Game Predictions – 1/28/2006

So far so good with the new prediction algorithm. It went 2 for 2 last night. There are 10 games tonight, I am hoping it will get at least 7 correct. I have also decided, completely for your amusement of course, to place bets on all these predictions to see if it can be used to actually make a profit. I have signed up with Bowmans Sportbook, deposited $100 and will bet $5 on every game as predicted by my prediction algorithm. I’ll keep you updated on how things go.

Since you have predictions for all games now I am just going to try to point out some potential upsets and/or where the betting lines do not agree with the predictions (read: potential favourable bet conditions).

Dallas/Detoit – Dallas is listed as the ‘good’ favourite but I think this one could go either way. The betting line actually favours Detroit (-115) over Dallas (-105).

Phoenix/San Jose – The algorithm lists Phoenix as a bit of a favourite to win but this is one that I think could go San Jose’s way. The Bowman betting line currently has San Jose -135 and Phoenix +115 so betters are siding with San Jose too.

Home Team Road Team Predicted Winner Confidence
Philadelphia Tampa Bay Philadelphia Good
Dallas Detroit Dallas Good
Los Angeles Anaheim Los Angeles Strong
Boston NY Islanders Boston Good
Toronto Montreal Toronto Good
NY Rangers Pittsburgh NY Rangers Strong
Carolina Atlanta Carolina Strong
Columbus Nashville Nashville Good
Phoenix San Jose Phoenix Some
Colorado Vancouver Colorado Good

This article has 16 Comments

  1. Good luck with the games against lines. I’ve been tracking on my own since the 03-04 season with vegas lines and one point of advice I can offer is to stick with laying the same amount ($5, $10, whatever) on every game no matter what.

    As opposed to laying whatever you would need to win $5 (ie, if someone is a -300 favorite, you’d have to wager $15 to win 5), keeping the same amount per game is the best way to go for hockey, in my experience.

    Best of luck.

  2. I have several questions today before I get to my predictions. First, how long does it take you daily to get the results of your algorithm formula? Second, since I am 15 and can’t be in the casinos, I don’t know what the odds mean (-300, +130). Could you please run that across me? Third, what changes did you make to develop your new algorithm? Fourth, how does the new algorithm eliminate pick-’em games? Fifth, what is your opinion on the Ducks dropping the “Mighty” in their name next year?

    Now, here are some upsets in my opinion:
    Anaheim/Los Angeles: I pick LA to win because of luck. This is the 6th game I am going to this year, and in all of them, the Kings lost the last game at home, and at the game I went to, the Kings won. This is the same reason as to why I picked Edmonton to beat the Kings Thursday.

    Atlanta/Carolina: Carolina is one of the best teams in the NHL, and Atlanta has done ok recently. They are finally breaking out of that season-long slump and could be an upset possibility.

    Before I get to my predictions, I need to say that I have also changed how I do my predictions. The stats are all the same, but now my “Strong” predictions have been changed to games which I feel I am 75% confident in. The “Good” predictions are those which are 60% confident, and “Some” predictions are basically a coin-flip, or 50%. The standings remain the same, so don’t worry about it.

    75% confidence: Rangers over Penguins, Predators over BlueJackets
    60% confidence: Kings over Mighty Ducks, Hurricanes over Thrashers, Bruins over Islanders, Canadiens over Maple Leafs, Sharks over Coyotes
    50% confidence: Stars over Red Wings, Flyers over Lightning, Avalanche over Canucks

    Overall Record:
    75%: 42-9-5, 79.5%
    60%: 73-43-13, 61.6%
    50%: 53-48-11, 52.2%
    Home: 94-44-16, 66.2%
    Road: 74-56-13, 56.3%
    Overall: 168-100-29, 61.4%

  3. First, how long does it take you daily to get the results of your algorithm formula?

    Very quickly. Once I update my excel sheets and then output them to an ascii file I just run a program I wrote which cranks out the predictions.

    Second, since I am 15 and can’t be in the casinos, I don’t know what the odds mean (-300, +130).

    A + indicates how much you will win on a $100 bet while a – indicates how much you need to bet to win $100. For example, a +130 means if you bet $100 you would get $230 back (your $100 bet plus $130 win). A -130 money line means that you need to bet $130 to win $100. In this case if you bet $130 and won you would get $230 back.

    Third, what changes did you make to develop your new algorithm?

