Wrong approach?

[MJC922]

Originally Posted by Hosshead

From what I see in the results from the Speculus article, he did pretty damn good with show betting.

A 56% Profit !!

He hit 71% of the show bets (whether there was a profit or break even)

And from what I calculate ( I may be wrong), the Avg. show price was about U.S $4.38
Thats a pretty nice avg. show price. (with a 71% hit rate)
Were these large fields, or just longer odds picks?

The next question is why the show and not the win?
I read that these selections were posted in the newspaper the morning of the race.
Could that have influenced anything?

His approach is on the right path IMO. First though the sample size < 80 plays is way too small to draw much of a conclusion. For example I have a mechanical method which had 170 plays in May with 48% wins and made 8%, now we're halfway into June and the profit is down to 1%and 45% wins. Looking at my median payoff it's 4-5 and losing the take, profit or no profit I'm not very confident at the moment. Second observation: How are the horses on his list selected? If it's random I don't see a lot of upside. One problem that I always ran into was finding enough horses worth 'following', that is, I always looked for horses that I would consider purchasing due to perceived hidden ability /class, unfortunately if you're following a single track there aren't many of them. If following multiple tracks it's hard to closely observe behavior, warm ups, post race, and not the least of which is to know the class of the others you're entered against.

[Hosshead]

Imagine playing Even Money Favorites and having a 71% Win Percentage.

That is what Speculus did with his show bets.

Of course I'm using a flat bet example, which he was not, but still, it's a good profit.

[Dave Schwartz]

For example I have a mechanical method which had 170 plays in May with 48% wins and made 8%, now we're halfway into June and the profit is down to 1%and 45% wins.

I have seen a system winning at 2,000 plays and losing badly by 3,000.

A lot has to do with how much the profit depends upon longshots.

I do not subscribe to the theory that you have to toss those because they may be what makes your system "sing." You just won't know until you get enough plays.

But the fact that you don't know for sure does not mean you cannot play the system.

[speculus]

Originally Posted by Hosshead

From what I see in the results from the Speculus article, he did pretty damn good with show betting.

A 56% Profit !!

He hit 71% of the show bets (whether there was a profit or break even)

And from what I calculate ( I may be wrong), the Avg. show price was about U.S $4.38
Thats a pretty nice avg. show price. (with a 71% hit rate)
Were these large fields, or just longer odds picks?

The next question is why the show and not the win?
I read that these selections were posted in the newspaper the morning of the race.
Could that have influenced anything?

No. Maybe you have made an error in calculation. The strike rate of overall 70% for SHOW is right, but maybe you made some error about avg odds.

The maximum show payout was 3.40 to 1, which in the U.S. context would mean a show dividend of $8.80. But that was max, not average.

Of course, it can be said that because of the popularity of the experiment, all selections of the column started getting hammered in the show pool, and I believe the average (ROI) was down to 10 or 11% profit on flat-bet basis, or in the US context, an average return of $2.20 for a $2 bet.

However, due to the peculiar nature of the money management practiced (which you can find at the bottom of the FAQ link), the profit was enhanced to 150%.

Of course it must be said that all ground rules (as I called them) of the experiment, including that of money management, were set well before the experiment began, and each rule was strictly adhered to till the end.

As regards your next question, the performance of the same selections in the "Recommended bet" category, which was not restricted to WIN only but had a mix of SHOW & SHP (SHP stands for Second Horse Pool, you get paid if and only if the horse finishes second; this pool is hugely popular as most bettors hedge their WIN bet with SHP), was, disastrous, especially because of the same money management method that enhanced SHOW performance!

I feel there are two important lessons to be drawn from this: One, if you have a good handicapping method, you should not burden it with too many layers of decision making, that will only deduct from the efficacy of your model. And two, any money management method based on an anti-martingale principle is the friend of 'high strike rate' but enemy of 'low strike rate'.

Interestingly, if you consider only the WIN category, then barring the six races in which the system had more than one horse to follow, there were 64 races (or bets) & 19 of them won (29.7%) with avg odds being 1.72 to 1, resulting in a flat-bet loss of 19% which is close to the standard takeout rate.

Also, besides these 19 winners, another 19 came in second & 16 finished third--giving a combined strike rate of 77% for SHOW--surely not bad for a method which doesn't even look at any other horse except its own miniscule short list.

[speculus]

Originally Posted by Dave Schwartz

A lot has to do with how much the profit depends upon longshots.

The more I think the more I feel that only a method with either an exceptionally high strike rate or exceptionally high avg odds is perhaps the only solution to be a consistent winner (as over long term) at this game. Any other method (with mid-range or "average" values for either, or both) is fraught with potential (long term) failure though it may occasionally (short term) show brilliance in flashes.

[MJC922]

Originally Posted by Dave Schwartz

I have seen a system winning at 2,000 plays and losing badly by 3,000.

