PDA

View Full Version : Would some of you test this for me, please?


Gramps
5th September 2001, 02:58.54 PM
I have a small database currently, compared to some of you, about 71,000 lines, which would be a bit short of 9,000 races.

I've finally gotten Access to do what I want it to, and I've been fine tuning what I call the "filter" query. The core of the filter simply takes all the low-percentage factors and filters them out.

The "bet all" query with nothing plugged in, and using paceline 5, is:

Bets: 71,580
Wins: 8,701
Pct : 12.16 %
Avg Mutuel: $ 12.38
ROI: 0.752

Here are the parameters for my filter core:

tTYPE: <> "2yr" And <> "2yf"
nLAY : <=60
nEQC: <>1 And <>2 And <>3
tRS : <>"R" And <>"S"
fFAV : 0
nSPCA: <>1 And <>2 And <>3 And <>4
nSTAR: 1


For those of you who don't use HTR, this basically means to

1) Exclude races for 2 year olds,

2) Eliminate horses who:
a) Haven't raced in over 60 days
b) Are making their first start on Lasix
c) Have the equipment change of "blinkers on" today
d) Have a running style of "Sustained" (late runner) and "Rear" (very late runner)
e) Were favored in their last race
f) First time starters
g) Just broke their maiden in their last race
h) Are moving from a sprint to route
i) Are moving from a route to sprint

3) Include horses who got exactly one "K star" in their rating (K-stars range from zero to five).

All of those are logical to me. My line of thinking, which may not be yours, is this: Two year olds are hard to handicap and are overbet because there are many overmatches that the stables know about but we don't. Horses making their first start on Lasix are usually overbet. Horses adding blinkers are often doing so because they have a problem keeping their mind on their business. Late running horses don't win their fair share of races. Last-race favorites are usually overbet in today's race. Horses changing distances usually are doing so because the trainer is looking for something that works. And the one K Star singles out horses who have a shot, but who aren't so obvious that everybody else sees them (as is often the case with more stars).

For non HTR and or Access users, that would be time-consuming to do on paper, but it takes Access less than 10 seconds to run through my 71,000+ line (about 9,000 races) database once those are plugged into the query.

With those parameters, I came up with the following results, again, using paceline 5:

Bets: 8292
Wins: 1041
Pct: 12.55 %
Avg Mutuel: $ 15.25
ROI: 0.957

And that ROI boggles me. I still have better than one out of nine horses remaining (almost one per race), and little more than a 4% loss by just using common-sense elimination and one rating (the K-star)?

If any of you have time, would you mind running this query through your Access bases and posting your results?

MikeDee
5th September 2001, 03:23.15 PM
Gramps you didn't mention surface turf, and dirt and wet?

I'll run it tomorrow morning. I only have the tracks I am curently following and can wager at but it will give you an idea.

George
5th September 2001, 03:30.45 PM
Gramps,
I am not a keeper of "large" databases so really can't help out. Think Carl, Mike, or Rick may be able to run it on a big one.

Just wanted to comment that I liked your concept and to agree it is quite interesting to get a 0.95 roi by merely using some basic logical eliminators.

An area that I have done a lot of testing on in last couple years is layoffs. You might want to reconsider or at least retest leaving out your layoff <60 parameter. IMO some of the best bets now that we have better racing thru chemistry, are long layoff horses. Trainers no longer "race" a layoff horse back into peak form. Not sure they ever did.

Will be very interested if somebody runs your concept on large data.

Rick
5th September 2001, 04:56.38 PM
The "bet all" query with nothing plugged in, and using paceline 5, is:

Bets: 71,580
Wins: 8,701
Pct : 12.16 %
Avg Mutuel: $ 12.38
ROI: 0.752

I have to say after reading that you better go back and check your roi report.

I would expect a net loss of about 18% (takeout) betting every horse in every race.

Check that out and then rerun everything.

Rick
5th September 2001, 05:03.19 PM
Worse than that.

ALL_MSA Misc Query ROI Report
05-Sep-01
Total Bets 521362
Total Amount Bet 1042724
Wins 63853
Pct. 12.25%
Amount Won 793368.20
Profit/Loss -249355.80
Pct Profit Loss -23.91%
Avg. Payout 12.42

A bit worse than minus 18% when betting every horse in every race in db for last year.

Check your figures.

Carl
5th September 2001, 05:14.00 PM
Rick,

I think a problem with our ROI Calculator is in races with entries it counts the horses 1 and 1A as two separate bets. So even if one half of the entry wins, we get credited with a "losing horse". And if both halves lose---well that is so depressing I don't even want to go there.

