Computer-Robotic Wagering

[Boscar Obarra]

Late odds changes have been a part of the game for years, since parimutuels I\'d say.

 And I have some sad news, most of the late odds changes you see are adjustments to prices that are out of line with \'fair odds\' , and not someone making a score.

 That horse at 3-1 you liked, that went 9/5 at the bell? Was never a 3-1 shot to begin with, and was never going to leave the gate at that price. If they shut off betting 2 minutes before post time, it would still go 9/5.

[miff]

Fair Odds Line ranks right up there with overlay/underlay and the tooth fairy.

[JimP]

I don\'t care whether the final odds are \"fair odds\" or not. I just like to have a decent read on what the final odds will be when I bet. I played for many years with pretty good confidence on what the will-pays were. It got more and more difficult in recent years to do that. So I stopped playing. Except for a few big days like the BC and the Triple Crown races where I still play a little.

[Boscar Obarra]

Everyone , especially big bettors , want a good \'read\' on the final odds, especially in the low odds ranges.  

  That\'s why the odds change, if the price is \'right\', then it becomes wrong, as all the late bettors try to pick up the free money.

  This is nothing new, nothing to do with computer betting. The only difference is, in the bad old days, you had to be standing at the $100 Window punching out tickets one at a time.

  Don\'t ya\'ll miss the thrill of holding a winning stack of those yellow ducats?

[Boscar Obarra]

Agreed. The term is used loosely to describe what price people/bots are willing to take, below which they will pass.

[Mathcapper]

JimP,

Have to say, I haven't seen any evidence of a breakdown in the correlation between the Will Pays and final win odds in recent years. And I'm not talking about a small sample either. I track the Will Pays for hundreds, if not thousands of races every year, and have habitually done so for close to 35 years.

Does every horse in every race go off right in-line with its Will Pay? Of course not. But in general, on an overall basis, they come pretty close. If anything, the correlation seems to have gotten better over the years, which may be attributable to the computer guys pushing each pool towards what they consider to be fair value.

Make sure, as Bit noted, that you're not using longshot winners from the prior race in your estimation. There are a couple of other issues related to the limitations of using just the one set of Will Pays rather than the entire set of DD combinations to estimate the odds, but in general, if you're doing it right, you should be seeing pretty good correlation.

For instance, in the last race I did today, the bigA finale, it was clear from the Will Pays before the first flash of the tote was even up that that Pletcher firster was going to go off at odds-on of around 1/4.

These are the estimates I had, along with the corresponding final odds:

#3: 6.6-1  => 6.9-1
#6: 9.2-1 => 10.3-1
#8: 28-1 => 31-1
#9: 21-1 => 18.3-1
#10: 0.3-1 => 0.3-1
#11: 12-1 => 22-1
#12: 16-1 => 9.9-1

These results are pretty typical of what I see on a race-to-race basis. There might be one or two horses that are a little out of line, but overall they're usually pretty close.

As Boscar noted, and as I see happen all the time, the vast majority of these late odds shifts are simply the win odds moving closer into line with the odds in the doubles, or to what is the public's probability estimate. And like he says, that's not something that's going to change if you close the betting 2 minutes to post.

Btw, the giant pools you see on days like the BC and Triple Crown don't necessarily mean the correlations are going to be better. In the 2013 Belmont for instance, Palace Malice was 13-1 in the win pool, but he was 7-1 in the doubles, which is a pretty big disparity even in a regular pool, never mind in the massive pools like we see in the Belmont. When these kinds of things happen, it's sometimes a sign of some "smart money" action, be it inside information/drugs, the computer guys, or even us fellow high rolling sheet players.

The win pool is generally the more efficient pool and where you see the smart money vis-a-vis the doubles, but in this case, I took it to mean that some serious money was pounding the horse's odds down in the doubles to well below the public\'s win probability estimate.


Rocky R.

