An experiment in mathematical models of conflict and cooperation between intelligent, rational decision-makers.


NHL 2016-17 regular season win totals

With the NHL regular season nearing, here is a rundown of value my model has found for this year’s regular season win totals.

A few things before I get to the picks:


1. The color-coding on #1 and #2 goalies is not a reflection of talent, it is merely a depiction of goalie situations that are/may be flexible. Either the team’s narrative reveals the flexibility, or my analysis tells me that there may be a flexible goalie situation, based on goalie talent.

I weigh the expected share of playing time that each #1 and #2 should see, but the dual-yellow coding means I am not as sure about those playing time splits. Single-yellow coding simply means I am not sure who the back-up will be.

Luckily, the teams I am finding value on have pretty stable goalie situations.

2. I am using BetOnline’s lines for these RSW’s. It’s just easier to quote one book, and BetOnline is very accessible. I can’t use Pinny, because a lot of people don’t have access to Pinny. You will need to use multiple books, though, to get much down at these limits.

You will definitely want to shop these lines, and opt for the option with lower chalk when lines diverge. I added 0.5 to the Hurricanes’ RSW because they are especially chalky at BOL.

3. The color-coding in the “plays” column is as follows: red=definite under play, green=definite over play, yellow=borderline play in the direction noted (both borderline plays are overs).

10% is the threshold, as arbitrary as it may be. The percentage is expressed as (expected wins – RSW line) / (RSW line).

Because I’ve built all of the rosters from scratch, daily NHL picks will begin on day 1 (October 12th). In the past I’ve waited until the first games are played by each team, in order to parse new lineups into the roster database. I’ve never been more prepared for any single season so start.

Without further adieu, here are the plays, listed alphabetically, not by strength. I will list them together first, and then list/write a little bit about them.

1. Buffalo Sabres UNDER 36 regular season wins (-12.57% definite play)

2. Calgary Flames OVER 39.5 regular season wins (+11.42% definite play)

3. Carolina Hurricanes OVER 36 regular season wins (+16.45% definite play)

4. Colorado Avalanche UNDER 38.5 regular season wins (-11.05% definite play)

5. Detroit Red Wings OVER 39 regular season wins (+9.37% BORDERLINE play)

6. New Jersey Devils OVER 37 regular season wins (+12.78% definite play)

7. Tampa Bay Lightning UNDER 48.5 regular season wins (+10.81% definite play)

8. Toronto Maple Leafs OVER 36 regular season wins (+9.83% BORDERLINE play)

Of the two borderline plays, I would much rather back Toronto.

1. Buffalo Sabres UNDER 36 regular season wins (-12.57% definite play)

The Sabres are headed in the right direction, with regard to skaters, but they should have saved at least one of the decent goalies they’ve had in the last couple of years (Johnson, Neuvirth, or even Enroth).

Keep in mind that Johnson started 40 games for them last year. It’s going to be a rough year with the tandem of Lehner and Nilsson in the crease.

If you are a Buffalo fan boy, you may think Lehner is decent because of his seemingly-decent save percentage. You would be wrong, though. Save percentage needs to be put into context. But I won’t get into the details of that.

Where it could go wrong: If the Sabres come to their senses and sign a decent goalie mid-season. We want Lehner and Nilsson all season, Ullmark would be fine too.

2. Calgary Flames OVER 39.5 regular season wins (+11.42% definite play)

Calgary is way better-off in the crease this year. Elliott is a major upgrade over Ramo/Hiller/Ortio, and Johnson is a pretty good back-up.

Where it could go wrong: Serious injuries to key skaters (Gaudreau, Giordano, Hamilton, Frolik, Backlund), and/or a serious injury to Elliott.

Monahan is fine, but overrated. My key skater notes will only include guys who show a good amount of value above replacement.

3. Carolina Hurricanes OVER 36 regular season wins (+16.45% definite play)

The dual-yellow coding on Carolina’s goalies isn’t really an issue, because Ward and Lack are comparably below-average. Their goalie tandem isn’t terrible, but Carolina could be really good with a marquee goalie.

Carolina has trouble creating chances- usually kicking harmless shots on net. However, they do have a good defense, and a fantastic penalty kill.

Where it could go wrong: Serious injuries to key skaters (Faulk, Staal, Lindholm, Nestrasil, Teravainen), and/or serious injuries to both goalies and a really bad fill-in goalie.

4. Colorado Avalanche UNDER 38.5 regular season wins (-11.05% definite play)

It’s never a good sign when your head coach just hangs ’em up abruptly.

The Avalanche are a good defense away from being the New Jersey Devils. They don’t work well together on offense, and they are just bad on defense.

Where it could go wrong: If the Avs give Pickard a healthy amount of starts, as strange as it sounds. Varlamov is severely overrated, due to the context in which his save percentage lands.

All it takes is a disciplined opponent to absolutely crush the Avs.

5. Detroit Red Wings OVER 39 regular season wins (+9.37% BORDERLINE play)

The loss of Datsyuk is probably the root of a lot of this value. He was Detroit’s best player, and people love to overreact to that kind of stuff.