    The biggest change is I am using a much more robust method for accounting for strength of schedule. This is why you find teams in tough divisions rise dramatically and teams in weak divisions fall dramatically in my power rankings. Another change I made is to count shootouts as ties when computing starting winning percentages because I don’t feel shootouts are really that indicitave of how good a team is or its ability to win.

    Fourth, how does the new algorithm eliminate pick-’em games?

    Basically what I do is crank out a number which if positive indicates a home team win and if negative indicates a road team win. Then I say that really big numbers become strong, moderately big numbers become good, and smaller numbers become some confidence levels. Under the old system I say if the numbers were close to zero they were too close to call. Now I am saying that since the algorithm is more robust and allows for better prediction I eliminate that close to zero “too close to call” zone.

    Fifth, what is your opinion on the Ducks dropping the “Mighty” in their name next year?

    Looking at where they are sitting in the standings I think they realized that calling themselves “Mighty” is kind of misleading. That said, Anaheim Ducks or Ducks of Anaheim sounds kind of dumb too.

  4. Carolina has been so good and Atlanta hasn’t been scoring as well as they are able to so you have to really like the Hurricanes chances. They should win.

  5. Thanks David. Another thing, why is Nashville only listed as a ‘good’ favorite? Nashville is clearly the better team here and they’ve won the past 4 meetings. Also, Vokoun is coming off of back-to-back wins against #1 Detroit…why isn’t this a strong pick?

  6. I can’t say for sure because that is just what the prediction algorithm predicted but Columbus being at home is definitely a factor. Plus it doesn’t look at short term influences like Nashville beating Detroit twice in a row. Besides, Columbus has won 3 of 4 and 7 of 10.

  7. Perhaps the algorithm is a bit flawed then? The fact that it doesn’t take into effect how teams have been playing recently could be a bit misleading. I mean, what if a team has one or two key players injured? That would seriously affect a team’s ability to win and it wouldn’t even be looked at by your algorithm. For example, what if Hasek injured himself and was out for the rest of the season? The Sens would clearly be in trouble but the algorithm would overlook this and probably still think they were just as strong. Another thing is winning streaks and player streaks; these are both things that would happen “recently” and wouldn’t be taken into effect by the algorithm since it only looks at the entire season and not what’s happening recently.

    Just a few things I thought I might run by…thanks again for the input David, great work.

  8. Another example is in today’s game involving the Kings/Ducks. The program probably didn’t look at the fact that the Kings are still without their leading scorer Pavol Demitra and their recent poor play. This is probably why you listed the Kings as the strong favorites, but they’re getting mangled 6-1 at home. Again, these factors are just too important to be overlooked.

  9. CORRECTION! The Kings only lost 6-2. Pavol Demitra did get an assist. Also, I think Carolina should be a good favourite to win because, prior to the game, they retired Ron Francis’ jersey with a 30-minute celebration. This is the kind of thing that motivates a team to win. Montreal retired two jerseys in the same game and won, Edmonton retired somebody and won, and look what happened when MSG retired Mark Messier. They won in overtime, so I think Carolina will keep the trend going. Also, I think that from now on, I will keep track of bets based on my predictions just for fun, starting with today’s games (if I can get the odds).

  10. Also, I need some help finding out how much I win or lose for each game that I bet $100 in. (ie: I pick NY, -380 odds, so how much do I win? I bet on LA, -155, so how much do I lose? I pick Dallas, Even, so do I win 100? I just don’t understand how much you get or lose in certain scenarios. Please help me!

  11. John: It is very difficult to factor in injuries and stuff like that into an automated prediction algorithm. I’ve have looked at things like winning/losing streaks and what not but haven’t found any reliable improvements in the success rate. Also, as I write this the Blue Jackets and Predators are tied 3-3 heading into the third.

    Kevin: It takes a bit of math. If New York is -380 that means it takes $380 to win $100. If you bet $10 you would win $10 * 100/380 or $2.63.

  12. So the formula is ($bet*100)/(odds)? Also, if you bet 100 on a team with a +125 and win, you get $125? This is much easier than I thought! I can’t wait to turn 21 so I can put this to use.

  13. That’s why sometimes when you make a prediction you can’t solely rely on that algorithm. I think taking injuries and streaks into account is something that just makes sense and would proabably even improve your success rate. In any case, it looks like your program is doing just fine on its own. Good luck on those bets!

  14. Using the odds I found on, and betting $100 on each team I predicted, I have lost $296.05 today, which is a lot lower than what I would’ve liked. With three games tomorrow, I hope I can do better.

Comments are closed.