Kelly Revisited

[TreadHead]

Not to belabor the point, but your \"worst case\" scenario depicted a person who was absolutely perfect at identifying fair value even money horses, because those horses won 50% of the time.

I would love to see a significant amount of data captured around horses that were identified at fair value of even money and how often those horses won, by person, to see who is truly good at assessing fair value and who isnt.  Where you assessed the value before the race and documented it, regardless of whether or not you placed a bet, and then how often those horse actually won.

If you got 100 astute handicappers and did this long enough, say a couple hundred even money horses identified per person, Im betting that the results would be all over the map in terms of %won.

Although if you averaged all of them out, you are correct that it might approach something like 50%, that doesn\'t help ME as an individual in following the criteria if my specific results are something much more or much less than 50%.

It\'s too bad it is not practical to do a study like this, because I think the results would be very interesting.

[Boscar Obarra]

Not my observation.  You keep saying it, but I look at these quite often and there is wide variance in the \'will pay implied odds\' vs the wins odds in the next race.

 Better than nothing but not reliable at all if you\'re looking for near exact odds to use.

[Mathcapper]

Boscar - How exactly are you calculating the implied odds?

If you\'re using the first leg winner's odds along with the DD Will Pays from that winner it\'s inadequate.

For one thing, if that winner\'s odds weren\'t quite the same in the win pool as they were in the DD pool, then the implied odds for all the horses in the 2nd leg will be off.

The implied odds can also be off when you only use the DD Will Pays for just one horse, especially if the winner of the first leg was a longshot (favorites in the 2nd leg tend to get underbet in the doubles, while longshots get overbet). I\'ve found that using an average of the first 3 favorites from the first leg yields the best results.

Also, you need to make sure you\'re making the necessary adjustment for takeout when calculating the equivalent win parlays used to estimate the implied odds.

At NYRA tracks, the DD\'s should pay +15.5% more than the equivalent win parlays because of the difference in takeout (16% on win bets, 18.5% on DD bets) and the fact that you\'re only getting hit with the takeout once:

DD payout = (1-.185)/(1-.16)^2 - 1 = +15.5% more than equivalent win parlay

The process above really necessitates the use of a spreadsheet, but it is a closed-form solution and I find very good overall correlation with it. There are always some horses that are a bit off, but getting individual precision is not what's important, it's the overall averages that are important. For instance, when I look at all the horses that were 7/2 in the doubles over time, I find that the average final odds for all those horses was exactly 7/2. Same for 2-1, 10-1, 4/5 and pretty much everything else.

Please forgive me if these are things you're already aware of and have already taken into account.

Best,

Rocky

[Mathcapper]

TreadHead Wrote:
-------------------------------------------------------
> I would love to see a significant amount of data captured...where you
> assessed the value before the race and documented it, regardless of
> whether or not you placed a bet, and then how often those horses actually won.

Tread – This is exactly what you need to do for yourself if you want to use the Kelly Criterion effectively.

What you almost certainly will find is that your overlays are not winning at the rate you think they should be.

Until you've gotten your handicapping to the point where you see that your overlays are winning close to the rate at which your fair odds imply, you don't want to even think about using the Kelly formula (or any other progressive wagering strategy for that matter).

[miff]

\"From what I understand, Benter has been playing in the North American market for at least a decade now. He has a staff of only 3-4 people and has said himself that a typical team wagers around $1M per day across roughly 150 races, or around $350-$400M per year. Woods never left Hong Kong\"


A computer whale rebater out of the Dakotas claims there is no way for US pools to absorb the volume of a Benter type handle.Know of at least two math based computer software groups that went bust in the last few years in spite of very sophisticated \"Benter\" type software being employed.

Believe TGJB may also know of bust outs with a similar profile.Bee says the Bowery upscale now,so I\'ll guess the \"math/computer\" horse players
are now hanging under the Brooklyn Bridge.

Given the onerous takeout,pool size and the ever present X factor, US racing is indominiteable regardless of any system/discipline.

[Mathcapper]

Not sure if or how the US pools are able to absorb that kind of volume, but those words came directly from Benter\'s mouth when he was asked what a typical computer team might bet in the North American market.