[JimP]

Mathcapper, thanks for the response. I understand how the DD Will Pays can be used to approximate Win Pool Will Pays. But that wasn\'t the subject I was commenting on. My reference to will-pays was strictly Win Pool related. When the odds I saw posted in the Win Pool no longer correlated to what a Win ticket would pay, I lost interest in wagering. That was the only point I was making. I understand that Boscar thinks that nothing has really changed in that area. Well, I guess we just see it differently. I first started wagering on horse races a half-century ago. In my opinion there was a dramatic change that occurred a few years ago and it effected my play. I also understand that an analysis of a DD pool can potentially provide an indication on what the Win Pool will pay. But that analysis is unreliable in many cases and in fact a DD pool doesn\'t always exist. My only point was that for me, this became a big enough issue that I have essentially stopped playing. That\'s my only point. Just one man\'s opinion. Just one customer lost to horse racing. Maybe I\'m the only one. I\'ve commented on this enough. Probably too much. I\'ll drop the subject and go back to lurking. I enjoy following all the comments on this board.

[Mathcapper]

Jim,

Sorry to see a fellow horseplayer leave the game for the most part. I was hoping I could give you some ammunition to rejoin the battle.

No need to go back to lurking - it\'s always good to hear different perspectives.

Best,

Rocky

[Mathcapper]

GC,

If it's true, I'm surprised that these guys aren't playing the Pick 6, or the superexotics like the tris, supers, and Pk4's, as others in this thread have speculated.

I could understand it if these computer guys were mostly "arb" guys, since they can't see beyond the exacta pools. I keep hearing about such groups and how they\'re able to generate profits off their rebates alone, supposedly without any handicapping(?). I guess they could be arbing the exacta combos to the win pools in the same way that Ziemba disciples arbed the place and show pools, but I don't know anything about them or if it's even true at all.

I do know a fair amount about the PhD-level mathematician teams you mentioned though. A lot of info on these teams is publicly available, often straight from the horse's mouth.

Bill Benter for instance, who is a brilliant and remarkably humble fellow for someone who just may be the most successful gambler of all-time, has given numerous public presentations over the years at various venues like the International Conference on Gambling & Risk Taking, as well as having published scholastic papers on the subject of multi-factor based computer handicapping.

From what they've said, these guys LOVE the superexotics. In his own words, Bill has said,

"Exotic bets offer some of the highest advantage wagering opportunities. This results from the multiplicative effect on overall advantage of combining more than one advantage horse...When you string them together, you go from what may be a very small or marginal advantage to a very large advantage. In ultra-exotic bets such as the pick-six, even a handicapping model with only a modest predictive ability can produce high-advantage bets...For a bettor in possession of accurate probability estimates which differ from the public estimates, 'the more exotic the bet, the higher the advantage'...Some professional players make only exotic wagers to capitalize on this effect."

He even gave a presentation at one of the ICGRT conferences dedicated entirely to how to play the Triple Trio (3 consecutive trifectas) bet in Hong Kong.

He acknowledges the difficulty of projecting the public's betting in pools they cannot see. This is made more difficult by the fact that the public doesn't bet horses in the superexotics in the same proportion to the win pool (they tend to overbet favorites). They deal with it through things like empirical research about public betting behavior instead of trying to predict it apriori based on assumptions of rational pubic behavior.

They also use things like the discounted Harville formula to estimate vertical superexotic probabilities, which is exactly what I use myself to estimate payouts for exactas, tris and supers.

As far as whether these computer teams have a deleterious effect on the rest of the public, Bill freely admitted that this is certainly so. Like Jerry said, it's a zero sum game.

By Bill's own estimate, the effect of the computer teams is to increase the effective takeout by around two percentage points. He also says that the maximum amount a computer team can expect to extract from any given racetrack is about ¼ of a percent of the total pool. And as was pointed out elsewhere in this thread, they're not concerned with ROI but rather the net increase in their bankroll. They make wagers on all bets they deem to be advantage bets as determined by their expected value, up to the point of their maximum expectation (I've posted both the EV formula and the Maximum Expectation formula they use on this board somewhere before).

Much of the effect of these teams can be seen in the declining odds on favorites over the years. There used to be a fairly well-pronounced phenomenon known as the favorite-longshot bias in U.S. racing. Favorites, particularly odds-on favorites, were underbet by the public. Not enough to show a profit, but enough to maybe cut the track take in half or so. Not so anymore. Favorites now win at a higher rate, and pay lower prices, and the bias is now virtually gone. This effect is further compounded by the fact that the teams get substantial rebates, meaning they can drive down the prices even further below what they would normally be able to do and still show a profit.

The money these computer teams are siphoning out of the pools by sucking the value out of overlays is certainly detrimental to the rest of us, but it doesn't mean the game can no longer be beat.