This and breakage move us away from the 18% overall loss to a number 30% or so greater. Don't really mind it, I am used to it, and if the spot play shows "break even" I can theorically make a profit hehe.

later,
Carl

Gramps
5th September 2001, 07:21.03 PM
Mike: It's on all surfaces. Since the parameters for the core are very generic in nature, I figured they would work on all surfaces and distances. Of course, I'll check it by surface later.

Rick: it's a 25% loss (a 0.75 ROI) and that's been pretty constant. I'm using every track for the past 50 or so days, except the really small ones (Marquis, Araphoe, Yavapai, etc.). That isn't far off of what you came up with (-23.91% which is an roi of about 0.76) and the average mutuel is about the same. The norm would be about 22%, I'd think. because of breakage and the fact that shorter prices win a little more than their share and the big longshots win a lot less.

I'm still interested in how my settings work on other databases.

Rick
5th September 2001, 09:32.30 PM
Sorry I didn't understand your original post.

Here's your query on my 1 year file:
ALL_MSA Misc Query ROI Report
05-Sep-01
Total Bets 60778
Total Amount Bet 121556
Wins 7023
Pct. 11.56%
Amount Won 105965.20
Profit/Loss -15590.80
Pct Profit Loss -12.83%
Avg. Payout 15.09

Gramps
5th September 2001, 10:18.55 PM
I suspected that there were too many long priced horses in there boosting the price. But when I set nODDS < 20 (for non HTR users, that's post time odds), it upped the win percentage a little but didn't change the ROI.

The same criteria with nODDS < 20 added:

Bets: 7084
Wins: 1005
Pct: 14.19 %
Avg Mutuel: $ 13.48
ROI 0.956

I don't know what to make of it, myself. Does anybody else get results that correspond to mine or Rick's? Rick: can you run it for the past 45 days only (since July 15th) to see if something might be amiss in my database?

Rick
5th September 2001, 11:33.00 PM
15 Jul to 12 Aug
ALL_MSA Misc Query ROI Report
05-Sep-01
Total Bets 5487
Total Amount Bet 10974
Wins 693
Pct. 12.63%
Amount Won 10204.60
Profit/Loss -769.40
Pct Profit Loss -7.01%
Avg. Payout 14.73
Thats all I got till next longshots report.

Gramps
6th September 2001, 12:03.19 AM
That looks a lot closer--I'm assuming you you use paceline 5-- and the difference could be in the extra tracks I run. That probably means that this works better in the summer than it does at other times, which would make sense.

MikeDee
6th September 2001, 08:24.30 AM
Here's wat I got from my data. this covers most but not races at these tracks during 99,00,01
ap 00 and 01 only, ct 01 only, gld 00 and 01 only, mnr 00 and 01 only, pen some 00 and 01 only
Trk nWin nBet Win% aMut ROI$
AP 163 1586 10% $16.31 0.84
BEL 200 1761 11% $17.07 0.97
CRC 379 2971 13% $15.87 1.01
CT 211 1607 13% $14.92 0.98
DEL 417 3277 13% $14.42 0.92
DMR 85 813 10% $15.38 0.80
FE 320 2565 12% $14.07 0.88
GLD 217 1855 12% $15.00 0.88
LAD 203 1914 11% $15.10 0.80
MNR 247 2117 12% $16.82 0.98
PEN 267 2384 11% $14.51 0.81
PHA 474 3857 12% $14.66 0.90
PRM 209 1970 11% $16.36 0.87
RKM 247 1874 13% $14.60 0.96
TDN 335 3147 11% $14.03 0.75
TP 231 2010 11% $14.71 0.85
WO 371 3154 12% $14.04 0.83
Total 4576 38862 12% 0.88

Gramps
6th September 2001, 03:22.49 PM
Thanks for the help. I guess my database is too small (71,000 lines and 8,000 plays), or the results are seasonally high. I ran the test again with maximum nODDS of 30 and also of 20, and the ROI didn't change much. I'm guessing that as the database grows, the ROI will regress back towards the 0.88.

Mike: How did you get it to export all the data by track?

MikeDee
6th September 2001, 03:47.44 PM
I have a vba module that runs roi by track and puts the results in a roi table. then I can query the roi table. I export the roi query to excel and paste the spreadsheet to a message and it works pretty real well.

I put together a package on how to do this for cliff and hurricane. If you want it let me know I can send it to you and walk you through getting in going in your db.

Gramps
6th September 2001, 06:06.34 PM
Thanks for the offer. I would appreciate it if you would send that and help me get it up and running.

Now, I need to get the office 2000 cd out and install excel. :D

MikeDee
7th September 2001, 10:31.26 AM
gramps send me a email message at mrdezo@ameritech.net I will attach the module and sene it to you with instructions