Robotic Pic3 Testing

[Ray2000]

Robotic Pic3s tossing out Predicted 1st and 2nd Crowd favorites.
Results to follow.
Playing these on paper to start.


MR R1....6 / 82 / 8
MR R2....82 / 8 / 528
MR R3....8 / 528 / 41
MR R4....528 / 41 / 2
MR R5....41 / 2 / 1
MR R6....2 / 1 / 2
MR R7....1 / 2 / 43
MR R8....2 / 43 / 2
MR R9....43 / 2 / 58
MR R10....2 / 58 / 18
MR R11....58 / 18 / 5

MEA R3....24 / 8 / 63
MEA R6....43 / 3 / 5
MEA R9....81 / 83 / 97
MEA R12....2 / 16 / 72

MOH R1....95 / 5 / 32
MOH R2....5 / 32 / 98
MOH R3....32 / 98 / 510
MOH R4....98 / 510 / 6
MOH R5....510 / 6 / 93
MOH R6....6 / 93 / 6
MOH R7....93 / 6 / 6
MOH R8....6 / 6 / 1
MOH R9....6 / 1 / 9110
MOH R10....1 / 9110 / 8

[Sinner369]

Is this a new version?

[Ray2000]

Sinner

I was reading:

http://www.econ.washington.edu/user/.../horse1105.pdf

about Pic 3 strategies and the advantage of eliminating crowd faves so I put together a method of predicting the favorite which I hope to be superior to Morning Line numbers.

The top 3 selections in each race are still from my regular computer program (robot). The new part is I'm tossing out any picks that the program "thinks" will be the favorite or 2nd choice on the tote.

This prediction method uses a different model for each race track so I'll be doing Meadows, Mohawk, Monticello, Pocono and Hoosier, all have multiple pic3 wagers each day.

Might be informative ...

[Sinner369]

Keep us posted on the results...........I play Pick 3's myself and usually on the WEG circuit..........Mohawk and Woodbine......!

[mrroyboy]

Good work Ray

[Ray2000]

Terrible start with bad accident and all
Going for longshots in this run so hit rate will be low.
Ignoring Monti R1 and 2,

Results Todate 
$1 base tickets, all Scratches treated as a loss	
				Fav was correctly predicted..			
Date	Track	Cost 	Return	Bot 	ML 
3-Jun	MR	39	0	8/12	8/12
3-Jun	MEA	18	0	11/15	13/15
3-Jun	MOH	35	0	7/11	7/11
		92

Tuesday, 6/4

HOP R1....7 / 3 / 2
HOP R2....3 / 2 / 7
HOP R3....2 / 7 / 57
HOP R4....7 / 57 / 6107
HOP R5....57 / 6107 / 6
HOP R6....6107 / 6 / 9
HOP R7....6 / 9 / 259
HOP R8....9 / 259 / 6
HOP R9....259 / 6 / 105
HOP R10....6 / 105 / 52
HOP R11....105 / 52 / 419

MEA R3....71 / 12 / 6
MEA R6....1 / 1 / 2
MEA R9....7 / 31 / 64
MEA R12....349 / 2 / 19

MR R1....3 / 18 / 18
MR R2....18 / 18 / 6
MR R3....18 / 6 / 1
MR R4....6 / 1 / 54
MR R5....1 / 54 / 5
MR R6....54 / 5 / 5
MR R7....5 / 5 / 5
MR R8....5 / 5 / 53
MR R9....5 / 53 / 4

PCD R1....45 / 9 / 1
PCD R4....128 / 68 / 4
PCD R7....291 / 35 / 7
PCD R10....3 / 45 / 82

[Ray2000]

Not good and might have to reprogram the bot to gather Pic 3 results from a different source.
The Racing channel reports pic3s Willy-nilly.

Results Todate 
$1 base tickets all Scratches treated as a loss	
				*Fav was correctly predicted..			
Track	Cost 	Return	Bot* 	ML* 
MR	60	0	15/22	15/22
MEA	33	0	21/30	24/30
MOH	35	0	7/11	7/11
HOP	47	0	8/13	9/13
PCD	18	0	9/14	9/14
	193

Weds, 6/5
HOP R1....1 | 12 | 27
HOP R2....12 | 27 | 8
HOP R3....27 | 8 | 6
HOP R4....8 | 6 | 54
HOP R5....6 | 54 | 3
HOP R6....54 | 3 | 5
HOP R7....3 | 5 | 974
HOP R8....5 | 974 | 265
HOP R9....974 | 265 | 52
HOP R10....265 | 52 | 35
HOP R11....52 | 35 | 410

MEA R3....51 | 2 | 4
MEA R6....74 | 8 | 5
MEA R9....28 | 31 | 2
MEA R12....8 | 78 | 395

MR R1....8 | 1 | 23
MR R2....1 | 23 | 7
MR R3....23 | 7 | 5
MR R4....7 | 5 | 276
MR R5....5 | 276 | 5
MR R6....276 | 5 | 12
MR R7....5 | 12 | 1
MR R8....12 | 1 | 6
MR R9....1 | 6 | 81
MR R10....6 | 81 | 1
MR R11....81 | 1 | 83

PCD R1....17 | 63 | 9
PCD R4....7 | 6 | 3
PCD R7....54 | 14 | 31
PCD R10....54 | 8 | 97

[melman]

A pic3 play for me today at the Meadows.

