weak opponents on off nights. Of course, there is no guarantee that it will work every time but I expect that doing this would give another small edge to your team. Here’s some data to help you get that edge:
This shows how many games each team has against a weak opponent while playing on an off night. I determined the “weak” opponents using data from last season. I simply added each team’s actual goals against and expected goals against and then selected the ten worst teams by this metric (data from Natural Stat Trick). Is this a perfect predictor for this season? Probably not, but you can see that it will likely be advantageous to pick your late skaters from Edmonton rather than St. Louis.
Goalies
I’m not going to call this an insight, but I do have a thought about selecting goalies.
Let’s assume you’re picking a goalie late in the draft and the options are bad. You’re looking at goalies who are not expected to play many games, but you still want to grab one more goalie. Among the options is a goalie from the same team as one of the goalies you drafted earlier. Should you take the tandem?
It could make sense to get the tandem in this situation for two reasons. First, you know this goalie’s schedule will be uncorrelated to the schedule of one of your existing goalies – they both play on the same team and therefore will not both be starting a game on the same day. This should give the tandem goalie a pretty big advantage over the other options, assuming they are all expected to play a similar number of games during the course of the season. The second reason for grabbing the tandem is the one that you often hear: “insurance”. If your main goalie is injured then your backup gets elevated to being a starter. I wouldn’t allow the “insurance” reason to affect my early drafting, but I think it does provide an additional reason to get the tandem when you’re looking at nothing but bad options late in the draft.
The End Of The Article
I haven’t covered everything that’s relevant to drafting your best ball team. Things like positional scarcity and average draft position are relevant in best ball, just like any other fantasy hockey draft. So of course you’ll want to keep those things in mind while drafting.
That’s it. I hope you enjoyed this article and maybe I’ll see you in the next best ball draft at Apples and Ginos. Cheers!
Apples and Ginos 