CFL Week 10 Predictions

Ottawa at Winnipeg
BC at Toronto
Montreal at Edmonton
Calgary at Saskatchewan

Week 10 of the CFL kicks off with two moderate favourite home teams; Winnipeg and Toronto. If these two were not playing at home then they would only be very slight faovurites over their respective opposition.

Edmonton are the biggest favourite of the week over Montreal, who are still by far the lowest rated team in the league. I'm still holding on for a breakthrough performance from Johnny Manziel though.

I have a soft spot for Saskatchewan, as they have won me a lot of money this year, but Calgary, the highest rated team in the league, are, unsurprisingly, a fairly big favourite to beat them.

Why I Only Bet On Winners (and Not Point Spreads)

My impression of American Football betting is that most people bet on the point spread between teams. For those unfamiliar with the concept of points spread betting, there is a really good explanation here (it even uses an NFL example!). This impression has mostly formed because of the 30 for 30 episode on Jimmy The Greek.

I only bet on the winner though, sometimes referred to as the 'money line' (unless you are having an online chat with your bookmaker and they don't understand the term 'money line'...). Another explanation from oddsshark on money lines here.

I take this approach for two main reasons. Firstly, teams don't compete with the point spread in mind (well, unless they or someone they know has a bet on it...). Instead, they play to win. If a team is 1 point up with a minute to play they will run out the clock, not carry on playing to beat a 3.5 point spread.

The other issue I have with the point spread is that, Elo modelling, in my opinion, is far more suited to pure a win or loss situation for American Football. Fivethirtyeight had to derive some contrived method, incorporating logarithmic functions, in order to make their model work with point spreads. Whenever you add assumptions into any predictions, you add sources of error. Fivethirtyeight admit themselves, in that link above, that their point spread model is not accurate enough to beat the bookmakers. In fact, they are only correct 51% of the time. My model picks the correct winner ~64% of the time.

My CFL Model Is More Accurate Than The CFL.ca Writers

This might be a premature brag that ends up biting me at the end of the season but I'm going for it anyway. My CFL model is better at pcking CFL game winners than the writers on the official CFL website. I don't know who these writers are and whether their opinions are that respected, I barely even the rules of the sport. But after week 9 of the CFL the respective predictions are sitting at:

• My model: 25-9 
• Matthew Cauz: 24-10 
• Chris O'Leary: 24-10 
• Marshall Ferguson: 23-11 
• Pat Steinberg: 22-12
• Total Pick 'Em Players: 22-12 (at least I think that's what the number at the bottom this page represents)
• Jim Morris: 21-13
• Jamie Nye: 18-16

We'll see how this finishes at the end of season.

CFL Week 9 Recap

I placed one bet on the CFL week 9 games: on Ottawa to beat Montreal. Not the most exciting proposition as they were a pretty heavy favourite. Ottawa ended up winning, although it was only 24-17 and secured in the last few seconds. This keeps my 100% betting record for the 2018 CFL season going. That 100% only equates to 4 out of 4 bets though. My ROI has dropped to, albeit still ridiculously high, 191% from 286% as Ottawa were such a large favourite.

From a predictions points of view, the model got 2 out of the 3 games for the week correct. The BC Lions pulled out a bit of a surprise win over Edmonton so my overall rate for the season is 73.5%, 24 correct and 9 wrong.

Underestimating Saskatchewan Was A Bettor's Dream

Both the pundits and the bookies severely under-rated the Saskatchewan Roughriders in weeks 4 and 6, which enabled very favourable conditions for bettors.

This occurred because in week 3 they lost to Montreal. Now Montreal might be the worst team in the league (although their fortunes will be on the up now they have Johnny Football! /s) but just because a team loses to them once, doesn’t mean they are necessarily worse than them. If you look at the CFL’s own power rankings they went from the 6th best team in week 3, to the 9th (and last) team for week 4 and up to 8th going in to week 6. TSN had them going from 7th ranked to 9th going in to week 4 and 8th going in to week 6.

These pundit underratings, or over compensation for a loss to the worst team, was also reflected in the bookmaker’s odds. From my modelling point of view this equated to massive differences between the betting odds and my Elo ratings for their next two games, in weeks 4 and 6 (no game in week 5), which is apparent in the graph below for the Saskatchewan Roughridgers edge. Their edge goes from a fairly low margin to jump up and then back down again.

It was these two large edges that created the environment for good betting spots. I had them 56% probability of beating Hamilton in week 4 and 47% chance of beating Hamilton in week 6. I managed to get a 2.2 to 1 win then a 4.4 to 1 win on what were roughly coin tosses. Thank you bookmakers.