R3 5712
R4 732
R5 4

Ray it's a Milke Wild Man Wilder special.

[Ray2000]

OK mel, good luck

I'm thinking of speeding this test run up a bit by not tossing out the predicting 2nd fave, just the top fave. That will increase the cost quite a bit, but I'm playing these on paper anyway.

Meadows R3 now becomes...

MEA R3....51 | 12 | 49

I'll buy that ticket for real...

[mrroyboy]

Ray what you are checking is good but it may not be best for pic'3 etc

[Ray2000]

Yes roy, this is not working out and collecting data manually is a pain.
Also
Robot 'Favorite Predictor' is falling behind Morning Line.
May have to revisit model.

Results Todate ($1 base) 
*Fave correctly predicted %			
Track	P3Cost 	Return	Bot* 	ML* 
MR	82	0	60%	66%
MEA	47	0	58%	67%
MOH	38	0	58%	58%
HOP	113	0	69%	73%
PCD	35	0	72%	79%
	315

Thursday
HOP R1....6 | 43 | 2
HOP R2....43 | 2 | 478
Race 3 is now a Pick4
HOP R4....478 | 410 | 2
HOP R5....410 | 2 | 7
HOP R6....2 | 7 | 379
HOP R7....7 | 379 | 5
HOP R8....379 | 5 | 23
HOP R9....5 | 23 | 10
HOP R10....23 | 10 | 52
HOP R11....10 | 52 | 8

MOH R1....19 | 4610 | 8
MOH R2....4610 | 8 | 83
MOH R3....8 | 83 | 59
MOH R4....83 | 59 | 67
MOH R5....59 | 67 | 51
MOH R6....67 | 51 | 107
MOH R7....51 | 107 | 97
MOH R8....107 | 97 | 148
MOH R9....97 | 148 | 26
MOH R10....148 | 26 | 1

MR R1....5 | 51 | 4
MR R2....51 | 4 | 5
MR R3....4 | 5 | 41
MR R4....5 | 41 | 3
MR R5....41 | 3 | 42
MR R6....3 | 42 | 71
MR R7....42 | 71 | 76
MR R8....71 | 76 | 387
MR R9....76 | 387 | 1
MR R10....387 | 1 | 461

[SchagFactorToWin]

I've tested some P3 methods. I found that tossing the ML fav didn't work- they win too often. I pick up some nice paying P3s with 2 favs and a long shot.

Here's what I tested over 4000 races:
1) my top pick (avg. Win odds=2:1): positive ROI, low win rate (exp. win rate 4.1%, actual win rate 4.2%), small bet size, but less actual $ profit than (3) below.
2) top 3 picks: losing ROI, good win rate (exp. win rate 33.9%, actual win rate 34.4%), large bet size
3) top 2 picks: positive ROI, decent win rate (exp. win rate 16.7%, actual win rate 18.1%- not statistically significant of a diff.), reasonable bet size.

I keep a bank and if the balance at the end of the day is greater than my initial bank (the balance I started with originally, not that day's starting balance), that is my profit. With the Top 2 method, profits graph a nice wave pattern. Although the bank loses money 2/3 of the days, it makes a profit about 1 out of every 8 days.

Alas, pool sizes limit the bets to the $2-4 range.

[Ray2000]

SFTW

Very Interesting, thx for sharing

My original reasoning for this testing was... if I could identify the over-bet horses in the pick3 pool, particularly in the 2nd and 3rd legs where no real time tote info is available, then removing them from my ticket should increase the ROI. This, of course, is the same as trying to identify False Favorites in any Win Pool, a very hard task. The first step was to at least do a better job (no pun intended) than the Morning Line in predicting the favorite, False or otherwise, and then manually decide to 'Single it' or 'Toss it'

So far, the method I'm using to identify Faves is not working, and as you have shown, tossing the ML is not a good approach.

What would be nice (and I'm not sure the big bots don't have this capability) would be to have access to the n x n x n Pick3 matrix showing probable payouts...

[melman]

Guys I play a lot of pic3's and I think an important helpful item is the first leg. If I can toss the fav in the first leg I bet more money. It appears that more than just a few pic players really love the idea of being "live" after leg one. This can set up many overlaid pic3's.

[Ray2000]

Good point, mel

Makes me wonder how times when I liked a longshot in leg2 or leg3 and have used a leg 1 obviously overbet fave on my ticket (possibly doubling the investment!) just because I didn't want to miss out on the longshot if it comes in.