HKJC Racing - William Benter (Hong Kong Syndicate Fame) Presentation!

[jfdinneen]

Originally Posted by Wizard of Odds

Well said, and truly you are erudite:

Wizard,

Thank you for the kind words.

The alternate approach you propose sounds very promising and directly addressing the continuing internal conflict between Type-I and Type-II errors is laudable.

Your reply reinforces two general points worth making. First, given that the Hong Kong betting syndicates and equivalents in other jurisdications appear to be predominantly using logit / probit models we should assume they have farmed that specific approach to the limit and that we must seek alternate models to compete successfully. In addition, by focusing on exotic bets, we are less disadvantaged than in the standard WPS markets. Even Benter concedes this point when he states that "...In ultra-exotic bets such as the Pick-6, even a handicapping model with only modest predictive ability can produce advantage bets..." (Benter, 1994). Secondly, I would encourage all handicappers to validate their approaches against the pari-mutuel baseline and, unless their models account for statistically significant amounts of additional variance in race results over and above that already accounted for by the publice win estimates, they should be wary of basing any long-term wagering strategy on such models.

Finally, you inadvertently attributed to me a contribution to another thread which I do not believe I made - please re-check and correct the attribution so that the real author can take the credit.


Best wishes,

John

[classhandicapper]

Originally Posted by cj

First, I agree with most of what you say.

However, there is a definite difference between maximizing ROI and maximizing profit. This is something that it took me some time to figure out and implement correctly, but my profits have increased nicely despite the fact that my ROI has dipped somewhat. Of course, you still have to have a positive ROI, or at least positive after rebate.

I agree with you completely.

Actually, I think there's a balancing act that must occur between ROI and number of races you get to play. All else being equal you can risk much more money per race with a bigger ROI edge without risking ruin. More importantly, you can also be more certain that you actually have a sustainable edge as opposed to an extended run of good luck or one that can easily be wiped out if your primary information becomes more widespread (speed figures for example) or stops working (trainer patterns for example).

Right now, I have a very narrow edge on my place betting so far this year, but I think I'll need another thousand or more bets to really be sure I have something worth the trouble. Even then, if more people start trying to exploit the place pools in a similar way to me, my edge will evaporate and I may not even understand that until my small ROI profit turns into a small ROI loss.

I have no such risk if I inisist on a high margin of safety on my win bets based on more detailed and comprehensive handicapping.

I'm mostly an advocate of high ROIs because I'm actually not trying to maximize my total profit. I'm trying to mazimize my profit within time and pleasure restraints. I get no pleasure out of handicapping, following, and betting battle scarred claimers at 3rd string tracks. I get an incredible amount of pleasure out of following the best horses in the country. So I want to squeeze every dollar I can out of those types of races. I also don't like playing horses all day every day. I like to mix in some Hold'em Poker, Stock Market stuff etc....

I'm mostly looking at from my personal perspective.

[sjk]

jfdinneen,

For those of us who are terminology-challenged could you explain what is meant by logit/probit model?

[jfdinneen]

Originally Posted by sjk

jfdinneen,

For those of us who are terminology-challenged could you explain what is meant by logit/probit model?

sjk,

When I find myself similarly challenged the first port of call is Wikipedia!

Best wishes,

John

[Wiley]

Originally Posted by classhandicapper

I don't have enough of a statistics background or a good enough database to develop a winning model of the type described here. I'm sure it can be done, but I really have to question how accurate it can be even if it is effective.

Class, according to Bentley, he has already created the model and has been reaping the benefits for many years now. The art, according to his presentation, comes in the evaluation of the expected outcome/weighted parameters for each horse. It looks like in Hong Kong racing provides much more detailed information on each horse (weight of horse prior to race - could be a good form indicator which we don't have here, veterinary details, less medication perhaps?) so it is a much more controlled environment to work in (some of the articles indicate this too) so the anaylsis of each horse in my opinion would be more definite.

