10/29, 1pm ET: Hockey Prop Updates
A slow Monday sports-wise, but a busy Monday for those of us in the content-production business, as we are all grinding to create stuff for you to read the rest of the week. That being said, I wanted to give an update on how the hockey SOG model is doing, and what other things we may be able to glean from it’s output. I’ve ran it every day since the last post (insert Bill Belichick “no days off” soundbyte here). I did, however, stop betting the total shot over/unders, because not only were they essentially a 0 ROI (neither positive nor negative) through about 2 weeks, but the percentage edges have actually gone down even more (rarely would there be one over 3%) where now it’s less than 1% in a lot of cases, and no edges to be had anywhere.
The surprising thing is that as those total shot over/unders have become “tighter” or more accurately assembled, the differences between the teams are still not being represented properly. The team vs. team SOG portion of the model is now 47-30 with a 23.66% ROI. Everything about what’s happening in that market is being pretty accurately predicted by the model, including all the larger plays winning with greater frequency, etc. etc. A couple other people have begun betting into some of these (maybe because of me, maybe not, who knows, there’s like no money in these pools) but honestly there isn’t any long-term potential in this market specifically so whatever. Everyone can have it. Even if I ruined my own edge. I would be limited anyway if I bet the limit on these that much longer, or at least I think I would. By the way, tonight, there’s a sub-5-percent edge in Calgary-Toronto (aka no bet), and a 6.61% edge in favor of Vancouver +4.5, except someone hit it the second it got put up and now it’s -135. You can still bet it small and that’s fine, you’re getting the best of it. It’s just not by much.
The really intriguing thing is to consider some extrapolation of this success…i.e., if this is good, and these numbers are great, what else can they accurately be applied to in order to win money? There’s basically no scientific evidence that more shots leads to either more goals or more winning, so the overall game markets are not a good use of this shot-on-goal information. Shots almost occur separately from the determining of good teams and bad teams. However, goalie saves is offered in a lot of places (Bovada and BetOnline being 2 of the biggest) and it’s fairly easy to take this shot-on-goal information, divide the posted game over/under (which should be efficient) in half, and apply half of those goals to one team and half to the other to get save predictions. Here, let’s do that in the first game that has no edge so you can see what I mean.
Calgary (29.68 projected SOG) vs. Toronto (34.02 projected SOG)
Over/Under for game at 5Dimes: 6
Calgary’s SOG (29.68) minus Calgary’s half of the total (3) = 26.68 projected saves for Toronto’s goaltender tonight.
You could use the team-totals if you want, but they’re juiced pretty heavily one way or the other so it’s almost like you’re using half the over/under anyway. In cases where a team is heavily favored, I’m not sure how underdogs do in the NHL, so I might rather just use the over/under anyway and cut it in half. Then I’m not really taking a position on who scores those (6) goals.
I’ve done the goalie saves experiment for 5 days, betting any save # that was off by 2 or more for a flat 1 unit, and wouldn’t you know it, it works there too (because really, why wouldn’t it if the underlying information is sound?) The goalie saves portion of the model is 20-10 with a 22.25% ROI so far. Maybe this is all a house of cards and VERY limited trials, but it’s still fun to see what’s what. During certain weekdays when I want to write this all up, I’ll give out the full projected SOG for everyone (maybe on busy hockey nights like Thursdays where there isn’t much else going on), and then if your book has saves instead of SOG you can just bet that instead. Or maybe this will fail miserably.
P.S. For the player shots-on-goal, I’m still playing around with a lot of different criteria. I think I have a methodology I really like (taking historical 2/3/4 year averages for players, as well as projections from reputable media, and compare them to how players are doing this season, looking for outliers) but I’d like to test it more and set some more parameters for what is a bet, and what is not, before giving those out. Such a methodology would also be applicable to “will ___ score a goal tonight” and “will ___ score a point tonight” if I can get it to all look right. Fun stuff.