Best Puck Manifesto - 2025 Updates

Last summer, I spent the vast majority of my waking hours that I hadn’t already sold to my employer learning to code. By which I mean, discovering new ways to ask ChatGPT to fix both of our mistakes at all hours of the night. I did this to deep dive the data Underdog Fantasy released for the Best Puck Classic, the biggest season-long NHL tournament on the planet, then shared it with dozens of my closest friends on the internet. You can find the output of my efforts here (vol 1), here (vol 2) and here (vol 3). I might be biased, but I enjoyed recently going back through each work, as it was written for me, a degenerate, by me, a sicko. I highly encourage you to do the same, as even I was surprised at some of the strategy bits that I wrote about and have since largely forgotten, even moreso than the evidence provided (which admittedly was one year’s worth of data, which I tried to contextualize at each turn).

2025-26 has offered up a new flavor of the Underdog Best Puck Classic, where the contest looks and feels similar to last year, but there’s an extra Flex scored, and an extra team advances out of the regular season and into the playoffs (where all the money is!). Both of these changes, in some form, open up the playing field. To me, this means that results are less dependent on having that one player who goes apeshit, and more about everything else.

Resetting the Stage

There are a million things I want to do here with a year to think about the game and all sorts of new data available, and frankly, there’s a lot we could say about last year’s results and what we learned, and how that compares to the Manifestos I produced. In a world where people gave a shit, I’d love to spend hours a week navel-gazing and using that to look-ahead, much like how the majority of the fantasy football industry spends all spring and most of the summer.

Since people are simply less engaged with the NHL, though, I want to be as direct as possible, since we’ll only get one or two chances to convert most people to the small pond of NHL Best Puck. I can’t spend all hours of the day donating time to an endeavor that already is my hobby and second job for no pay, after all!

What that means is that in some cases, you’ll have to trust me, or you’ll have to ask a more direct question in Discord or on a stream, as I have everything I produced last year for 2024-25 as a comparison, but I don’t have the bandwidth to present it and expand upon it.

Let’s get into some of the most interesting and impactful things I’ve dug up when investigating how to beat this year’s version of the Best Puck Classic.

All data from Underdog Fantasy, Natural Stat Trick, and QuantHockey

Stacking and Advancement

The function of stacking still applied in a similar way to last year - when a team disappointed, they dragged you down, but when they were good, they elevated your squad. No NHL team was more representative of this than the Columbus Blue Jackets. Drafters (there were n=2203 of them) who took exactly one CBJ player (which doesn’t include Johnny Gaudreau, who was drafted 215 times last year) advanced at a 34.3% rate in the new rules/format. This is a noticeable lift over the theoretical 25% advancement rate (in 25-26, top three in each 12-team pod will advance). That said, the n=517 who took two CBJ players advanced at a 56.1% rate and the n=71 who took a three-man CBJ stack advanced 66% of teams!

Some of the story is simply being more likely to capture the “league winner” within your stack (3 CBJ players gives you a pretty good chance of having the 61% advancing Werenski!), but there are other benefits inherent to stacking even if you don’t capture the Fantasy MVP (Werenski was a mid-round pick, making his explosion quite valuable).

Stacking and Upside

Because the new format doesn’t have a clean set of data to use to model my “EV” process from last year (where I could use the real life payouts for playoff performance and tweak it to extend to more than just one “trial” of the playoffs, all using only last year’s data and no “fake” data (AKA sims)), my main weapon is neutralized. Rather than try to sim out several “could be” playoff brackets that quite honestly would really be at risk of having fundamental flaws, in thinking about how to apply the new scoring and advancement protocols, I realized that with more teams advancing, I had to crank up the dial more toward upside on a draft-level, since the playoffs will matter more. Meaning, if all 12 teams advanced to the playoffs out of your draft, all that matters is the two-week upside of the playoffs, and ultimately the majority of that would depend on the finals.

Since we are moving from 2 teams per pod (plus wild cards… I know I know) to 3, we’re focusing more on upside, in some increment.

How do you win a championship, then? Well, you need to hit the nuts in the finals.. but you also have to get there. You see above that stacking can bring advance rates up… if you get the right team, but since you often have to reach ahead of ADP and/or sacrifice positional needs to complete a stack, there’s a penalty to stacking more often than not.

This leads me to trying to maximize playoff output, in the short sprints (two-week playoff rounds). What I am looking for is somewhere within those two-week windows. And those windows are something I can model based on last year’s real results, even if they aren’t technically the playoffs.

Out of the ~13k teams drafted, I want to figure out what sorts of teams had true top percentile finishes multiple times. I made this a clean “top 200”, or just about 1.5%ile. And since some players were drafted with known injuries (but clear second-half value upon return) or as young players with upside (but in need of time to ascend), I started the clock at week 12, or just about January 1st. Similar to last year’s playoff EV exercise, I feel that the ~halfway point is a proper balance of increasing sample size while accounting for some minor situations that drafters may be explicitly accounting for when drafting that hurt short-term/early-season value in exchange for long-term/late-season upside.

For 23 of the 32 NHL teams, stacking them offered more upside, somewhere along the way, than only taking one player. This is an astounding finding, and even more promising is that the 3-man stacks show extremely well in aggregate when focusing on multispike %. If last year was about the ministacks (2 players stacked), this year might just be about the 3-man stacks.

We see multispike % peak at 1.7% for 4 stacked players, but it’s worth noting that in a setup where 25% of teams advance, these teams would have only advanced at an 18% clip. It’s probably not worth that steep of a drop-off.

We see ministacks lead to the highest advancement rates, with good spike_rate, where 9.0% of teams with a ministack have 1 or more top 200 week. It’s just that 3 stacked players hits that sweet spot of sacrificing very little advancement for a noticeable increase in multispike rate, which you’re going to want to win a tournament.

