Quick explanation:
I keep weighted power ratings which are updated daily and put into a formula (that I'm trying to improve all the time) to get an estimated moneyline on each game. Expressing that moneyline in terms of each team's chances to win (example, a -150 favorite has a 60.0% chance to win), I compare my ML with the real ML. (To get the real ML I average Pinny's 2 moneylines -- example, a 130/140 ML says the favorite is 57.43% likely to win, which translates to a shade above ML -135).
The recordkeeping is based on the closing Pinny ML as best I can track it. I compare that with my ML. So if it closes at Pinny at that 130/140 and I give that favorite only a 51.6% chance to win, then ML-OFF system pick is the dog at a strength of 5.8% (57.43 minus 51.60), which is rounded to 6.
For the first half-month of the season, if we played every team where the ML-OFF was 5% or more, we'd have 71 plays, which went 31-40 straight up and 41-30 ATS and made a profit either way.
Again in this Block I, as in the testing I did with last year's results, the concept of plays being "too good" -- or my ML being "too far off" -- reared its head. So far in 2006-07, at 15% ML-OFF or higher, the results take a turn for the worse. Just playing the plays in the 5%-15% range would have gone 26-29 SU for +23.8 units, with ATS plays 34-21 (or about +11 units if all plays were 1 unit and juice was -110.)
This is promising, although on a small sample. It's a less-than-sexy approach in that you are asked to be on almost every big dog, and on a lot of unappealing teams. For example, so far it's featured these teams most:
POR 9x
MEM 7x
NY 6x
CHA, BOS 5x
ATL 4x
Also, because the power ratings are based on the last 10 games, these early results are still somewhat connected (on a decreasing basis) to part of last year's results. (I divided last season into 10 chunks to make up the "last 10 games" while the system gets started.) In the next couple of weeks the ratings will get to be all about this year.
Next post will have the summary, and the third post the daily data.
I keep weighted power ratings which are updated daily and put into a formula (that I'm trying to improve all the time) to get an estimated moneyline on each game. Expressing that moneyline in terms of each team's chances to win (example, a -150 favorite has a 60.0% chance to win), I compare my ML with the real ML. (To get the real ML I average Pinny's 2 moneylines -- example, a 130/140 ML says the favorite is 57.43% likely to win, which translates to a shade above ML -135).
The recordkeeping is based on the closing Pinny ML as best I can track it. I compare that with my ML. So if it closes at Pinny at that 130/140 and I give that favorite only a 51.6% chance to win, then ML-OFF system pick is the dog at a strength of 5.8% (57.43 minus 51.60), which is rounded to 6.
For the first half-month of the season, if we played every team where the ML-OFF was 5% or more, we'd have 71 plays, which went 31-40 straight up and 41-30 ATS and made a profit either way.
Again in this Block I, as in the testing I did with last year's results, the concept of plays being "too good" -- or my ML being "too far off" -- reared its head. So far in 2006-07, at 15% ML-OFF or higher, the results take a turn for the worse. Just playing the plays in the 5%-15% range would have gone 26-29 SU for +23.8 units, with ATS plays 34-21 (or about +11 units if all plays were 1 unit and juice was -110.)
This is promising, although on a small sample. It's a less-than-sexy approach in that you are asked to be on almost every big dog, and on a lot of unappealing teams. For example, so far it's featured these teams most:
POR 9x
MEM 7x
NY 6x
CHA, BOS 5x
ATL 4x
Also, because the power ratings are based on the last 10 games, these early results are still somewhat connected (on a decreasing basis) to part of last year's results. (I divided last season into 10 chunks to make up the "last 10 games" while the system gets started.) In the next couple of weeks the ratings will get to be all about this year.
Next post will have the summary, and the third post the daily data.