Where it could go wrong: A serious injury to Mrazek, or too many starts for Howard. Serious skater injuries are less of a worry here, because there aren’t a whole lot of stand-out Detroit skaters. Detroit is a team of uniform talent.

6. New Jersey Devils OVER 37 regular season wins (+12.78% definite play)

We’re back for more Devils RSW overs this year. Schneider had some injury issues last year, but they still went soaring over.

Defense is the name of New Jersey’s game. They will be “in” almost every single game they play this season.

Where it could go wrong: A serious injury to Schneider, as he is one of the best goalies in hockey. Like Detroit, New Jersey has uniform talent, so key skater injuries aren’t as much of a worry.

7. Tampa Bay Lightning UNDER 48.5 regular season wins (+10.81% definite play)

I was expecting this under spot to be taken by Washington, Pittsburgh, or Chicago, because there is always an under to be played on darlings/teams who killed it the previous season. I’m just surprised it’s Tampa.

There’s not a whole lot to say about Tampa. They are a good team, but not a 49-win team.

Where it could go wrong: Nothing unforeseen to worry about here. This play could certainly lose, but everything will have to go right for Tampa. Anything but their optimal lineup and goalies will probably result in under 49 wins.

8. Toronto Maple Leafs OVER 36 regular season wins (+9.83% BORDERLINE play)

Ah, the youth movement. The Toronto Marlies (the Leafs AHL affiliate) has been a fixture atop the AHL standings since 2011-12. We are starting to see that talent filter into their NHL roster, most notably William Nylander last season.

This season, along with the top Marlies we got a glimpse of last season, they will add more young talent in Auston Matthews, Mitch Marner, and Nikita Zaitsev (from the KHL).

The young Leafs will undoubtedly make a ton of mistakes this season, but they should also surprise a lot of top teams.

What could go wrong: A slow acclimation process for the young talent, and/or a serious injury to Frederick Andersen.

Sports Betting Quick Tips

I want to take the time to run through some of the most crucial things I’ve learned in my time sport betting.  I’ve been doing this professionally for several years, and I’ve definitely  paid my “bettor’s tuition” in order to learn and, more importantly, follow these principles.

  • Keep detailed personal records.

Moving from third-party trackers to detailed in-house documentation has helped me a great deal.  When you use a third party tracker it isn’t as accurate, detailed, or personal.  You can add whatever metrics you want to your in-house records to make them more detailed and helpful to your growth as a sports bettor.

Here is a link to my detailed in-house records: Record Keeping Archives

  • Performance tracking goes far beyond W/L, win percentage, and units gained.

You need to track market agreement to truly gauge the value you are cultivating.  CLV (closing line value) is a very important tool for assessing market agreement.  It’s like the stock market; if you bet a play 5 minutes before close just because you are too busy/lazy or indecisive to make a move earlier, it’s like buying a stock and selling it 5 minutes later at the same price.  But there are exceptions, like when new information breaks late and you have to make your move within 5 minutes of close; but it’s likely you will grab CLV in those 5 minutes if that’s the case.

W/L records are full of variance, but there isn’t any variance in CLV.  Either the market agrees and is heavy enough on your side to move, or it isn’t.  There aren’t enough accidental bets (variance) in the world to influence your CLV substantially.  If you are beating the market consistently and not churning a profit, you can be fairly certain that your luck will change.  You are on the right track if you are grabbing a ton of CLV consistently.

I don’t take anyone who claims to be a professional sports bettor seriously unless they track CLV and beat the closing line consistently.

  • Use models to create your own spreads and expected win percentages.

If you are betting with your gut, you’ve already lost.  You have to create balanced probability distributions that are accurate.  This isn’t easy to do, and requires a ton of knowledge and trial/error.

  • Bankroll management is as important as the picks you are betting.

A sports books biggest boon, along with parlays, is the fact that most people are AWFUL at bankroll management.  I’ve met accountants, financial advisers, investment consultants, etc. who don’t have a clue about how to manage a gambling bankroll.

There are pretty much two viable options:

Your first option is a compounded flat bet.  Establish a bankroll, let’s say $5,000, and bet no more than 3% of your bankroll per pick.  After each month  you compound your bankroll, so if you started with $5,000 and had a good first month, winning $600 @ $150 per pick ($5,000*.03= $150) you will have a new bankroll of $5,600 and a new unit of $168 ($5,600*.03= $168).  It works the same way, but your unit decreases, if you have an awful first month and the cycle continues in perpetuity.

If you are someone who has managed to create an accurate model that has shown an ability to create BALANCED probability distributions that can pinpoint, within reason, how often an event will happen over an enormous sample of bets, then you can use Kelly Criterion.  I won’t go into Kelly Criterion in detail, but I will say that you should never surpass 1/2 of Kelly’s recommended stake.  If you have a large bankroll with life-changing amounts of money, never surpass 1/4 of Kelly’s recommended stake.  Here is a Kelly calculator for you to play with.  Note that the odds are flipped, -110 is 10/11, not 11/10 so something like -140 is 10/14 not 14/10.  Remember to compound your bankroll monthly.

