2022-2023 NHL Preparations

Today we go through example preparation for the upcoming season.

coffee
September 30, 2022

2022-2023 NHL Preparations

The NHL season is slowly approaching with the preseason already underway. The first puck of the 2022-2023 preseason dropped last weekend. This means it's time to shake the rust off our models and get everything prepped for the 82 game season. In preparation for the upcoming season, let's go over some housekeeping and basic risk modeling and general betting theory for the fast approaching season.

For any new season on any sport we want to be reviewing team/conference changes, abnormal schedule rotations, rule changes…the factors are overwhelming if we are not prepped before the season starts. Those having individual player models for props or bottom up modeling will also want to review big trades, roster realignments and especially review all expected 32 starting goalies. In the sport of hockey goalies are very similar to what a pitcher would be in the MLB. With only 32 and a handful of more split game teams it's not an impossible task. Modeling In this sense doesn't necessarily apply only to quantitative models either. It's broad encompassing, even for those sharps of you that can simply watch game flow and bet with edge.

As far as rule changes go for NHL 2022-23 there is really only one major change. Referees will not have the ability to nullify a major penalty after video review. This season they will be able to reverse a major penalty to a minor penalty. Is every power play minute crucial to the outcome of a hockey game? Everyone models differently, but generally power plays have a huge impact on how a hockey game progresses and especially on the Total Goals for the game. Adjust your models accordingly or at the least be on the lookout come season time for whether the books have compensated (if even needed) for this rule change correctly on the Totals. Wondering how to assess the impact on your model? I have included a link here to a great paper by Patrice Marek on an example of effects from previous NHL rule changes.

We will also want to plan out what tools we will be using throughout the grind. A number of tools are now available that can help the sports bettor develop an edge. There are also a number of "tools" that should be avoided. Separating the two is something I will leave to the reader. It's our hope that sportsbettingintel.com will make it on that list of helpful resources as we provide a set of both free and premium tools to our users. There is no reason to limit your tool set either. When building a house it would be very difficult to get by with only a hammer. You will generally need a number of tools to help. While you could conceivably use a screw driver to hammer in a nail, it's generally worth the money to buy a hammer as it will pay off in the long run. This holds true for sports betting. Tools, data and analytics can really pay dividends and help achieve a higher level of profitability offsetting whatever little costs there may be.

Up-to-date with rule changes, have our models cleaned up, our data providers are in line, the sports betting tools and resources we will be using are cataloged and planned for, what's next? Without proper planning and risk assessment sports wagering can be more akin to rolling a couple dice at a craps table. Great boilerplate for any profession right? Planning and risk assessment are key in most industries. For the sharp sports bettor it is no different. Just a casual bettor? I urge you to follow along still and spend a little time preparing. We'll keep things light.

First things first. We need to know our opportunity available for the upcoming season. NHL.com has 1,420 games listed for 2022-2023. Knowing that each team plays 82 regular season games (covid be damned) we can deduce this leaves 108 preseason games using the simple math below. Plenty of time to shake out the rust we mentioned before. Perhaps for the more ambitious bettors out there 108 events to create more opportunity.

\[REGULAR \ SEASON = {82 \ GAMES * 32 \ TEAMS \over (2 \ TEAMS \ EACH \ GAME)} = 1,312 \ GAMES\]

\[PRESEASON \ GAMES = {1,420 \ TOTAL \ GAMES - 1,312 \ SEASON \ GAMES} = 108 \ PRESEASON \ GAMES\]

That math is easy enough right?? We could always google how many NHL preseason games and find the answer, but what's the fun in that? Better to ease into things for later anyway. 108 preseason games is a lot, but let's not snooze on getting things done now. For the sake of going forward we will have to make some assumptions to create a base line for continuing. Let's make the assumption that we are betting only Moneyline -INC OT on the NHL games and we have identified that we will be betting around 20% of the matches. This would equate to betting 2 or 3 games on a normal big card day for the NHL. If your strategy is only betting 1 game on a big slate day, or 1 game every other slow slate day, just cut the percentage in half to 10%. Now the biggie, and this differs for everyone. For the sake of the article let's put our bet size somewhere between a buck ($100 average bettor) and nickel bettor ($500 average bettor). We will use a static bet size of $200 per bet. First thing we will look at is our maximum season loss.