Yes, the pools are larger, and overlays aren't absolute. One man's overlay is another's underlay. As far as I know, these teams aren't using performance figures, specifically not Jerry's. There are still plenty of overlays to be found, especially when you string them together in the superexotics, they might just be a little harder to come by.

Rocky R.

[Wild Again]

Rocky

I\'m not throwing stones.  As far as the gambler is concerned, horse racing is not a zero sum game.

It is a negative sum game.  The takeout is deducted from the pool.

The difference between a zero sum game where the computer robotic wagering systems have a slight edge thru ability to place bets late and a negative sum game where they receive a rebate and can always act last is a considerable difference.

While all my words may be assertions i.e. I have no proof.  If it walks like a duck, and it talks like a duck, it ain\'t a mongoose.

Computer Assisted Batch Wagering is bad for me. A small time recreational player.

Thanks

John Perona

[miff]

If all venues ban the bots,handle goes down by app $2 billion assuming bots don\'t play anymore.Pool liquidity already an issue at many venues esp during slow times, like now thru at least Dec.

The bot issue has more moving parts than one would think.

[BitPlayer]

That\'s the second time I\'ve noticed pool liquidity mentioned in this thread.  It\'s not clear to me what it is or who it benefits.

If liquidity refers to the ability to make a large bet without significantly affecting the odds, wouldn\'t the bots need it more than the recreational player?

As a recreational player, isn\'t it better for me to bet without the bots and risk the effect that my bet will have on my edge than to bet with the bots and not have an edge to begin with?

My theory on the economics of it all is that the net takeout on bot bets is the price the bots are paying the track for the privilege of taking unsophisticated money (that would include mine) from their pools.  It works for the tracks (or at least they think it does) because many players budget how much they are willing to bet rather than how much they are willing to lose.

[Mathcapper]

John,

If I implied that computer teams are not bad for the overall player, that was not my intention.

By \"zero sum game,\" I was referring to the fact that if one group is winning, by definition that means the rest of the public must be losing that much more. In effect, the computer guys act as an additional rake on the pool. The track takes their 20% or so out, then these winning computer teams take out another chunk.

As I mentioned, Bill Benter himself estimated that the teams have the effect of raising the takeout on the rest of the public by around 2 percentage points. So yes, that\'s certainly bad for the player.

What I wouldn\'t necessarily agree with is that their edge is due solely to their ability to place bets late and to receive a rebate.

The teams that I know of use sophisticated 80+ factor multinomial logit models to help them identify positive expectation bets, and they bet these overlays accordingly. Big money sharpies, whether they use computer-assisted wagering or whether they\'re standing at the $50 window watching the gate close,have always tried put their bets down at the last possible second to hide their action from sycophantic bettors.

Yes, the rebates do allow them to bet down overlays more so than the general public and still show a profit. That may be part of the reason why we\'ve seen the prices of winning favorites decline over the years (along with the neutralizaion of the favorite-longshot bias I mentioned in another post). Compared to the fact that these teams are sucking the value out of overlays across the board in general though, I would think this effect is rather minimal.

If they\'re actually using CAW to past-post however, that\'s an entirely different matter.

Rocky

[GChandler]

CAW players leaving the game would likely cause T\'Bred handle to fall by 3 billion (annualized) almost immediately.  The effect on tracks like GP, AQU & SA would be less than at some other tracks but still noticeable.  Smaller tracks like  HAW, TuP, TP, MNR, Los Al,etc would close within 60 days because they would be unable to pay purses.  All of those cheap horse have to go somewhere and since many them can\'t compete anywhere else a few thousand would have to be put down.

Therein lies the problem.   You can say that without the bots more players would be attracted to the sport.  Not with an average takeoput of 21% they wouldn\'t, not to mention the fact that we have already lost 1-2 generations of gamblers.  There isn\'t anything you could to get large numbers of 18-30 years that have never bet horses to take up the game.  Maybe, just maybe if you lowered the takeout to 10%, which neither the tracks nor the horsemen would support, you could get the generation behind the 18-30 group back to the game.  Maybe.

To borrow from Churchill \"there\'s nothing worse than having CAW players in the pools except for not having them at all.\"

[miff]

Exacta in 7th at Aqueduct only paid $459.00. Guess the bots knocked it down to the \"fair odds\"