A lot has to do with how much the profit depends upon longshots.

I do not subscribe to the theory that you have to toss those because they may be what makes your system "sing." You just won't know until you get enough plays.

But the fact that you don't know for sure does not mean you cannot play the system.

In my opinion when it comes to identifying a profitable system we unfortunately need a better measure than ROI. True ROI is all that really matters once we're putting down hard cash but I feel for test purposes it's leads us down blind alleys more often than not. For example I'm coming off a condition study where I sequenced nearly 2 million pp lines. I then queried up every sequence winning over 20% of the time with ROI over 20% and with 15 wins or more for sample size. At this point there were just over 170 'spot plays' that qualified for closer scrutiny. When I dug deeper though I found that only about two dozen were still able to show a profit when median payoff was the measure as opposed to avg payoff (ROI). Most of them when you examined the payoffs were small samples of 50 starters and just 'caught' one random 25-1 shot with the rest of the prices under $10. IMO these bombs are random, they will occur but we need to ignore them until we can project how often. To this end I have a strong suspicion median payoff is the key metric. Either median payoff is by itself all that matters or there should be a way to project the true avg payoff over time using median payoff.

[Hosshead]

re: Show Bets

Originally Posted by speculus

No. Maybe you have made an error in calculation. The strike rate of overall 70% for SHOW is right, but maybe you made some error about avg odds.

Yes it appears an error because I was calculating on a flat bet.
But your Money Management system has increased the flat bet profit immensely.

Originally Posted by speculus

However, due to the peculiar nature of the money management practiced (which you can find at the bottom of the FAQ link), the profit was enhanced to 150%.

Trying to understand about the 150% profit, as I thought it was 56%

Originally Posted by speculus

I believe the average (ROI) was down to 10 or 11% profit on flat-bet basis, or in the US context, an average return of $2.20 for a $2 bet..

Now that I know your flat-bet ROI, I calculated that your hit rate of 71% and about a 10% profit overall = about a $3.10 Payoff for each ($2) bet that you won/hit.
But again, your MM system comes to the rescue !

Only natural to ask:
Could your MM system turn a 10% flat bet loss into a profit?

I found this whole experiment interesting Speculus, and the fact that you documented this in real time, out on a limb, In A Public Newspaper , whether win or lose, is commended. Good going !
_________

[Robert Goren]

Originally Posted by Dave Schwartz

I have seen a system winning at 2,000 plays and losing badly by 3,000.

A lot has to do with how much the profit depends upon longshots.

I do not subscribe to the theory that you have to toss those because they may be what makes your system "sing." You just won't know until you get enough plays.

But the fact that you don't know for sure does not mean you cannot play the system.

And it could very well be profitable again at 4000 plays. Systems that pick random huge pay offs are tough to know about. To be sure, it has be profitable without the boxcar payoffs. If it profitable if throw out the largest 5% of the payoffs, then you might have something, maybe.

[speculus]

Originally Posted by Hosshead

....Trying to understand about the 150% profit, as I thought it was 56%.....

If I had not varied my bet size and wagered a constant amount of first bet (500) on all 76 bets, I would have bet 38k and made profit of approximately 3.8k (about 10% profit on turnover), but due to the peculiar way of managing money, I ended up making about 5.6k profit as you can see. I was referring to that enhancement (5.6k/3.8k) as about 150%.

Originally Posted by Hosshead

....Only natural to ask:
Could your MM system turn a 10% flat bet loss into a profit?.....

NO, not yet. Though at times, I have found it doing something similar, I must say that could be due to the peculiar WIN-LOSS sequence in that particular sample.

However, in most cases it would reduce the loss, like say turning a 10% negative into a 4% negative, and that may be thought of as a significant achievement if you consider the existence of rebates.

But before Mike steps in with the charge of violation of terms, let me make clear it is NOT FOR SALE.

[speculus]

Originally Posted by speculus

NO, not yet. Though at times, I have found it doing something similar, I must say that could be due to the peculiar WIN-LOSS sequence in that particular sample.

However, in most cases it would reduce the loss, like say turning a 10% negative into a 4% negative, and that may be thought of as a significant achievement if you consider the existence of rebates.

But before Mike steps in with the charge of violation of terms, let me make clear it is NOT FOR SALE.

Just to avoid the confusion, Hosshead.

What I wrote above is NOT related to the simplistic but peculiar money management method I adopted during the My Stable experiment to which you responded, but it was related to my MM research as described on this page or this upload, which too, incidentally, were recommendations published in my selections column on the race club's official website.

By the way, I wonder if any other race club in the world has ever done it---making available profitable recommendations FREE for three straight seasons now as my column, despite the phenomenal following it enjoys, has managed to show flat-bet profit on recommendations since it started: Mumbai (2011-12), Pune (2012) & Mumbai (2012-13).