  • There are no such things as “locks”

This is one of the biggest mistakes I see amateurs make.  It is a cardinal bankroll management sin to bet something as if the result has already been decided.

Everything falls on a probability distribution.  Any possibility is possible, that’s why it’s called a possibility.  Just because a team gets mashed by 30 points doesn’t mean that if they run it back the next day the losing team from the day before can’t win by 30.  Both events fall somewhere on the probability distribution.

Another related issue is bettors measuring their expected value based on a lopsided, or a series of lopsided results that go for or against them.  Singular results are irrelevant for future predictive purposes.  The event happened, but that doesn’t mean that the result represents the median distribution.  You could land on the tails (unlikely part) of a distribution for 10 plays in a row; they are all unexpected results that are very unlikely, but they will happen occasionally, and sometimes frequently over a small sample.  If you land on the right side of these distribution tails you may think you are a god and you will get pounded back to Earth for your irrational exuberance.  If you are on the wrong side of these distribution tails you may completely scrap a good model because you are using the wrong information to judge your future expected value.  Either way you are getting fooled by singular results and you deserve what’s coming to you.

If anyone ever refers to a bet as a “lock”; run away, don’t walk.  I’m not saying to fade them, just don’t let them have anything to do with your betting experience.

  • Shopping lines is very important.  Create multiple betting accounts (outs) so you can get the best number available to you and increase the amount of money you can get down.

The heading is pretty self-explanatory.  Just have enough outs at your disposal so you can shop for the best line.  As I mentioned above, 10 cents makes a huge difference and some of the various outs you have could differ by 10 cents regularly.

You always need to understand what kind of lines your outs offer.  For MLB and NHL you need to have 10-cent books like BetOnline, 5Dimes, Pinnacle, etc. A 20-cent book will offer -110/-110 (2.38% vig, -4.76% ROI), but a 10-cent book offers -105/-105 (1.22% vig, -2.44% ROI).  All vig will spread out a little more the further you get away from 100 (you know this if you understand vig conversions), so a 20-cent book will offer something like +170/-200, while a 10-cent book will offer something like +175/-195.  Most books only offer 20-cent lines on things like NFL spreads/totals, NCAAFB spreads/totals, NBA spreads/totals, and NCAAB spreads/totals, but there are a ton of books that offer 10-cent lines on MLB and NHL spreads/totals.

  • Understand sample sizes and what constitutes relevant information.

If you think the following is a relevant piece of information, stop betting right now and do not resume until you understand why it isn’t relevant: “Patriots are 23-9-2 ATS in their last 34 games after accumulating less than 150 yards passing in their previous game.”  That’s on the tame side of situational trends, some include 10 years of team data.  Do you think the Patriots were the same team 10 years ago as they are today?  They could have the exact same record for 10 years straight and be completely different teams each year.

I won’t go into full detail about how awful trend betting is, I’ll just give an examples of how lazy and foolish it is.  Can you predict how your morning commute is going to go because the last 12 times you ate eggs and bacon for breakfast you didn’t hit a single red light on your way to work? No. The same applies for sports betting.  There are so many stupid trends that I could write about them for the rest of my life and never run out of material.  And yet trend betting is the life blood for your average tout and Joe Schmo buys into it.

I need to be clear, situational trends are way different from overall functional trends in sports.  There are useful oddities hidden in the dynamics of sports that are very useful.  Here’s an obvious one: there was a time that odds makers split college basketball first half and second half totals 50/50.  I think everybody now knows that odds makers allow for 7-8 additional points in the second half of college basketball games.  The NBA is pretty close to a 50/50 split on first half and second half totals though.  Those kinds of functional trends are still hidden in some sports, although they are much harder to find now than they were in the past.  The key is that they are trends that encompass the game, not a single team or situation.

Sample size is related to situation trends.  Trends have a very short shelf life because they are just small runs of major variance that stick out due to small sample size.  I wouldn’t actually call it a shelf life because situational trends should never be on the proverbial shelf.  You have to understand that variance is a function of sample size; the smaller the sample, the larger the variance.  This is important to remember when building the models I discussed earlier.

An excellent book about sorting through relevant and irrelevant information is “The Signal and the Noise” by Nate Silver.  It is a fairly casual read that provides the big picture of evaluating data.

I will update this page with more advice in the future.


I’m not saying these principles are anything groundbreaking, or anything that a lot of you don’t already know/practice.  I’m just laying them all out there no matter how elementary some of them may be.

Choose to understand and practice these principles or not, it doesn’t really make any difference to me personally.  The sports betting world is full of false dreams and misinformation, so I know that many amateur sports bettors have trouble deciphering what is true and helpful from what is complete BS.  Most amateurs gravitate toward the BS because that’s where they find their cozy false dreams fed to them by shameless profiteers and overall uneducated people who don’t know any better.

However, if you follow these principles, you don’t need anyone to give you picks.  You can bet safely and enjoy picking your own games.  Although you’ll need to master these principles and then some to turn a large and consistent profit.  If you aren’t content with safe, fun, recreational betting, and want to receive the bets I personally play on a daily basis, you can sign up here.