The guy does not sound like a handicapper so I am guessing he got help from people who do know horse racing to define these parameters or developed them through trial and error analysis or just through brute computing force of every definable piece of data covering every possible rateable factor comes his resulting odds line. From there I am guessing it is all 'cookbook' programmed to the amount of the bet based on the variance with real odds.
I don't know if it is still true today but in one of the articles it said the Hong Kong pools are filled with money from players that use less defining methods like numerology or the 'likes the color of the saddle cloth' bettors that we need more of over here so Bentley is easily able to exploit these less sophisticated bettors for profit which at 24% with a 19% take out is unreal.
I also like the fact that the Hong Kong racing commision looks at these guys as troubling to the average bettor but find it difficult to limit or ban them based on their 19% cut of the very large amounts that they bet.
I guess to play into these pools one must be in Hong Kong? I know when you play Sha Tin here it goes what into the Canadian pools I think?
Thanks for the great link jfdinneen!

[Wizard of Odds]

Originally Posted by jfdinneen

Wizard,

Thank you for the kind words.

The alternate approach you propose sounds very promising and directly addressing the continuing internal conflict between Type-I and Type-II errors is laudable.

Your reply reinforces two general points worth making. First, given that the Hong Kong betting syndicates and equivalents in other jurisdications appear to be predominantly using logit / probit models we should assume they have farmed that specific approach to the limit and that we must seek alternate models to compete successfully. In addition, by focusing on exotic bets, we are less disadvantaged than in the standard WPS markets. Even Benter concedes this point when he states that "...In ultra-exotic bets such as the Pick-6, even a handicapping model with only modest predictive ability can produce advantage bets..." (Benter, 1994). Secondly, I would encourage all handicappers to validate their approaches against the pari-mutuel baseline and, unless their models account for statistically significant amounts of additional variance in race results over and above that already accounted for by the publice win estimates, they should be wary of basing any long-term wagering strategy on such models.

Finally, you inadvertently attributed to me a contribution to another thread which I do not believe I made - please re-check and correct the attribution so that the real author can take the credit.


Best wishes,

John

I stand corrected regarding the attribution.
I believe Game Theory (another statistician/modeller amongst us) is properly attributable. Apologies to GT.

All the Best,

Wizard

[classhandicapper]

Originally Posted by Wiley

Class, according to Bentley, he has already created the model and has been reaping the benefits for many years now.

I don't doubt that. I'm wondering whether a model like that can maximize the potential accuracy of an odds line and thus the ROI and profits vs. the more traditional approach

To me, the game is all about making the most accurate odds line possible or at least recognizing situations where there is likely to be a big inefficiency in the pools.

The more accurate the odds line, the more situations you will find that are actually profitable and the more likely you will avoid simply mixing a lot of moderately bad bets with a bunch of great bets for a slightly positive result.

I'd rather just make the great bets and leave off the bad ones.

It's always been my view that "even" the vast majority of "winning horseplayers" spin huge amounts of money through the windows with little or no edge on the majority of their bets. A good sized percentage of their bets may even be bad ones. They just don't have an accurate enough odds line to know which is which or crave lots of action. They are just happy to net out to a profit.

However, IMO, most of the profit is located within the small percentage of bets that they could have easily identified as the best overlays beforehand.

The trick is be able to tell the marginally profitable bets from the losers. Then you really have something. I'm still working on that.

[stu]

Originally Posted by MichaelNunamaker

You wrote "there is a definite difference between maximizing ROI and maximizing profit."

Indeed, my ROI this year is only about 55% of what it was last year. However, I've wagered over three times as much money. So, at the end of the day, my profits are considerably greater this year than last. If my goal was ROI maximization, I would drop 99.9% of my wagers and only bet on the very very best plays I have. I'd probably quadruple my ROI, and almost totally wipe out my profits in the process.

I agree. A better measure of a horseplayer is rate of bankroll growth. It is hard to use for comparisons since most of us would be unlikely to share this figure with others.

So a frequency of profittable opportunities may be a better measure.

What would you choose an ROI of +50% with one play a week or an ROI of +20% with twenty plays a week?

stu

[classhandicapper]

Originally Posted by stu

What would you choose an ROI of +50% with one play a week or an ROI of +20% with twenty plays a week?

stu

I just want to clarify my point.

What would you rather have, 100 plays per week that net you 5% or the ability to identify the 10 plays per week out of that 100 that produce 90% of all those profits over the course of the year.

For example:

100 plays at $2 = $200 total bet = $10 profit = 5% edge.

10 plays at $2 = $20 total bet = $9 profit = 90% of the profits = 45% edge.

