Welcome to part 2 of my round one preview, which was split into 2 parts because Dallas-Vegas and Edmonton-LA play their first match on Monday, a night that will also have 2 game twos. This is ostensibly a standard picks post for Monday, but still under the R1 Preview banner. I may have teased that this would preview 6 game twos, but later realized that I’m not waiting until the end of Canucks/Nashville to post my pick for that game two. My extra effort promoting part 1 on Twitter resulted in 4 clicks from that platform, and those were probably subscribers browsing my Tweet history searching for parlay polls (thanks to everyone who voted, minus those who selected Toronto).
Before we go any further, it’s time for my regularly scheduled obligatory *DISCLAIMER* it needs be noted that I’m not betting with real money. These are all fictional wagers in a spreadsheet. My mission is to engage in a mass betting campaign, picking a winner of every single game, every over/under, because it provides a complete dataset for macroeconomic analysis, which can be shared with you, shedding light on what worked and what failed. I’m also tracking the results of betting every outcome, to help me (and you) uncover previously unknown or newly emerging profit vectors. What started as a thought experiment has evolved into much more.
Getting back to business; one key statistic about game two in round one of the last 2 playoffs, home teams are 12-4. This might be the product of road teams going 10-6 in game one, making the higher seeded host more desperate to win game two. This is likely the behavioral science explanation for the “zig zag” phenomenon. Do or die can be a powerful motivator, but some teams (or Mitch Marner) buckle under the stress. As discussed in part 1, betting the loser of the previous game is a very popular strategy and you can often see it in the line prices. A team can be -140 on the ML, lose 5-1, then be -170 the next game. That’s very common in playoffs, and it’s a direct response to public betting habits.
This was discussed in the introductions of my 2 new “zig zag” models. The fact that oddsmakers nerf the lines (especially when the higher seed is the one who lost) does increase difficulty extracting profit from this angle. Building my “Orthodox Zig Zag” model was a tough slog and required an uncomfortably large number of cuts. Would it have been worth my time to explore the anti-zag angle? Well betting $100 moneyline on the team that won the previous game in the previous 3 playoffs resulted in a $900 loss, but the bulk of that came from road teams who won previously. Betting $100 on home teams who won the previous game yielded a $200 loss on the moneyline, but a $350 gain -1.5 goals. No home team has lost yet.
Since 2014, home teams that lost game one (any round) went 38-24 in game twos. In that span, road teams only went 62-73 in game ones. So, the recent trend of road teams being successful in the opening game does not sustain going further backwards in time. My new team of playoff specific models have most of their money thus far on road moneyline or home -1.5 goals, and on opening night, H-1.5 went 2-0, VML 0-2. That was very nearly 3-0 and 0-3 until Steven Stamkos scored a goal to kill the Florida puckline in the dying seconds. Certainly, the early evidence is pointing towards a return to home dominance, but I logged my two visitor picks tomorrow before today’s games and am not changing that.
Part 1 introduced my new team of betting models, my 9 “Cupwraiths” who went 8-10 on opening day (they had 8 home bets, 10 road), while my other “Expected Goals Value Voodoo” algorithm went 2-0 (I still have no idea why that design worked, but there was a clear visible pattern when looking at the inputs) so that’s 10-10 altogether. That post did not introduce my 4 new over/under models, who went 5-3 opening day (in round one game one of the last 2 playoffs, unders went 8-6-2. In game two, overs went 10-6). None of the new OU mods are overly complicated (unlike my “next gen” zig zag) and won’t take long to describe here.
1. Prime Optimus
The first model built did not have a specific angle acting as the primary input, simply taking every game where the opening total was 5.5, then looking at which input variable had the biggest difference (in standard deviations) between overs and unders. In the case of 5.5, the optimal variable was combined save percentage of the two teams last 5 games (as visitor or host). If that’s greater than or equal to .911, it makes a minimum bet on the under. If it’s less than .911, it makes a max bet on the over. Easy peasy, rinse and repeat for 6 and 6.5 totals. It generated a 16% return on 177 historical wagers. It went 2-0 on Saturday.
2. Cumulative Goals
In my betting spreadsheet, this one is named CumGoals, but not sure that title is appropriate for a public moniker. The process is very simple, are the cumulative goals (averaged per game) in a series above or below the betting total? If it’s >=0.8 then max bet on the over, if it is less than -0.9, max bet on the under. There are 4 compartments in between the extremes, mostly making minimum bets on one or the other. For game ones, it uses average goals last 5 games against only other playoff teams. CumGs bet both unders Saturday, going 1-1.
