Sygamel's 2005 NFL Regular Season Picks Log

[DougOzzzz]

[ QUOTE ]
[ QUOTE ]
dont forget the more you bet, the less mistakes you make also.

as long as you can deal with times like this, you will be ok.

you seem to have a decent grasp on handicapping (your picks usually fall in line with the guys i follow) and i personally think your betting teschniques could use some tuning up, but time will tell.

cant wait for another blood bath this week [img]/images/graemlins/smile.gif[/img]

[/ QUOTE ]

Multicollinearity, i.e. the ability of two similar trends to predict the same result, thereby artificially inflating the win probability of a specific team, is the single biggest problem in trend systems. Before every season, I will spend hours upon hours making sure I've eliminated its influence by going through all my trends to ensure no multicollinearity exists. Because so many of my trends favor underdogs, this problem can become particularly acute during periods like the last 9 weeks. I am certain, however, that I covered all the bases before the season but will go over them again all the same.

If it's a question of pedigree, i.e. the problem is the person making the system and whether it's truly valuable and mathematically sound, I scored a 720 on the GRE analytical section and 800 on the GRE math section 3 years ago. As I see it, the problem that exists within the system is the potential over-reliance on trends -- through the first 130 games of this pro football season, it is of this that I am primarily, and probably entirely, guilty. I will need to correct this in some fashion.

[/ QUOTE ]

What I find interesting about your trend system is that somehow you use it to predict what the "true" spread should be. To me it seems not too difficult to evaluate several trends and say that Team X is likely to cover - however I have no idea how you translate that into an actual evaluation of the true point spread in a game.

I always thought that the biggest problem with trend systems was that many trends may simply be the result of random variation. Many stats I see thrown out by people trying to look smart tend to fall in this category, IMO. It could be that you are failing to properly regress each trend to the mean. Then again, it sounds like you have more experience with this than I.

It does seem though that your system too easily evaluates a game as a 3-5 unit bet. Your picks agree with my sources most of the time, but I am rarely ever told to put more than 2 units on a game. I pretty much maxed out on the Pats last night though, going with 2 units on the cover, 1 on 1H +3 EV, and 1 on the money line.

I am curious whether you track the success of your different grade of picks. If your 1 unit bets come through at a 55% clip, then your 5 unit bets should be coming through close to 2/3rds of the time. I'm sure you know this, though.

Anyway, hopefully things will turn around. The squares have to be feeling really good about their favorites now, and they have some money to boot. Hopefully this means the lines will be especially inflated and we can clean up the next few weeks.