True Favorites

[sligg]

I'm wondering if I'm playing games with myself using a spread sheet for analysis?

I have 5 months of results logged into a spread sheet. I like to tweak to improve results. True Favorites has not had a losing month.

No tweaks:

Bets: 68
Won: 54
Won %: 79.4
$ ret: 169.7
$ net: 33.7
roi %: 24.8

Tweak: added a recency restriction:

Bets: 43
Won: 36
Won %: 83.7
$ ret: 115.2
$ net: 29.2
roi %: 34

Tweak: added an odds rule to the recency tweak:

Bets: 33
Won: 28
Won %: 84.8
$ ret: 95.6
$ net: 29.6
roi %: 44.8

Impressive to say the least. But all is not rosy.

1. Limited spot play, only 6.6 bets in a month from on average of 5400 races per month.

2. The recency rule is doable, last race less than 30 days.

3. Odds rule: 3/5 minimum, 1.5 maximum. With all the late money this may be questionable.

4. Maximum run out: 1

That's it. True Favorites is a black box method. I handicap 180 races across 20 tracks in about 15 minutes a day. Most of the bets come from the small tracks so amount of money you can bet may be limited without reducing the payout.

[hcap]

No tweaks is more reliable IMHO. You took your method developed on past data and had very very impressive results going forward. Not only that but your total earned-$ ret: 169.7- is greater than the others.

Excellent.
Have you tried keying your choice in exactas?

[Hosshead]

..
I have 5 months of results logged into a spread sheet. I like to tweak to improve results. True Favorites has not had a losing month.

1. No tweaks:

.[/B] Bets: 68
Won: 54
Won %: 79.4
$ ret: 169.7
$ net: 33.7
roi %: 24.8

2. Tweak: added a recency restriction:

Bets: 43
Won: 36
Won %: 83.7
$ ret: 115.2
$ net: 29.2
roi %: 34

3. Tweak: added an odds rule to the recency tweak:

Bets: 33
Won: 28
Won %: 84.8
$ ret: 95.6
$ net: 29.6
roi %: 44.8

Impressive to say the least. But all is not rosy.

1. Limited spot play, only 6.6 bets in a month from on average of 5400 races per month.

2. The recency rule is doable, last race less than 30 days.

3. Odds rule: 3/5 minimum, 1.5 maximum. With all the late money this may be questionable.

4. Maximum run out: 1

That's it. True Favorites is a black box method. I handicap 180 races across 20 tracks in about 15 minutes a day. Most of the bets come from the small tracks so amount of money you can bet may be limited without reducing the payout.
Sal,

It looks like you can get the most plays (approx 1 play every 2 days) with method 1 . That is also the method you can make the most $'s per month.

Results (profits) of $100 win bets for the 5 months of plays you listed:
Method 1. - $1686.40
Method 2. - $1462.00
Method 3. - $1478.00

I wouldn't worry about trying to narrow it down any further with recency rules.
Not enough races to split up the sample.
Keep up the good work !
------ Hoss

[lsosa54]

Outstanding, Sal - where else could you get that return on 15 minutes a day?!

[Tom Barrister]

The more you tweak it, the more you're backfitting the method to the results. Essentially, you're finding some factor or combination of factors that happens to filter enough losers out so that the results improve.

For example, if you looked hard enough, you might find that horses whose name begins with the letter "S" (or whatever letter) are only 2 for 8 with an ROI of 0.45 (55 cent loss on the dollar). Would you add a rule to eliminate all horses whose name begins with "S"? Of course not, because it doesn't make any sense.

Trying to squeeze every last ounce of profit from a method is doing just about the same thing---you're using the results to model the method, and that usually doesn't work on a limited sample, even though the reason (recency) may make sense. Horses picked by your method who are racing over 30 days ago are still hitting a healthy 72% (18 for 25). In the next six months, you may hit 80% winners from these kind of horses. Many competent trainers don't race certain horses more often than every five or six weeks. A 90 day rule might be more in line with reality, but again, if the win percentage is still good with the over-90-days horses, you might want to reconsider.

68 plays is a very small sample. It's useless, in my opinion, to try to filter it at this point. As the saying goes: "If it ain't broke, don't fix it."

[sligg]

No tweaks is more reliable IMHO. You took your method developed on past data and had very very impressive results going forward. Not only that but your total earned-$ ret: 169.7- is greater than the others.

Excellent.
Have you tried keying your choice in exactas?

No. What would I use for the second horse? That would require further handicapping which I am trying to avoid. I like the comfort of a black box especially if the % winners holds up. At my stage in life, money is not the ultimate end.

[sligg]

The more you tweak it, the more you're backfitting the method to the results. Essentially, you're finding some factor or combination of factors that happens to filter enough losers out so that the results improve.

For example, if you looked hard enough, you might find that horses whose name begins with the letter "S" (or whatever letter) are only 2 for 8 with an ROI of 0.45 (55 cent loss on the dollar). Would you add a rule to eliminate all horses whose name begins with "S"? Of course not, because it doesn't make any sense.

Trying to squeeze every last ounce of profit from a method is doing just about the same thing---you're using the results to model the method, and that usually doesn't work on a limited sample, even though the reason (recency) may make sense. Horses picked by your method who are racing over 30 days ago are still hitting a healthy 72% (18 for 25). In the next six months, you may hit 80% winners from these kind of horses. Many competent trainers don't race certain horses more often than every five or six weeks. A 90 day rule might be more in line with reality, but again, if the win percentage is still good with the over-90-days horses, you might want to reconsider.

68 plays is a very small sample. It's useless, in my opinion, to try to filter it at this point. As the saying goes: "If it ain't broke, don't fix it."

You are right, the sample is much to small to be definitive on any tweaks. As I have said in previous posts my end goal is having fun and not making money so that's why I play around with tweaks.