3. Avg Goals Last 5 Games 2.0
My former primary over/under algorithm (for more than a year) was simply averaging the number of goals scored in each team’s last 5 games. Crudely simple, but was remarkably effective after its introduction in the early stages of the December 2021 Omicron scoring boom. This one ostensibly does the same thing, but only uses road games for the road team (visa versa for home) against other playoff teams. The difference between this model and Cumulative Goals is that the extremes were the least profitable compartments, such that >= 1.3 it actually makes a minimum bet on the under (the opposite extreme is also a min bet on the under).
4. Expected Goals Playoffs
There are a few different over/under models/algorithms in my advisory team who use expected goal numbers (all strengths) from Natural Stat Trick to make picks, one is in my Small Council, while the other is a model and was my best OU performer by far in the 4th quarter. That means each of the aggregators in the pick graphics have at least 1/5 of their recommendation coming directly from this statistic. This one doesn’t have a snappy name, simply called xGPO in my spreadsheet. This model bet both unders Saturday, going 1-1.
Monday Picks
One other piece of business before we get to the picks, all data from my 5 lowest performing regular season models was dropped from the pick graphics. Now it’s just top 5, next 5, and the newbies. The “model % invested last 30 days” and “model return on $1” do not include data for the new models, or the 5 worst models. But, the “models % invested” for this specific game (the top one) DOES include the new playoff models. In total, there are 20 models/algorithms contributing to the “largest model position” (that’s likely to shrink as poor performers will get benched). For over/under, my Small Council is continuing in its current form (despite bad performance down the stretch), but lost top billing to a raw aggregation of my 5 best performing Q4 contributors.
That being said, all the other models/algorithms are still active and logging picks every game, so if any emerge as hot, they’ll get bumped up. Any of the top 10 who go cold may get demoted. Any of the Cupwraiths who lose a large majority of their early bets will stop having their picks shared in pick graphics until they heat back up (if ever). Tough to say, some might work in round one, but not later rounds, or visa versa. There will be a few parlay options at the end, and a Twitter poll. Those polls are the last little bit of value that Twitter is providing. By the way, Elon can call his new platform whatever he wants, but they’re never changing the Twitter.com address.
One other note of interest: The 14 new playoff models went 0-14 on the Tampa-Florida game. They all had either Tampa, Florida -1.5 goals, and all 4 OU had over 5.5. The Stamkos goal with 9 seconds left cost them $2,300 (not real money). That game didn’t finish until all the writing above had been written (except for one comment at the end of a paragraph added afterwards) but I’d like to get this out soon so I’m not re-editing that for new information. They went 0-14 in the first game today. Caps-Rangers is nearly finished.
Vegas Golden Knights at Dallas Stars
My guess is that the Dallas Stars would have preferred playing LA in the opening round instead of the defending champion who bulked up at the trade deadline, but it also feels unlikely that Dallas was the preferred round one opponent for Las Vegas. Dallas was my pick to win the series before even looking at the betting lines, given they were far more reliable than Vegas down the stretch. Both teams boast impressive firepower at the top of the line-up, but Dallas does have the edge in goal. Struggles in net after the injury to Mark Stone is the reason Vegas came unnecessarily close to missing the playoffs, but Logan Thompson did improve in the final couple weeks.
Vegas managed to finish 12-8 in the 4th quarter, which is less impressive than the Stars 15-4. My betting models overwhelmingly preferred Dallas in game one, soliciting 85% of their total money. Of all the road teams in the opening game of round one, my 9 new playoff models liked Vegas the least, and by a considerable margin. Dallas was their second biggest bet on home teams -1.5 goals. From my perspective though, the potential return of Mark Stone does make this a dangerous Dallas puckline bet, so I’ll be playing it safer on the moneyline. My series bet in Dallas at -130, winning in 6 games at +450. I’m also taking over 5.5 goals, but my new playoff OU models did not like that pick.
Los Angeles Kings at Edmonton Oilers
For the 3rd time in as many years, the Edmonton Oilers and LA Kings will be meeting in round one of the playoffs. Reviewing the betting results from the previous 2 series, home teams +1.5 goals (both favorites and dogs) went 11-2 in those 13 games, I’m just not paying -465 to bet that here. The Oilers are once again my series pick at -192 (last year they were -220), and I’m once again picking them to win in 6 games (+425) except I wrote a different answer when filling out my bracket this season (7 games). That was likely an error, since I was trying to put the same answer on both bracket and blog (other than picking Florida when it’s even odds). Last year this was the only series I picked in the correct number of games (at +400).