\[MAXIMUM \ LOSS = {TOTAL \ GAMES \ BET * \ BET \ SIZE} \]

\[OUR \ MAXIMUM \ LOSS = {(1,312 * 0.2) * 200} = $52,400.00 \]

Yep, if we are betting $200 on every bet and we take 20% of the offered games we could theoretically lose all 262 of our bets for a total loss of over $52K on the season. Enough to shy away any bettor right? Not really. This scenario is so highly unlikely its not really worth considering unless we are using 100% of our net worth (please never use your entire net worth as your bank roll). For those of you interested there is a 0.00000000000000000000000000000000000…00010226549804791400% chance of this loss occurring. Yes the ellipses are a series of 0's I had to take out because it just looked weird. Apologies to everyone on mobile. ...And for the proverbial more likely to be struck by lightning? You are far more likely to be struck by lightning. Almost 10x more likely, but I think everyone gets the picture now. So how do we go about modeling our max risk then? One commonly used way is to model a bell curve and drop two standard deviations out to go about finding a more appropriate measure of max loss. Simulating the random outcome 10,000 times of placing 262 bets over the course of the season we can see the breakdown below.

σ is a quick term for Standard Deviation. The lightly orange shaded area shows total winnings within 1 σ. The lightly shaded red shows total winnings from the simulation within 2 σ. As we are looking for the 2 Standard Deviation move on our losses we can see that ~ $6,463 is a good estimate of our max loss on the NHL season, knowing in the back of our head that it can actually be higher. Once we were to hit this number on our losses it would strongly signify our model has a flaw to it and we need to reassess our strategy. (This all assumes we are wagering on -110 lines, No Tie Moneyline bets, from a reputable sportsbook)

Now for the key point that I know a lot of you more seasoned sports bettors have already figured out. This number may be a good approximation for our max loss in this scenario, but it is an awful measure of how well our system is actually doing. Though our occurrences are not that high the chances of even this 2 σ event happening are not great. How do we measure what we'd expect to lose then on fair markets as a baseline? This part actually becomes a bit easier. Same as before we will assume -110 lines for each bet. Leaving aside any kind of shading from the book (a topic perhaps for another blog) This will mean we are paying 2.38% juice on each bet. Knowing this we can calculate our season long loss assuming we were to flip a coin every time to pick our bets.

\[TOTAL \ EXPECTED \ LOSS = {TOTAL \ BETS * TOTAL \ BET \ SIZE * JUICE}\]

\[TOTAL \ EXPECTED \ LOSS = {262 * $200 * (2.38 * 0.01)} = $1,247\]

Total expected loss on making random bets throughout the season would be $1,247. Another way to calculate, just use our total bet amount expected throughout the season @ $52,400 (our matched max loss from earlier) multiplied by our juice 0.0238. This $1,247 is the number we have been working toward. If at the end of the season our total P/L comes in at -$1,247 we can assume that our model has no edge. The number of occurrences (bets) being at 262 gives us a small sample size, but we have to work with the data we have. New to the sports wagering world? Using these formulas and logic one can plug in their own numbers to calculate what the bet size should be to maintain a comfortable risk level. This logic should only be used as a guide to make more informed Risk/Reward decisions. It is always possible to lose every dollar you bet.

Of course this is a baseline and not considering using tools like sportsbettingintel.com provides for line shopping. When finding the best lines we can often find synthetic no holds across books or lowered juice using best lines. On a synthetic no hold we can calculate our expected loss to be @ $0 by this same logic. This is a topic for another time though. Finding synthetic no hold and low hold becomes a very valuable mecahnism for a sport like hockey especially, where lines very often have more juice than other sports.

Do your work, be prepared, choose the right tools. Nothing wrong with playing for fun, but if your goal is to be profitable put in the work yourself or use tools like SBI to help you save time.


Posted by Aces High