I\'m not surprised to hear other computer teams have gone bust. Many would-be copiers have tried but could not get their R-squared high enough to show a consistent profit.

As far as the U.S market being indomitable,according to Benter that\'s not true:

\"Teams are playing around the world in most of the big racing markets very successfully. I made a telephone call last night to a fellow and asked him if he\'d ever come across a racing market he couldn\'t beat with a computer system and he said no. In all of the ones he tried, which were many, they\'ve always been able to beat it using one of these systems. So this seems to be something which works more or less everywhere.\" - Bill Benter, 12th International Conference on Gambling & Risk Taking

[Boscar Obarra]

Yes, I am aware of all those factors.

 And while it may be true that on AVERAGE, the odds converge, individual races will have a wide variance, which was my main point.

 I don\'t think Kelly works well under that scenario, but since I don\'t use it, I leave that problem to those that do.

[Boscar Obarra]

I see some very large batch bets come in, at all mtp times , that routinely CRUSH the returns on exacta/dd combinations.

 Unless these combinations were massive overlays to begin with , I don\'t see how they are not squeezing the value right out of them.

 Something paying 200 when it would have paid 350 without you, well, you better be good.  Could be they are agnostic about the payoffs, concerned only about  the net return on the race.

[miff]

\"Could be they are agnostic about the payoffs, concerned only about the net return on the race\"


Box,

You just described the rebate whale who is working with a software program.

Mike

[Boscar Obarra]

Yes, I knew that, but still, on occasion , they absolutely crush certain numbers to the point of absurdity.

 And they rarely win.

[Mathcapper]

It may not be just the arb guys trying to grind out rebates. What could also be  happening is that because these fundamental modeling-based computer teams are betting so much money, pool size is their limiting factor. So rather than using the Kelly formula of edge/odds, they use a formula that maximizes their expected value.

So if they assess fair odds for a particular exacta combination at 12-1 and it's currently 20-1, they don't just bet it down until it's no longer 20-1, they bet it down to the point that maximizes their expected return, which might be, say 14-1.

The specific formula used is quite daunting, so don't say I didn't warn you, but here it is:

Maximum EV = -A_i + sqr rt{[A_ip(Σ(A_j) – A_i)(1-T)]/(1+p(1-T))}

Don't ask me what the variables mean because I have no idea, all I know is that it maximizes their return on the bet.

[miff]

Box,

Agree,but the pro rebater is so far hedged into the pool that the \"crush\" of a certain combo doesn\'t matter. At the end of the race,if they can collect 90+% of their total wager outlay,they beat the race, assuming a rebate of 10+%.Huge grind, no guarantee of long term profit.

If you are a player, light a candle for all whale rebaters who add great liquidity and depth to the shrinking US pools, down app 2 billion from its peak.

Mike

[Mathcapper]

Boscar – I understand your point, but that kind of individual-race precision is unnecessary. No one's ever been able to be that precise. There have always been last minute changes in odds, yet Kelly practitioners have been and continue to use the formula with great success as long as they know that their overall edge is what they think it is.

Some Kelly users don't even look at individual races. They base their bets on their overall edge and their overall odds. For instance, they may know that their long-term ROI on win bets is +10% and that the average odds of all the horses they wager on is 5/2. So their Kelly bet is the same for every race: +10%/2.5 = 4% of their bankroll. This isn't the best way to utilize Kelly, but it's another way to do it.

Regardless, the final odds of any particular race isn't all that important. If you have a +10% edge and you're only betting what you think will end up being 5/2 shots, it doesn't really matter if one particular horse goes off at 2-1 or 8/5, while another one goes off at 3-1 or 7/2. That just means that sometimes your edge is a little bigger than what you thought it was going to be and sometimes it's a little smaller. What's important is that on a long-term basis, your edge really is +10% and not -5% or -10%. The fact that they go off at 5/2 on average indicates that you have optimized your returns using the Kelly formula.

[TGJB]

I\'ll tell you, man, if the NYRA thing and that insane surtax on ADW bets go through, the game may be in real trouble. Overall this is one amazing clusterf--k, how not to run a business. How to run it into the ground.

