Money

[crestridge]

This is a question for Dave Schwartz: I know you have discussed money management in detail, and it has been discussed in detail on this forum, and I'm not intending to open "the box" on this, but what is your current thoughts on the Kelly criteria? Do you use your HMI system or do you now use "Kelly" or something other? Just curious.

CREST

[Overlay]

Here's a quote from Dave (on 30 August of this year) in the "Change Your Game Forever" thread that he started (http://www.paceadvantage.com/forum/showthread.php?t=105408):

"IMHO, Kelly doesn't work. It can't work. Oh, it works in theory but not in reality.

Kelly is based upon 2 things: Profitability and odds. Thus, an advantage of 20% with odds of 2/1 calls for a 10% wager. We all get that. Great formula.

However, we NEVER really know the hit rate - it is only our perception.

And we NEVER really know the odds - because they change so much after we've made our bets."

[thaskalos]

Here's a quote from Dave (on 30 August of this year) in the "Change Your Game Forever" thread that he started (http://www.paceadvantage.com/forum/showthread.php?t=105408):

"IMHO, Kelly doesn't work. It can't work. Oh, it works in theory but not in reality.

Kelly is based upon 2 things: Profitability and odds. Thus, an advantage of 20% with odds of 2/1 calls for a 10% wager. We all get that. Great formula.

However, we NEVER really know the hit rate - it is only our perception.

And we NEVER really know the odds - because they change so much after we've made our bets."

I think that Dave's opinion has shifted on this subject.

I remember that he was a proponent of Kelly-type wagering not long ago.

[Overlay]

I think that Dave's opinion has shifted on this subject.

I remember that he was a proponent of Kelly-type wagering not long ago.
How recent is the "not long ago" that you refer to? The quote from Dave that I posted was from less than five weeks ago.

[thaskalos]

How recent is the "not long ago" that you refer to? The quote from Dave that I posted was from less than five weeks ago.

What I meant was that he used to think Kelly wagering was the way to go about a year ago...and that was the wagering method he used in his own play. I remember him saying so when he was talking about his "Renegade Handicapper" product.

He has obviously changed his mind on the topic of Kelly recently.

[classhandicapper]

The only thing that works well for me is "flat" bets.

It's hard enough for me to handicap a race, form an opinion, evaluate how strong an opinion it is, think about what price I need to make a play, figure out what pool to make the play in etc... I don't like adding another complication to the mix like trying to figure out what the bet size should be or to try to use those things to determine it.

[Dave Schwartz]

The answer is, "Yes, I am for Kelly." My second answer is, "I am not a fan of Kelly."

How can it be both ways? I am for it and also against it?

The answer is that the typical application of Kelly is impractical, as per Overlay's quote of mine from above. (Thank you.)

It does not work because (as I stated) the two components we need - hit rate and payoff - we do not have. Therefore, betting a percentage of your bankroll does not work in the sense that the wagers will be optimized (as promised).

However... I believe Kelly can work in the sense that you will wager larger amounts on more profitable/consistent horses. The key is that the wagers must be relative to some STATIC number or against a CURRENT SESSION bankroll.

In other words, Kelly can work for you if you have done the homework and have realistic expectations about what it will do for you. If you want it to create relative bet sizes from one race to the next, it will do that. If you want to do it relative to your bankroll (as the promise of Kelly would lead you to believe)... well, that just does not work.

As for my own play... we are just starting to play on a serious level. During our testing we could not use HMI because it would mask the true $nets of the test (since HMI becomes a component of the result). When we begin playing (later this month I hope) we will use HMI.

[DJofSD]

Dave, thanks for the information.

Are you referring to fully or partial Kelly?

[Dave Schwartz]

Doesn't matter. It is all a work of fiction.

There is "Full Kelly" and there is "percent of bankroll." The point is that to OPTIMIZE when you bet a percentage OF BANKROLL you must KNOW 2 things:

1. Hit rate
2. Odds

We have a PERCEIVED hit rate but we have nothing that resembles accurate odds.

Thus, the entire concept of Kelly (which is optimum bet size) is out the window for horseplayers.

Still serves a purpose.

Perhaps I misspoke earlier. To be more clear, I use a version of the Renegade Handicapper approach along with something new to determine what my bet size should be in this race relative to available bankroll.

"Avilable Bankroll" is determined by HMI.

[crestridge]

Dave

Thanks much for offering some insight into this most important topic. It seems to me, if one could know one's edge, and develop an approximate end times odds algorithm (which maybe impossible), one would optimize growth of bankroll. It's a tough nut to crack.

[TJDave]

However, we NEVER really know the hit rate - it is only our perception.

And we NEVER really know the odds - because they change so much after we've made our bets."

That wouldn't be true for whales, would it?