2008 CFL MiniTracker

2008 CFL MINI-TRACKER Year-End Summary

--------------------------------------

Statistics were kept for W-L Straight Up (SU), W-L Against The Spread (ATS) and the average number of points each prediction was off from the actual result (MEAN ERROR). e.g.HOM was picked to win by 2 & VIS won by 3 = 5 pts. off. There are 3 sets of statistics: 1st Half (Weeks 1-9), 2nd Half (Weeks 10-22) and Full Season (Weeks 1-22). Week 22 was the Grey Cup. In weeks 1-9 statistics, different skills are needed to create the starting ratings and predict the early weeks of the season. Used for starting ratings can be last year's data, Pre-season results, predictions in publications, Sportsbooks' number of wins each team will have for the season (Over/Under Betting), manual ratings (which might consider roster/coaching changes), or some mystical schemes. Any, some, or all of the above are used. What proportions (weighted values) will be used for the combinations? What should be the initial Home Field Advantage? Then during the early weeks, what larger adjustments are made to the ratings to minimize the effects of the starting ratings? What is the declining slope of the adjustments each week? How many weeks before all portions of the starting ratings are phased out? Separate awards for the 1st half of the season gives recognition to those who have the best grasp of these special talents.

The Participants:

1. The early line (LINE) is the 1st odds I could find; usually Sportsbook CRIS on Monday afternoons. I do not update the Line except to change whole numbers up/down by only a 0.5 point, if the line changes in the next day or so. This eliminates a point spread = to the actual score. In the same way systems that predict in whole numbers are affected.
2. My system; Enhance Spread Projections (ESP).
3. The Sports Network (TSN); Chris Schultz makes the picks. He does not use a computer. I wanted to see how well this top CFL authority competed against the LINE and the computers. He makes his picks early in the week, so he does not have the benefit of the updated injury reports.
4. STAT FOX (FOX) from the NFL & NCAA Trackers.
5. Eric Hollobaugh's System (HOLL) from the NFL Tracker.
6. Kenneth Massey Ratings (MASS); His system produced its own predictions in 2008. In prior years I devised a method of converting his ratings to point spreads.
7. Steven Jens' System (JENS) from the NFL Tracker.
8. (RANK) This is where I generated Ratings from Rankings. I used "experts" from various websites who rank the teams from 1 to 8 each week. I gave the top team 10 points & scaled down to 3 for the 8th team. THE SCORE provided 6 individuals' Rankings and their totaled consensus by team. This was easily converted to Ratings in decimal format. For the other half of each team's Rating, I combined rankings from members of TSN, SPORTS NET, and the Winnipeg Free Press. I totaled & averaged them into decimal Ratings. I then adjusted each team's Rating slightly, + or -, using their Wins-Loses X 20%. Point spread predictions = Ratings +3.5 points for the Home Team. * The ATS for the LINE below is the record of the Favorites VS its line. W = the Favorite covered the point spread. L = the Underdog won with the points or outright.


               
                    WEEKS 1-9 
            SU                ATS         MEAN            
     W - L-T  PCT PLC  W - L-T  PCT PLC  ERROR PLC 
 
LINE 17-15-0 .531 (6)  15-17-0 .469 (*)  11.11 (4)   
ESP  18-14-0 .563 (3)  17-14-1 .548 (3)  10.98 (3)     
TSN  18-14-0 .563 (3)  17-15-0 .531 (5)  10.81 (2)  
FOX  15-17-0 .469 (8)  17-14-1 .548 (3)  11.13 (5)  
HOLL 20-12-0 .625 (2)  18-14-0 .563 (2)  12.00 (8)
JENS 18-14-0 .563 (3)  17-15-0 .531 (5)  11.56 (7) 
MASS 21-11-0 .656 (1)  16-16-0 .500 (7)  10.78 (1) 
RANK 16-16-0 .500 (7)  19-13-0 .593 (1)  11.50 (6)  
          
                    WEEKS 10-22
                 
LINE 32-13-0 .711 (2)  25-20-0 .556 (*)  11.89 (4)    
ESP  30-15-0 .667 (6)  23-22-0 .511 (6)  12.76 (7)    
TSN  31-14-0 .689 (5)  16-29-0 .356 (7)  13.84 (8)   
FOX  32-13-0 .711 (2)  31-14-0 .689 (1)  11.62 (2)      
HOLL 29-16-0 .644 (7)  27-18-0 .600 (3)  11.94 (5)
JENS 33-12-0 .733 (1)  25-20-0 .556 (4)  11.79 (3)  
MASS 32-13-0 .711 (2)  29-15-1 .659 (2)  11.36 (1) 
RANK 28-17-0 .622 (8)  25-20-0 .556 (4)  12.54 (6)  
 
                    WEEKS 1-22
 
LINE 49-28-0 .636 (3)  40-37-0 .519 (*)  11.56 (3)  
ESP  48-29-0 .623 (6)  40-36-1 .526 (6)  12.02 (6)      
TSN  49-28-0 .636 (3)  33-44-0 .429 (7)  12.58 (8) 
FOX  47-30-0 .610 (7)  48-28-1 .632 (1)  11.42 (2) 
HOLL 49-28-0 .636 (3)  45-32-0 .584 (3)  12.00 (5) 
JENS 51-26-0 .662 (2)  42-35-0 .545 (5)  11.69 (4)  
MASS 53-24-0 .688 (1)  45-31-1 .592 (2)  11.12 (1) 
RANK 44-33-0 .571 (8)  44-33-0 .571 (4)  12.11 (7)

I determined the best overall predictor for each of the 3 sets of statistics. I added where they finished in each of the 3 categories. In the 1st Half, MASS & ESP had a total score of 9 for the 3 categories. ESP fell to 2nd place because it tied for 3rd in SU W-L. TSN was a close 3rd with a score of 10. For the 2nd Half, we had an exact tie for 1st between FOX & MASS. Each had a score of 5. They even tied with LINE in SU W-L. JENS was not far behind with a score of 8. MASS dominated the Full Season Stats missing a perfect score of 3 by one. 2nd place went to FOX with a score of 10. For 3rd place, both JENS and HOLL had a score of 11. HOLL dropped to 4th when it shared 3rd place in SU W-L with 2 others. The LINE failed to take 1st place in any category which is unusual. RANK did not do well overall, but it managed to be in 1st ATS W-L in Weeks 1-9. I'm amazed that this system I contrived is competitive. Last year RANK tied for 2nd (missing 1st by 1) as the 2nd Half Overall Predictor. RANK was the the 2nd best Overall Predictor for the Full Season. Of course key to its success is the skill of the Rankers.

Joe Kasulis July 30, 2009