Well it sure has been a long time, hasn't it? After the hot dog eating contest there was a good month or so of basically no action for me, almost at all. It was probably the smallest volume month I've had in like a decade. All the free money Covid stuff, the stock market bets and the horse matchups and video games and the hot dog eating contest and the NFL Draft even, all that stuff either came and went or I lost access to betting on them. Things have picked up ever since the NHL playoffs started, and now especially with football in full swing. But before we get into what I'm up to now, let's do a little house keeping on the hot dogs and a little politics.
Joey Chestnut won the contest, of course. It wasn't even remotely close. I got that at -850 which was somehow still a steal. That won me .75 units. Miki Sudo DID win the womens group, however, and that cost me .5u. Both of them went over their totals (but not in my middle range) which netted me 1.93. So a final win of 2.18 units. Really not bad considering the amount of time and "research" this took. I wish I kept better records early on, I would love to know how much I'm up lifetime in the July 4th hot dog eating contest. Every year I just bet on Chestnut and usually the over, and usually have arbs too. One-off events that draw a ton of attention like that and the Draft and the Super Bowl always offer all kinds of opportunities.
The other development is the VP was picked! Kamala Harris! If you've been reading along, you know I've been following this market especially close. And even though my initial prediction was wrong, I feel like I was pretty on top of what happened. The Dems threw out a lot of names and nothing really stuck, so they kind of backed up into Kamala. No one can argue that Kamala was the clear choice all along, it was obvious they were looking at and floating the names of a lot of women. After Klobuchar became radioactive, it felt like a long, meandering walk to begrudgingly pick Kamala.
So how did we do? Well, pretty good it turns out. On paper at least. My final position on Kamala was 12u to win 20, which won. I had 12.8 units on everyone else, so I ended up plus 7.2 units in the VP market. I say 'on paper' because I didn't end up collecting all of that. A couple of my later Kamala bets got "deleted" before they settled, which is just awesome. It's the cost of business though I suppose. When you're a AAA ball player hitting off a tee, you can't cry when the other team goes home. Or something.
They also finally officially nominated Biden for the Democratic nominee, so those bets all settled. That was a win of 12.5 units. So I've scooped up a nice 19.7 (paper) units on politics and 2.18 on hot dogs since we last spoke. Since I decided to include stiffs and "deletions" in the official record (my reasoning being that the risk of not getting paid on winnings must be factored into any official record I have due to the nature of the markets I bet on), the official politics win was only 13 units. 15.18 additional units total since the last count.
So what now? I've added some political bets in the past couple months. A few states and I've added to the House and Senate positions as well as some more on Trump vs Biden. I need to update my own records though, once I do that I'll make a post with everything I have pending.
What I've really been focusing on though is NHL player shots on goal bets. I have been on fire with them this post-season, too. It sucks I can barely get anything down on them, and they take FOREVER to do, but it's been worth it. Up almost 40 units since early August. It's due to a combination of A) these props always being generally very soft B) one of my books that takes them has no idea what theyre doing C) my model is legit and D) maybe running a little good? Although to be honest I don't think that's true. I can't even get down a full unit on these plays most of the time. It's grinding out pieces of units every bet...that's an honest 40 units if you ask me.
This model started as an off-shoot to one of my very first models. It was for NHL player vs player total points. Compared to what I have now, it was very simple but it had a certain clever elegance to it. To this day, NHL player vs player props is one of my very best prop sub sets and it was profitable right out of the gate.
The basic logic behind it was to use a players Corsi per game for the current season as a starting point. I'd take this number as his 'true talent' level and assume any change in this number was due to an increase in actual skill, not random variance. Then I'd use some some combination of his career numbers and his last 3 years of "corsi into goals". There was never really an official name for this stat, but basically a players shooting percentage but using shots AT the net instead of shots ON the net (corsi instead of shots). Like shooting percentage, but even more so, that number fluctuates wildly and has a big impact on a players performance. It can be very fluky, even for an entire season. A player's 'corsi per game', or his shot attempts per game, however, is not so fluky. A sudden increase in a players corsi per game is much, much more likely to be the result of an increase of actual skill rather than a fluke, as opposed to how often his converts those chances into goals. (Or he's playing with better line mates which is fine too.) So I'd use current season numbers for Corsi per game, career/L3 years numbers for 'corsi into goals', adjust for the opponent, factor scoring effects in a little bit, and I'd get a pretty good number for the players expected goals.
Then for assists, I would assume each team would get the average number assists that their expected goal total would suggest, give a bump to primary assists and a ding for secondary assists, and bang. Add em up and you get expected points. Then use poisson to convert them into percentages against each other, which you then convert back into betting odds and you got yourself a model.
That was the original one that got this new one started. I eventually got kicked off of every account that offered head to head matchups, so I had to adjust. First thing I tried was just using the model to bet at a new account that offered over-unders on players goals and assists. This did not work for assists, and in the old format, didn't work for goals either. I figured out that the model was really only good at matching up players and finding ones who were running good or bad. The actual numbers they were spitting out weren't all that accurate (this new book is very sharp, too. Its a real-deal one, not a PPH mickey mouse club member.)
I then started looking at shots on goal. This new account offered them and I thought if I took the same logic as before but improved on it and applied it to shots instead of goals, I might have something. I scrapped the old model completely and started from scratch. I made so many improvements: I started using 'per minute' instead of per game and came up with a really good way to predict power play time for each team and player. I think that was probably the biggest improvement, as well as more accurately adjusting for the opponent. Lots of other things too that all add up and have come together to make a model I am most proud of. Of course it's in a market that almost no one offers and when they do it's for a couple months a year and you can only bet like 1 or 200 hundred on, and if you have a good week or two you'll get noticed like a sore thumb but hey, a profitable model is a profitable model, damnit!
I would love to post my plays on here for a few games but it probably won't happen. They take forever to do and I'm always firing right up until the game starts. I will try though. If I do you best get down on em. It's not even so much that the edges are huge, it's just deadly accurate. I'm fine with betting stuff only 5 cents off of my expected number which is a pretty low margin of error. I feel like I need to push every edge I have in this market though since it'll be gone soon.
That's it for now, expect another update relatively soon with some football stuff.