[miff]

Agree with much of what Crist says but NYRA needs to get real. The self praise reaped upon itself by NYRA CEO Chris Kay during the \"average\" statistical SPA meet is transparent and insulting to the the many who support NY racing at the windows.An all out NYRA shill-fest from CEO Kay to the talking heads on the TV feed...NYRA racing is wonderful.Many of us who wager $millions do not agree!

It seems that NYRA\'s primary mission is to get the financials in order so that Cuomo/Skorton/Megna et al can execute their plan to jettion NY racing.


DRF

Steven Crist:State making NYRA a tough sell

People who work at the New York Racing Association understandably bristle when anyone says that the annual Saratoga meeting is essentially a foolproof operation that runs itself. When I worked there more than 15 years ago, I bristled, too, arguing how hard the employees worked and how much time and effort it took to put on the show.
 
The truth is that both of these arguments are correct. Everyone at the NYRA does work hard during Saratoga, but it's still pretty hard to mess things up. Yes, there's too much racing and particularly too much bad racing, and a host of small annoyances that customers must navigate, but the numbers year in and year out suggest that the whole proposition works pretty well regardless of who is running the place. Over the last three years, Saratoga has posted essentially the same business numbers under three entirely different management scenarios – under an entrenched and experienced management team in 2011, under a leaderless and shorthanded crew in 2012, and under a whole new board and chief executive for this year's 40 days of sport.
 
The 2013 Saratoga meeting will be best remembered for two things:  extremely good racing in its biggest events, as is often the case, and almost freakishly good weather, which is rarely the case. Just as the weather finally returned to its usual poor form on closing weekend, however, the meet also ended on some unsettling notes – not on the racetrack, where stellar performances by rising stars in the juvenile ranks concluded the meet on a hopeful note, but in the boardroom, where the murky intentions of New York state, which took over the NYRA a year ago, became a bit clearer in an Aug. 28 board meeting.
 
Most of that meeting was devoted to an excessive video tribute to the meeting and replays of the many ceremonies and photo-ops staged by new management, but things suddenly shifted gears toward the end. The state's representatives laid out a specific goal of making NYRA profitable without the Aqueduct racino revenues that have buoyed the game for the last two years, a prelude to a reprivatization plan that is supposed to be formulated by 2015.
 
Reprivatization is either the light at tunnel's end, or an oncoming train, depending on what you think it means. The best-case scenario is one in which the state realizes it has a good thing going and that the best plan ahead is to keep NYRA a non-profit operation, putting any profits back into the game, preserving a cherished cultural institution, and continuing to support hundreds of thousands of jobs in direct and related industries. The worst involves the delusional notion of some politicians that the enterprise can be repackaged and transformed into something that can be sold to private operators.
 
It seems to me that we already wasted more than a decade on the latter notion, when the state did everything it could to discredit NYRA and put the franchise out to bid. For those who have already forgotten, here's what happened: Nobody was interested in bidding unless they could operate casinos at the tracks and keep most of the profits. Those bidders professed some phony affection for the game that most people saw through from the start, and when it became obvious that there was no profitable racing enterprise to be sold, every interested bidder dropped out except one – the NYRA, which actually wanted to continue putting on the best racing and use casino profits only to improve the sport and the facilities.
 
Nothing has changed to make some sort of "sale" of New York racing any more viable in 2015 than it was in 2005 or 1995. The land and the possibility of doing something different with it are always going to be more valuable than anything the racing can generate. Any plan to sell New York racing is merely a long-term plan to replace it with something else.
 
Even if it were an attractive business proposition, which it never will be, you would be buying a phantom – something that can be transformed overnight by legislators from marginally profitable to virtually worthless. State government has repeatedly proven that even written contracts guaranteeing racing days or casino revenues can and will be changed at the whim of any incoming administration.
 
Still, here we go again. NYRA's primary mission apparently is going to be cutting costs, despite finally having the revenue to make critically needed improvements, so that the state can claim it has worked its political magic and created something of value to sell. It feels like the beginning of a long and doomed charade.