The Hamilton trend line in the graph above shows that I also have the bookies consistently overrating Hamilton and thus creating negative edges at every game this year. This has not been by a particularly significant amount but all 3 of my bets on the CFL this year have been betting on teams to beat Hamilton. These have all been big, underdog winners with the third bet on Ottawa at nearly 2 to 1 when I had them as a slight favourite.

The third bet is that large Ottawa edge in week 7 on the graph. Now if you are interpreting that first, positive y-axis gridline as the threshold for when I place a bet, then I’m sorry to disappoint, but it’s not quite that simple.

CFL To Date

I am doing this catch-up article on the Canadian Football League as, although I have been betting on it all of this season, I only started this blog the other day.

It is 8 weeks into the CFL season and my betting has been incredibly profitable thus far. I’m sure I am riding the crest of wave of good fortune and the tide will turn soon but for now I am basking in the glory.

The table below shows all of the games through the first 8 weeks with my models predicted winners and actual winners. This equates to a 74% success rate in picking the game winners in 2018 so far. I don’t know if there are any real surprises in there. Maybe the week 4 game between Hamilton and Saskatchewan. I’ve got much more on that in a forthcoming article on Saskatchewan being underrated around that time of the season.

I have only placed 3 bets so far in the 2018 CFL season out of the 31 games through week 8. All winners, all on underdogs and all against Hamilton. These were:

Week 4, Saskatchewan hosting Hamilton at +222
Week 6, Saskatchewan at Hamilton at +447
Week 7, Ottawa at Hamilton at +192

I’ve graphed this out below:

This is evidently a low volume of bets but that's just what my model throws up, it is a fairly low volume system.

This gives me an ROI on my bets of 286%, which is laughably high and unsustainable – but I’ll enjoy it whilst it lasts.

CFL Week 9 Predictions

Edmonton at BC 
Hamilton at Winnipeg 
Montreal at Ottawa 

Three fairly straightforward games for week 9 of the 2018 CFL season. I have Edmonton as a fairly large favourite at the BC Lions. Winnipeg are a slightly bigger favourite at home to Hamilton. Ottawa finish off my picks as the biggest favourite of the week against Montreal. This is not that big a surprise as Montreal are the lowest rated team in the league by quite some margin. Hopefully Johnny Manziel will be starting for Montreal, just for the bants.

Introduction

Welcome to my blog!

I started this to capture and keep a record of the bets I make on various American Football leagues. These bets are based on a system I developed using a combination of the Elo rating system for the competing teams and an analysis of historical betting odds. I was inspired by fivethirtyeight and their modelling efforts but my models are 100% developed and maintained by me. Elo is a simple concept that anyone can use.

I run three separate models. One each for the National Football League, Canadian Football League and College Football FBS games. These are 64% accurate in predicting the winners of NFL or CFL games and 72% accurate for college. From a prediction point of view, these are mostly just simple Elo models with a few, bespoke, subtle quirks. The real trick comes in determining when to bet, this was the hardest part to evaluate and I'm always looking to enhance this aspect.

I am from the UK and don't really understand American Football at all. I followed the NFL last season for the first time ever whilst developing these models so I pretty much understand the rules but can't hold any sort of intelligent conversation around strategy and the nuances of the game. This does not matter though as the modelling is independent of my personal knowledge of what is going on.

What I’ll be doing on this blog going forward:

• When I talk about ‘edge’ I specifically mean the difference between how my Elo model rates a team’s chances of winning and the equivalent chance the bookie’s odds give them. A positive edge means the odds give them less of a chance that I do, negative edge means the odds have them at a higher probability of winning than I do. There are probably other terms I might use in a slightly non-standard way but can’t think of them just now.
• I will not be posting bets before the games happen, so you will you just have to trust me when I post them retrospectively. Perhaps I will just state in advance if I am betting on anything and how many bets there are per week but I am not going to guarantee I will do that as sometimes bets only become viable just before a game starts and I’ll be more interested in monitoring the lines moving than posting something on the internet that no one will read.
• I will not be posting many specific numbers involved (except maybe actual odds). I have put a lot of hours into creating these models, there is no real gain in me giving anything away for free. One thing you might notice is the low volume of bets. Bookmakers are good at their job (or they’re broke and out of business) so the odds tend to match up to my models pretty closely. This means there are not that many good betting spots to be found. To date, there have been 31 games in the CFL this year and I have placed 3 bets (all winners btw, not to spoil a forthcoming post…).
• I will post my predictions, in terms of winners, before the games start. Well...I might do, if I get around to doing it in time.