Computer Picks: January 15th—Recovering from System Crash

Updated: January 15, 2010 at 11:13 am by Jeff Tay

When I wrote in the opening section of yesterday’s edition of Computer Picks that Thursday was going to be a “tough test” for the system, I didn’t completely anticipate how ugly it was going to get. Then again, I was still basking in the glow of momentary success, and heedlessly entertaining fanciful thoughts that a system to picking NHL winners had just been perfected, and was at my disposal.

After Thursday night’s generous serving of humble pie brought me back down to earth though, I came to realize two very important lessons. The first of these two lessons was that, obviously, the system had not been perfected—but furthermore, that it was crucially important to analyze where the system was flawed and to try and address its issues. After all, how often do things go perfect on the first attempt? It’s all about the tweaks. The second lesson that I came to realize then, was related to the first lesson, in that I had actually overlooked one very important issue which made the system flawed; I was not factoring enough into the equation the level of motivation that specific teams would have for specific games. This lesson initially came in the form of a Thursday night upset of the Flyers by the Leafs.

Not long after the puck was dropped in this game, those who were watching were bearing witness to a highly-motivated Leafs team (looking to snap a 4-game losing streak and avenge a recent 6-2 loss) really pushing the pace against a Flyers team that was playing with a level of intensity equivalent to that of their pre-game skate. Despite how well the Flyers had been playing, and despite their noticeable edge in talent, motivation was the difference in this game, and it was more than enough to push the Leafs to an easy victory. Seeing that the system had picked Philadelphia to win this game 68% of the time, and then seeing that they barely even bothered to show up for it, a change in the system was in order.

After the night was over, I added additional stats for the computer to take into account—stats related to how teams played during corresponding losing/winning streaks. I also made some modifications so that effects of extremely irregular games (unusually high or low scoring or very lopsided games) would be mitigated in the calculations. After making these changes, I then had the computer make a prediction for Philadelphia vs. Toronto again, and the projected result afterwards was PHI 2 – TOR 2, with Toronto actually winning 52% of the time. Needless to say, I felt pretty positive about these changes (for the moment at least).

So, just how ugly was the rest of the day? Here are the numbers (look away if you are squeamish):

Correct Predictions on Winning Side: BUF, DET, FLA, MTL, STL, CHI, and PIT (7 out of 12)

Correct Moneyline Recommendations: BUF, FLA, PHX, PIT, and BOS (5 out of 11 for a losing night)

Correct Predictions on Over/Under: FLA/TBL u5.5, DAL/MTL o5.5, BOS/SJS u5.0, ANA/LAK u5.5 (4 out of 12)

Overall Record:

Predictions on Winning Sides: 14-11
Predictions on Over/Under: 13-11-1

It looks like Friday will be the debut of Computer Picks Ver. 1.1 (the one with the new tweaks). Here we go. Ready and launch:

Game #1: Toronto Maple Leafs 2 – Washington Capitals 7 (TOR 19% – WSH 81% to win)

Key Stats: I can’t believe I’m even seeing this after the changes that I just made to the system, and I’m almost embarrassed to publish it. One of the key contributors to this outlandish prediction though, is the fact that the Leafs haven’t won 2 consecutive games since December 14th, and their last 2 road games following wins have been a 6-2 loss to Philadelphia and a 3-1 loss to Edmonton. They’ve also given up 5 goals in each of their last 2 road back ends of back-to-backs where they played 3 games in 4 nights (to Boston and Montreal). Meanwhile, this also appears to be Washington’s prime situation. They’ve excelled at home when returning from road trips, and they’ve also done very well to follow wins with wins, and they tagged Florida for 6 goals the last time they returned home from a close win. Furthermore, Washington has also beaten the Leafs once at home this season by a score of 6-4.

Best Bet: Given how questionable Washington’s defense and goaltending have been lately, I certainly see this game having the potential to go Over 6.5. It is a big number though, but I wouldn’t want to test the under on it. As for the 7-2 projected score, I see it being a more modest 6-2 final. Actually, I do have my concerns about how many goals Washington allowed against Florida and Tampa Bay, and am not sure if that will carry over to their home stand here. Taking a shot with Washington -1.5 goals does appear to be more than within reason though.

Game#2: Nashville Predators 3 – Calgary Flames 2 (NSH 63% – CGY 37% to win)

Key Stats: This appears to be very much a situational prediction. Nashville has already won a game in Calgary this season, and has had some recent success against low-scoring home teams such as Calgary. Furthermore, they have been excellent road underdogs, showing a tremendous ability to score goals on teams that don’t consider them to be elite-level opponents. Meanwhile, there is a lot that doesn’t favour the spot here for Calgary, including the way that they have tended to decline towards the end of home stands, and the way they have struggled to score at home for a very long while now. They will have revenge on their minds, but they will need to buck some pretty significant trends in order to pull out a victory here.

Best Bet: Not too long ago, Nashville would have been about a +140 underdog here. Times have changed a bit, but there is still good value with a good road team like Nashville at +125. Their road record this season is actually better than Calgary’s home record. Meanwhile, there is also a slight lean towards the Under 5.5 if you fancy that.

Final Output: I think I’m going to be pretty surprised if the system can’t get any of these picks right. Saturday will be the true test though. The system has had its moments (good and bad) so far, and I still think that with the right tweaks it can be a great money-making tool. We’ll see if I still feel the same way after Saturday, I guess.