QUANTUM PICKS PINNACLE MATCHUPS 6/23

[formula_2002]

date track race/dist pgno horse LINE LINW=E PLAY SURFACE POST TIME
6/23/2006 BEL R5 8F 6 NOTHIN B -170 PLAY D 5:00
6/23/2006 BEL R5 8F 7 KIT KAT 154 D 5:00
6/23/2006 BEL R6 8F 5 HOT TRUC -107 PLAY T 5:32
6/23/2006 BEL R6 8F 7 DOVE WIN -109 T 5:32
6/23/2006 BEL R8 7F 6 ELEGANT -105 T 6:36
6/23/2006 BEL R8 7F 7 CLEVER F -111 PLAY T 6:36
6/23/2006 CD R7 6F 4 CASE CHA -110 PLAY D 5:47
6/23/2006 CD R7 6F 8 SLEWS IR -106 D 5:47
6/23/2006 CD R10 8 1 CHIEF EX 149 PLAY D 7:20
6/23/2006 CD R10 8 5 SHADOWLA -165 D 7:20
6/23/2006 CD R11 8 1 CONFLICT -120 PLAY T 7:49
6/23/2006 CD R11 8 3 CAPE AUG 104 T 7:49
6/23/2006 HOL R3 8. 8 CADILLAC -154 T 11:02
6/23/2006 HOL R3 8. 12 WAR GAEL 138 PLAY T 11:02
6/23/2006 HOL R7 8F 1 HOLDTHEH -136 T 12:54
6/23/2006 HOL R7 8F 2 DARK SOR 120 PLAY T 12:54
6/23/2006 HOL R8 7F 3 BEWITCHI -135 D 1:22
6/23/2006 HOL R8 7F 9 FEATURE 119 PLAY D 1:22

[ryesteve]

I can tell this model is very different than the one you've been using. Typically, I agree with every pick you make... it's just a matter of degree. But this time around, I'd probably go the other way on maybe half of these.

[formula_2002]

I can tell this model is very different than the one you've been using. Typically, I agree with every pick you make... it's just a matter of degree. But this time around, I'd probably go the other way on maybe half of these.
Somewhere in the process of setting up this system, I was shocked to find that 80% of the picks were losing...then i discovered that i had the plays reversed, i was playing the underlays. A simple correction got that resolved!!.

I haven't tested the whole thing..Success, or the lack of it will tell.

[ryesteve]

Something sounds a bit off there. Playing the underlays should cost you money, not result in a catastrophic low win%. I guess we'll see how it goes...

[formula_2002]

I'll change the 6th at bel to a "no play" race

[ryesteve]

I'll change the 6th at bel to a "no play" race
Does Pinnacle let you do that? :D

Turns out you had the winner, but given that the margin was about 6 inches, your assessment that it could've gone either way was correct.

[formula_2002]

Does Pinnacle let you do that? :D

Turns out you had the winner, but given that the margin was about 6 inches, your assessment that it could've gone either way was correct.

Rye, you have a point there. I eliminated the play because I re-ran the program after scratches, and it eliminated the 6th race. I seldom do that.
Since I want to get the earliest line possible I guess it should not be eliminated since the bet will be made prior to scratches..Hey it's a work in progress. In about one more month, it maye be a bit more then that.. :)
But then again, when I run the data the final time, with the results xrd file, that play will have been eliminated again.
Still have a bit of work to do before I meld the xrd file into the data base. After that is done,I'll be able to do a bit of statistical analysis

[formula_2002]

Going into HOL night time racing, QUANTUM,
has won 3 of 5 races, $119 profit, $611 played.
(even the canceled play, who's figures are excluded from the above, won.)

Going against the top choice in the 3 remaing races at HOl..tough going..

[ryesteve]

Since I want to get the earliest line possible I guess it should not be eliminated since the bet will be made prior to scratches
From the little I've seen so far, your new model seems a bit contrarian, so you might not necessarily benefit by taking the earliest possible line. I understand the dilemma though. Several times a week I'll discover that scratches have eliminated the perceived value I had at the time I made the bet. I suppose one way to avoid this would've been if my model used raw values instead of ranks, but then again, it might not work as well. Given the success I'm having, recollecting data and refitting the model isn't something I'm motivated to do. The other thing is that my record on those selections that should've been no-plays after scratches, is actually very good.

[formula_2002]

Here are the finals:
4 wins, 8 plays, $57 profit, $911 played.
"E" (expected winner) =3.8

Theoretically, in the long run, if actual winners>=expected winners, the program should make a min 10% profit

[Jeff P]

...Theoretically, in the long run, if actual winners>=expected winners, the program should make a min 10% profit. Joe, I'd say you already have the ability to make a LOT more than that with the selection criteria you already have. I've always maintained that WHEN you bet a matchup is every bit as important as who you bet. Focus - even a little bit - on anticipating and exploiting line movement and you'll see what I mean.