The Kings were identified by my new models as a juicy upset candidate, as the 52% number you see up there was their largest VML of all the opening games. My existing best models had a strong preference for Oilers -1.5 goals, and home teams -1.5 are off to a strong start (better without that late Stammer snipe). Given that my pick is Oilers in 6, they are also going to be my game one pick. However, I’m not tailing the top models on H-1.5, simply making a minimum bet on the Oilers ML at -162 (which has moved to -170). My Small Council wants to take under 6, but I’m going to tail the others who are 84% to 44% on over 6. Yes, unders are off to a hot start (until this Rangers game (maybe not)), so this will just be a minimum wager.
New York Islanders at Carolina Hurricanes
The New York Islanders very nearly seduced me into picking them game one, and played well enough against the Hurricanes that several points during the game, I regretted not picking the Islanders ML at +184. That game was much closer than +184, and Freddy Andersen was the only reason New York wasn’t winning after 2 periods. Yes, it’s likely Freddy will be back Monday, but seeing the Isles open as a +210 dog was confusing after the game I had just watched. Sadly, I was still 10% too cowardly to take NYI +210 (given that I have Canes going to the Stanley Cup final), so settled instead for +1.5 goals at -130.
Under 5.5 was my choice in game one and it hit. That was my only under of all the game ones, so at least I hit 100% of my R1-G1 unders. My models preferred that pick again for game two, and it looks like goals will be tough to come by in this series. The Islanders defense was blocking a crazy number of Carolina shots. The only thing that gives me pause about fully embracing unders going forward in this series was the high goal totals when they faced off in the regular season. It was also noted earlier that overs tend to be better in game two, although it’s a conclusion built on a relatively small sample size.
Toronto Maple Leafs at Boston Bruins
My betting models liked Toronto’s chances of a game one upset, but we found out shortly before puck drop that William Nylander would not be playing. As a Leafs hater, this should have been joyous news, but the team found their way into half my parlays. Most of those were democratically chosen, even the initial versions shared here were based on early polling results. That’s not my fault. There was a news story about Nylander missing practice that was framed on TSN as “gamesmanship has already begun” which gave me the impression that he would play. But I also didn’t read the story, which may have revealed the Nylander injury could possibly lead to not playing.
In my defense, I ostensibly laid out the definition of the “gambler’s fallacy” in my Leafs game one pick justification, though the Leafs were getting substantial votes before my preview had even been posted (I deliberately posted them early because Friday afternoon is more fertile for democracy than evening, and wanted to rank them on the blog by early results). My Cupwraiths are divided and my best models prefer Boston 56% to 10%. Given that I have the Leafs going to the conference final, I’m going to double down on my initial wager. My justification this time is based on my own personal theory that the longer a streak lasts, the more likely it is to end (still unproven, but it’s on the summer checklist).
My top models are doing a better job than me navigating these teams lately, but also 42% of their investment is on Toronto, so my choice isn’t entirely fallacy. But considering we have no idea what’s wrong with William Nylander and Samsonov does not look good, there is an increased risk on the visitor’s side. I’m almost talking myself out of this Leafs pick, but don’t want to re-do the picks graphic and would like to get this up soon. Over 5.5 goals hit in the last game, and both my Small Council and new model team want me hitting that again, so that’s the plan. Samsonov might be the biggest reason to bet Boston and the over.
Parlay Combinations:
I’m currently running 2 Twitter polls with my bets and my model bets. I’m not framing the question as a parlay, just which bet do you like better. When the polls wrap, I’ll give out a parlay with the best bets from both polls (unless they are conflicting). If you don’t want to wait for democracy or you’re not on Twitter, here’s the poll ordered by model investment.
Largest Model Positions
Dallas -1.5 PL, Los Angeles ML, Boston -1.5 PL, which currently parlay to +1899. Add Carolina -1.5 for +4098. You wanna get nuts?
My Bets
Edmonton ML, Dallas ML, Islanders +1.5 PL, which currently parlay to +397. Add Toronto ML for +964.
UPDATE: Early polling pointed to Dallas ML, Edmonton ML, Boston -1.5 PL and Carolina -1.5 PL as the clear favorites, so I cobbled together an impromptu final, but am going to leave the semis open for those of you who never had the chance to share your opinion on the original options. Or you can just parlay the top 3 right now.
Good luck! Next picks post coming tomorrow! Go Vote!