sports picking statistical formula

[fuzion1]

Do we have any number freaks in here? I was wondering if anyone has ever ran across or made a formula for picking winners. I'm positive it could be done. If you have one please share!

[captaincapper]

I've always had pretty good luck with this formula:

a^2+2ab+b^2 = Winner Winner Chicken Dinner

[canadiens11]

2+2 is fouuur
2 plus 2 is fouuur

[gmswerk]

I prefer the Brandon Lang formula.Flip a coin heads =fav. tails =underdog.

[HAMILROCK]

Originally Posted by gmswerk

I prefer the Brandon Lang formula.Flip a coin heads =fav. tails =underdog.

[reggie02]

Number freak right here!

To answer your question, i thought the same thing about 10 years ago. I'm a financial analyst/risk manager by day and hardcore gambler by night. I got sick of losing so started playing with stats and simulators. Basically, over the last 10 years, i've built numerous football models with every possible statistic you can find. I've compiled thousands of games already played and loaded them in models to help predict future outcomes and the bottom line is that it helped me lose less, but there is no magic formula. The models I run now are seriously complexe in that they factor in a team's abilities, their opponents abilities and their opponents' opponents abilities....if you follow me. Basically it's a way of standardizing the schedule or making things relative to who they played.

I've tested pretty much all the stats before I put them in models so that they are statistically significant. So every variable I use helps explain past outcomes, however it isn't as accurrate in predicting the future. I put together some numbers and the best model I have will be accurate within +/- 10 points 68% of the time. The deeper in the seson it gets, the more accurate it becomes because teams stabilize so I usually do very well in bowl seasons. I usually only play the games that have the largest deviations and do pretty well (at least break even most of the time), but sometimes force bets on games i'm watching and tend to lose more often on those.

By the way, I only do it for NCAA football. Last year i came in 35th place in espn bowl picking contest out of probably 50,000 people (29 out of 32 correct). The only advice I can give you is that all my models factor yds/point on defense and offense as the most important followed by home field advantage. I think home field is that high because so many upsets happen on home field and the model has no other place to account for that freak occurance from a statistical perspective. Next is offensive and defensive passing (yds/att) and then rushing off and def (yds/att). All models like offense rather than defense and special teams have never been statistically significant in any model I created, so I don't use them. I also use an offensive line stat that measures sacks allowed, 3rd down rate and interception rate. It doesn't work as well for defensive line, I think because the defense basically reacts to the offense and how potent they are or aren't. One last thing, I have noticed that the non BCS conferences tend to be almost impossible to model correctly due to the fact that they are way more inconsistent. If you're gonna use stats, I encourage you to focus on the better teams and the better conferences. For the weak teams, home field advantage becomes massive as the bad teams often perform poorly away from home. This will save you money, trust me!

For NCAA or pretty much any sport, you have to take into account the schedule strenght that each team plays. In the past, i would just take Sagarin or Anderson/Hester's numbers and multiply each team's stat by the SS number and it worked pretty well. Now I compute my own, but its really too much work.

There you go, statistical modelling 101. Feel free to ask away if you have any questions and good luck.

Reg

[Chicago RED]

Originally Posted by reggie02

Number freak right here!

To answer your question, i thought the same thing about 10 years ago. I'm a financial analyst/risk manager by day and hardcore gambler by night. I got sick of losing so started playing with stats and simulators. Basically, over the last 10 years, i've built numerous football models with every possible statistic you can find. I've compiled thousands of games already played and loaded them in models to help predict future outcomes and the bottom line is that it helped me lose less, but there is no magic formula. The models I run now are seriously complexe in that they factor in a team's abilities, their opponents abilities and their opponents' opponents abilities....if you follow me. Basically it's a way of standardizing the schedule or making things relative to who they played.

I've tested pretty much all the stats before I put them in models so that they are statistically significant. So every variable I use helps explain past outcomes, however it isn't as accurrate in predicting the future. I put together some numbers and the best model I have will be accurate within +/- 10 points 68% of the time. The deeper in the seson it gets, the more accurate it becomes because teams stabilize so I usually do very well in bowl seasons. I usually only play the games that have the largest deviations and do pretty well (at least break even most of the time), but sometimes force bets on games i'm watching and tend to lose more often on those.

By the way, I only do it for NCAA football. Last year i came in 35th place in espn bowl picking contest out of probably 50,000 people (29 out of 32 correct). The only advice I can give you is that all my models factor yds/point on defense and offense as the most important followed by home field advantage. I think home field is that high because so many upsets happen on home field and the model has no other place to account for that freak occurance from a statistical perspective. Next is offensive and defensive passing (yds/att) and then rushing off and def (yds/att). All models like offense rather than defense and special teams have never been statistically significant in any model I created, so I don't use them. I also use an offensive line stat that measures sacks allowed, 3rd down rate and interception rate. It doesn't work as well for defensive line, I think because the defense basically reacts to the offense and how potent they are or aren't. One last thing, I have noticed that the non BCS conferences tend to be almost impossible to model correctly due to the fact that they are way more inconsistent. If you're gonna use stats, I encourage you to focus on the better teams and the better conferences. For the weak teams, home field advantage becomes massive as the bad teams often perform poorly away from home. This will save you money, trust me!

For NCAA or pretty much any sport, you have to take into account the schedule strenght that each team plays. In the past, i would just take Sagarin or Anderson/Hester's numbers and multiply each team's stat by the SS number and it worked pretty well. Now I compute my own, but its really too much work.

There you go, statistical modelling 101. Feel free to ask away if you have any questions and good luck.

Reg

Great Post Reggie, nice work.

[cptkirk321]

Originally Posted by captaincapper

I've always had pretty good luck with this formula:

a^2+2ab+b^2 = Winner Winner Chicken Dinner

Winner Winner Chicken Dinner

[gmswerk]

Originally Posted by HAMILROCK

they only make movies about winners