I think in the right situation, 4-man stacks could be viable as a high-risk, high-reward endeavor, but for many of the reasons laid out last year I still think they are generally too thin to definitely need. With the extra flex, however, I think the 3rd man in the stack makes a lot more sense than I advocated for last year, and are in fact a core part of my 2025-26 strategy.

I have found myself coming back to this final section of my manifesto exercise from last year time and again. Sharing again, because I truly think if you take one thing away from my work, it’s this:

FYI, in 2025-26 I count ~85 weeks in the playoffs where a team plays 4 games (and 1 where Pittsburgh plays 5!), and only 22 instances of back-to-back 4 game weeks for a team. More than last year, but still well below the rate you’d expect to see them if “week number of games” was truly random.

Now that we score another player vs. last year, I think the benefits of having natural pockets like this is even more important this year!

ADP Capital and Checking in on ADP Risers/Fallers

One of the more interesting findings that I wasn’t sure how to handle last year was the insane correlation between players who were steamed up draft boards throughout the summer and the league winners. Last year, “sharp” drafters who bested the field (especially in the earliest set of drafts) by ADP closing line value were exceptionally successful:

This was a similar pattern to “current ADP” (excluding what the ADPs eventually closed at, and simply looking at how the pick compared to ADP in the draft applet), but more extreme (current ADP topped out around +35% and -55%).

Part of the reason that I thought the closing line value was such a boost had nothing to do with CLV. Eight of the top ten biggest risers were unequivocal smashes in 2023-24, highlighted by William Nylander, who flew from a 5-6 turn guy to the 2-3 turn over the course of the summer and had a massive season. So it was hard to tell if closing line value was an added boost on top of what we know when we’re in the draft lobby (i.e. does beating the field at close, especially early in the draft season, really matter?), or did we just run into a situation where well-drafted teams with closing line value happened to have the league-winners on them more often than other teams?

Enter 2024-25 data (based on how the 2024-25 contest actually functioned, not adjusted for the 25-26 settings):

Closing line ADP still mattered, and it mattered a lot. It wasn’t quite as extreme as the prior year, but let’s compare to how this looked vs. current ADP:

This looks… similar to closing ADP. Especially when you consider that bucket 1 contains most auto-drafters (who are not stacking, paying close attention to positional allocation, etc.), the 2+3 buckets average just about +30% EV, whether you look at Current or Closing line ADP.

Surely the field did it again, and steamed up only the best possible picks and we can pat ourselves on the back again and call that a true skill, right?

Crap. Seven of the top nine biggest risers over the course of the offseason were net drags on a team, two of which were only minorly negative in EV. Over half of these players were actively bad, where a guy like Connor Bedard simply didn’t produce anything close to what you needed from a round 2-3 center (or a 4-5 center, where he was originally being drafted!). Evan Bouchard is a tad unfair, because he was so tethered to McDavid (who got hurt for the playoffs and was generally bested by MacKinnon all year anyway), but I don’t think there’s anything going on that invalidates these findings.

Ultimately, I will say I am not overly concerned about CLV myself, in that I’m not trying to load up on players in the early offseason simply because I think they’ll move. I think part of that is that anyone I am particularly high on will probably get steamed anyway, because most of the high volume drafters are only/primarily consuming my content (this isn’t the most competitive media space…) and I think I’ve garnered enough trust over the years that people take bits and pieces from me, and these player takes certainly seem like the sort of thing that get soaked up.

In place of aggressively targeting CLV, what I tend to do is more of a mixed approach, where I’m happy to dive down the board in spots to complete a stack or get ahead of a position (such as a goalie avalanche), but I’m extra cognizant of pairing players I get who slide past ADP with players in the next round who are also at or after ADP. That way, if there does happen to be some line movement on either player, I can almost guarantee that I am getting a unique pairing to take advantage of whatever extra benefit there is to closing line value, all while taking advantage of the current ADP value that I am certain exists. After all, if I get two guys at the 6-7 turn who are late 6th rounders, and one of them moves up into the late 5th round, now that pairing of players either simply doesn’t happen, or the team I’ve built around them is “better” (since anyone getting the two players now must spend their early 5th round pick to get the climbing player, rather than the 6/7 pick I spent).

I think here is where I am supposed to find a snazzy way to mix together the stacking conversation with the ADP value discussion, tie everything in a bow, and give you succinct parting words of wisdom. But I don’t have that, because I believe (and the data supports) that stacking and ADP value are opposite sides of the same coin. You tend to have to sacrifice one to get the other.

When it works out, great, take it and don’t apologize, but it’s very hard (impossible?) to control both things simultaneously while on the clock. It takes a tremendous amount of luck to get a stacked player after ADP, so stay on your toes and take advantage of every opportunity as it does arise.

What you can control though, is to follow me on X @FakeMoods, and to subscribe to this newsletter. I’ll be writing more strategy pieces throughout the offseason, and I write ~once a week during the NHL DFS season as well. If you haven’t yet and you finished this article, you should definitely read the manifesto from last year (here (vol 1), here (vol 2) and here (vol 3)). Most of it still applies, despite the changes to the format this year.

You should also subscribe to DJ Mitchell’s YouTube channel, where we draft at least once a week (there are also some in the library if you want some background noise!) and stream live. I’ll be doing some stuff on my own personal YT channel very soon (but I’m not promising a ton on there!), so make sure you just follow/subscribe everywhere. And join the Discord (DM me if you want in the MSP Discord), if you somehow haven’t already.

That way, you won’t miss out on anything!

As always, thanks for reading!