db/pencil

[BIG HIT]

hi guys have question or rather need opion.I read that a test of thirty race or so well tell you the same info that 10,000 race will. now being a pencil person and have no db it seems you have the same hot and cold streaks.Which i thought were because i did not have enough races.One thing know for sure db are faster and more mistake free.

[GR1@HTR]

Big Hit. Me thinks 30 not enuf for sure. A while back I thought I found the money bank. On just a few weeks fo data I found a play that hit 44% w/ close to 2.0 ROI over 50 samples. Then guess what when I started to watch how it did....Lost 19 in ROW!!!!! Well anyways, 10,000 may not be enuf either. Don't think anybody really knows how many is enuf. To be honest, something that worked for last 50,000 races may not work for the next 50,000 races. In most cases, it is like chasing the golden pot at the end of the rainbow.

[Larry Hamilton]

never mind

[tilson]

Big Hit I am not a database guy more of a pencil and paper guy.....but i can tell you i have done Several past studies and 30 ain't even close you need like 175 races just to get a HINT if something is worthy of a deeper look......hell i was showing like a + 20% roi after 100 races a few different times......by the time i got to 220 races it was down to like +3%

[GR1@HTR]

Tilson,

You know what really sucks about those type of studies where we get positive roi over 100 races of 1.20 and just +3 after 200...The last 100 of the 200 showed a negative 20% ROI!!! Been there more times than I wish to speak of. In matter of fact I would venture any query that showed a postive ROI over 100 plays, that there is a greater and than 75% chance it will show a negative ROI over the next 100. Not saying that the total of the 200 plays will be negative, just the 2nd 100 plays following the 1st 100.

[Tim]

Nothing lasts forever.

Here is a list to show that even with a large sample, you still can get killed. It's an annual list of Gasper S. Moschera's performance as a NYRA trainer.

Year Starts Wins WinPct
---- ------ ---- --------
1992 554 134 24%
1993 492 112 23
1994 512 107 21
1995 528 106 20
1996 534 89 17
1997 445 83 19
1998 385 79 21
1999 347 46 13
2000 294 21 7
2001 104 15 14 (year to date)
---- ------ ---- ------
Total 4,195 792 19%

If, as a trainer intent handicapper, you conducted an analysis of Moschera's performance in mid 1999 his impressive stats looked great. If you then started betting on his horses you would have lost your shirt. What I'm saying is that it doesn't matter how good the past looks or how large your sample is, you still have to maintain a method to measure current performance.

I use rolling ROI's. I maintain 3 different ROI calculations for any betting methodology that I employ. Such as:

Rolling ROI calculation - Last 20 bets
Rolling ROI calculation - Last 50 bets
Standard ROI calculation

Depending on the methodology I modify the number of bets used or use a time based rolling ROI such as last 30 days. At any rate, the power if a rolling ROI calculation should be clear. I expect swings in the short ROI calculation but will stop dead in my tracks with I see movement in the longer ROI calc's. Rolling ROI's have saved me many a time from throwing good money after bad.


Tim

[Tom]

When I calculate my Beyer pars for different tracks, I not only look at the long term data to get a large enough sampel for each class, but I also look at the last two months to make sure things are still looking the same. Long term, $10K horses at NYRA have a Beyer par of about an 81-82. This year, that number is about a 75! Big difference. At FL, I had to actually throw out all of the data prior to 1999 becasue things had changed so radically. Computer databases allow you to look at things in certain time frames.
Tom