In this example, you are leaving $1 on the table by making fewer plays, but you can bet massively more per race with a 45% edge vs. a 5% edge without any fear of ruin. You can also be more certain that your edge is real and sustainable and not the result of a good run of luck or a temporary set of conditions.

IMHO, most people CAN identify their best overlays eack week.

However, they prefer lots of action and actually put a ton of money through the windows with little or no edge. You have to keep in mind that just because the net was positive 5% in this case, that does not mean that all the horses were equal overlays or that all of them were even overlays. Many might have been underlays while some of them may have been massive overlays.

I'm not sure where the balancing act is between ROI and getting more bets, but I think it's related to being able to identify clearly when you have a small/moderate edges as opposed to just throwing extra money through the windows. There's a grey area in there where many people are throwing money threw the windows for no good reason.

[cj]

Class,

You are forgetting one thing, volume is rewarded through rebates.

[sjk]

I don't make a bet unless it think it is a significant enough overlay that it will contribute to net winnings, but alas, 88% of the time it loses anyway. I bet more on what I consider the better overlays than on those that appear less favorable.

I would very much welcome any suggestions as to how to eliminate some of the losing bets. As far as I can tell losing a large percentage of the time goes with the territory.

[Geekyguy]

The best exercise for isolating your bread-and-butter angles is to play on low money for a day, or "tap out" your day's bankroll on a single race or a single bet.

Whatever you gravitate to is likely your best angle, and your ROI should bump up a dime or more with your "back to the wall" if you're a true pro.

If this were boxing, you wouldn't say that you'll start punching hard once you become champion, but rather that you'd become champion by learning to punch hard.

[classhandicapper]

Originally Posted by cj

Class,

You are forgetting one thing, volume is rewarded through rebates.

I understand. I am also trying to increase my volume but only because the rebates have created more potentially positive situations. That's why I am making some place bets these days (for example). They are now slightly positive where before they were negative.

It's hard for me to explain my point.

Assume I am talking about ROI after rebate.

IMO, when you are operating on a very narrow margin, you are almost certainly putting tons of money through the windows that has little or no positive expectation "with rebate" and probably don't even realize it.

Maybe this is the best way to say it:

You are trying to make sure that every bet you make actually has a positive expectation (with rebate). If you do, the net ROI should be fairly high because it will be the net of some marginally profitable bets and some very profitable bets. (for example - some 2-1 shots with a 35% chance of winning and some 10-1 shots with a 20% chance of winning).

If you are netting out to a very low ROI with rebates, it's screaming to you that you are making a lot of bad bets but they are netting out positive because the very big overlays are overcoming the bad bets.

If that's the case, IMO, you need to tighten up your play and eliminate some of your marginal and probably losing plays because as you increase your ROI you can bet more and have more confidence you are on solid ground. Plus it's the really big overlays that are where most of the profits are located.

I hope I explained it better.

[MichaelNunamaker]

Hi Classhandicapper,

I agree with you point and perhaps I can furnish a numerical example to illustrate your point. Say a handicapper has 1000 plays a year with a ROI of 10% and he can wager $100 per play at that level. He makes a $10,000 profit a year.

Now our handicapper reads this thread<G>, does some number crunching, and finds out he can split his plays into four groups, described as:

Group 1: 250 plays, ROI -20%
Group 2: 250 plays, ROI -2%
Group 3: 250 plays, ROI +2%
Group 4: 250 plays, ROI +40%

Our hero can now completely skip the group 1 and 2 plays, either reduce the amount wagered on (or skip entirely) the group 3 plays, and dramatically increase his wagers on the Group 4 horses. For example, say a $200 wager would reduce his group 4 ROI to 30%. Our hero would then be making $15,000 per year profit and yet only risking one half the money.

Mike Nunamaker

[hurrikane]

great thread,

hopefully over the weekend I will have a chance to digest it all. One thing that strikes me is what Mike N says. It takes a minute to realize that 100 bets at a 20% ROI a year is not nearly as nice as 3000 bets at 12% ROI.

One thing about Benter. He is betting into big pools. That's why he left here a went there. Also, he is calculating a defined quantity/quality as opposed to the nonsense here. Not so many class differentials and the US bs.

I love this stuff.....thanks everyone.