Pirateboy's NCAAF ML Dogs Experiment - W9

[pirateboy]

Thursday through Saturday 7-12 +.6u

I actually think this week was a bad week for the experiment, and I'm still up on it. I think UCF is a play tonight, but waiting til closer to gametime.

[kdog]

[ QUOTE ]
I actually think this week was a bad week for the experiment, and I'm still up on it. I think UCF is a play tonight, but waiting til closer to gametime

[/ QUOTE ]

UCF became a non play for this system when the line moved to +3.

[Nostrawho?]

Pirateboy,

Austiger asked earlier in the thread if you'd be applying the same principles to favorites who were below the break even ML but I think it got caught up in the rest of the questions going on. Anyway I think its an important point seeing that it should be an exact reflection of what you're doing as long as you're following the same principles. I know you could start another thread on this but I am just reitterating Austiger's question from earlier. Will you be trying this also?

[MyTurn2Raise]

[ QUOTE ]
[ QUOTE ]
I actually think this week was a bad week for the experiment, and I'm still up on it. I think UCF is a play tonight, but waiting til closer to gametime

[/ QUOTE ]

UCF became a non play for this system when the line moved to +3.

[/ QUOTE ]

no...it was still a play
Pirateboy is working off a more refined data set now and UCF was still a clear play tonight

[kdog]

[ QUOTE ]
no...it was still a play
Pirateboy is working off a more refined data set now and UCF was still a clear play tonight

[/ QUOTE ]

Mind shooting me a copy of this? I bet UCF at +3 about an hour before KO and ML was +135 which is not even close to the BE's we have been talking about.

[thelyingthief]

re. range of pointspread groupings:

to begin with, are they really that broad? the difference between 4.5 and 6.5 is an extra point, basically. you'd need to factor in the relative frequencies of the various data points if you want to seriously quantify differences between them, else you may be making errors in your refinement that are greater than that posed by the ranges selected by the author. you could end up compounding the impact of -EV events by assigning them a valence too great, given the universal set of which they are a part. in a general way, when attempting to quantify data of this kind, it is always better to err on the side of the less precise and more robust than your instincts might think. i bet another sport that is heavily stat based, and i have found it difficult, when creating ranges like this, to quantify overly nicely. you kind of have to accept fringe events, since there is insufficient evidence with which to make calculations that will accurately represent them. by that i mean, with a given range, say you have sufficient data to make a strong correlation, 99% confidence; when assigning value to the exceptions, or individual points in the data set, you will have nowhere near that kind of confidence, and in application will experience volatility in your results.

besides, the author may have examined this issue prior to publication, and found it insignificant. why don't you contact him and ask what the reasoning behind his separation of the data and if he considered the problems involved in making them so wide?

tlt

[kdog]

Well, as you've seen, the original groupings are set around the key numbers. 1-2.5, 3-6.5, 7-9.5, etc. So other than the 4 it's really not too bad. And truthfully I can't think of a better way to break it down and still have enough data points for relevance.

I've compared my breakevens to two weeks of ML's now and find most all to be quite close. Since I agree with your statement "in a general way, when attempting to quantify data of this kind, it is always better to err on the side of the less precise and more robust than your instincts might think." (Proceed with caution in other words.)I'm looking for a minimum difference between ML and my breakeven of greater than 5 before I place a bet. That seems like a good place to start.

[pirateboy]

I'm going to keep all my picks in this one thread, as to not clutter.

As MT2R said, I have a very refined data set, so those original groupings have now become halfpoint specific. UCF was a play, but I couldn't get to a comp fast enough, and I didn't want to post a play I didn't make.

That being said, our first Week 10 play is tonight:

Virginia Tech +120 (about a 4.5 cent edge)

[pirateboy]

Saturday plays with edge in cents in parenthesis

Rutgers EV (2)
Cincinnati +195 (14)
Maryland +125 (9)
NC State +345 (21)
Duke +650 (115)
Navy +140 (19)
Alabama +240 (12)
Arizona State +250 (10)
Buffalo +215 (2)
Michigan State +160 (10)
Idaho +140 (5)
New Mexico +145 (18)
Stanford +140 (5)
EMU +260 (5)
UAB +355 (69)
Florida State +235 (5)
San Diego State +155 (17)

Like I said, that's the edge in cents on a dollar.

[Jim T]

[ QUOTE ]
Saturday plays with edge in cents in parenthesis

Rutgers EV (2)
Cincinnati +195 (14)
Maryland +125 (9)
NC State +345 (21)
Duke +650 (115)
Navy +140 (19)
Alabama +240 (12)
Arizona State +250 (10)
Buffalo +215 (2)
Michigan State +160 (10)
Idaho +140 (5)
New Mexico +145 (18)
Stanford +140 (5)
EMU +260 (5)
UAB +355 (69)
Florida State +235 (5)
San Diego State +155 (17)

Like I said, that's the edge in cents on a dollar.

[/ QUOTE ]

I thought that this was based on road dogs, which would make more sense given the historical percentages - yet you have Duke and Alabama which are both at home.