-jp

.

[formula_2002]

Focus - even a little bit - on anticipating and exploiting line movement and you'll see what I mean.

-jp

.

Jeff, the "anticipating" part is difficult for me, while the "exploiting line movement" is something that I would like to do.

Say, the Pinnacle normalized probability for "A" IS 50% and my normalized probability is 55%, I have a 10% edge.
My odds are .81-1.
Using Kelly criterion, my bet size is 12% of bankroll. (edge/odds)

If Pinnacle's line takes "A" down to a 40 % probability, my edge goes to 37%, which would now call for a bet size of 46% of bankroll.

I would bet both situations.

I'm sure I would start off at no more than 10% kelly, bringing my play down to 1.2% and 4.5% of bankroll.
I would certainly run that though my data base before making any real $ plays :)

[ryesteve]

If Pinnacle's line takes "A" down to a 40 % probability, my edge goes to 37%, which would now call for a bet size of 46% of bankroll.

46%?! That sounds like a ruinous strategy. Have you run any sims on that?

On another note, did you look to see how your old model performed yesterday? My guess (based on how mine did, and how we used to be generally in agreement) is that it outperformed this new one by a wide margin.

[formula_2002]

46%?! That sounds like a ruinous strategy. Have you run any sims on that?

On another note, did you look to see how your old model performed yesterday? My guess (based on how mine did, and how we used to be generally in agreement) is that it outperformed this new one by a wide margin.

You have to read a bit further in my note, I would cut 46 dn to 4.6 , and would not commit any $ until I ran simulations (using my data base).

I did make an attemt to run the old program against yesterday's results, but it didn't happen.
I'll be running the new vs all the old some time this week.

[ryesteve]

You have to read a bit further in my note, I would cut 46 dn to 4.6
I did read further down, and understood what you said YOU would do... I was expressing surprise that a straight application of Kelly would recommend a betsize that high.

[formula_2002]

Somewhere in the process of setting up this system, I was shocked to find that 80% of the picks were losing...then i discovered that i had the plays reversed, i was playing the underlays. A simple correction got that resolved!!.

I haven't tested the whole thing..Success, or the lack of it will tell.

Is success more telling than "logic" ?
If I make the "logical play", that is when my odds line is less than the Pinnacle line, I get killed.

When I play the reverse, I get the following results;

24 wins, 44 plays, +$716, $4649 played.
that's 55% winners and a 15% profit...AND..
win or lose, the final track tote board odds (adjusted for the match play) sum to more than the Pinnacle odds.

Over the 44 plays, the track odds difference between the Pinnacle odds was $13.20 !!
Perhaps I should call the system "Reverse Quantum"!!

[ryesteve]

the final track tote board odds (adjusted for the match play) sum to more than the Pinnacle odds.
So are you saying that, for instance, in a situation where the Pinnacle line is even, your selection tends to go over at higher odds?? That would certainly be a recipe for disaster!

[formula_2002]

So are you saying that, for instance, in a situation where the Pinnacle line is even, your selection tends to go over at higher odds?? That would certainly be a recipe for disaster!


Yes (my higher odds would be the play..nuts, but that how it appears)


Rye,,i'm going to e-mail my tally sheet covering the "Quantum" period. I have to make one correction that I spotted.

Amoungst other things it list,
match up horses with Pinnalce lines, My lines, Edge,
final track odds, fin postion, $won, $lost $bet
normalized track odds for match play,
compares "bet" surface to "ran" surface (to insure race was not taken off of grass) $ difference in track and pinnical odds..etc

[formula_2002]

Oh, there are three other number I track,
number of expected winners base on tote board odds, Pinnacle odds and my odds.

The actual number of winners in this sample was 13 .
expected winners base on final tote board odds was about 10, Pinnacle was about 10 and mine was but 6+..
And that may explain the issue. My line is too low.

I'll leave it that way for nownad just use 'Kentucky Windage".

[ryesteve]

It sounds to me like maybe your model is using factors that may correlate well to value (ie they select horses who outperform their odds) but correlate weakly (or even negatively) to odds. In the context of matchups, that'd be a bad thing. If I had a model that did a good job of predicting final odds, I wouldn't even worry about trying to predict probabilities of winning. As you've said before, as long as you're on the "correct" side of the toteboard by a wide enough margin relative to the matchup line, profit is guaranteed.