Re: Robo Betting (1352 Views)
Posted by:
Mathcapper (IP Logged)
Date: August 14, 2013 05:59PM
Hi Frank – Agree with your rationale for playing the horizontal wagers (see study results below). Have to disagree though on those wagers not being a reflection of the prior race odds (see same study results below). The computer guys use a couple of different techniques to estimate the likely odds of upcoming races for which they cannot see into the pools. On a related note, some of the computer guys are indeed just arb guys looking to bet as much as they can and collect the rebates. Many of the teams however, actually have a great deal of “handicapping” programmed into their algorithms. They typically use a multinomial logit model composed of upwards of 100 factors (ie. performance figures, trainer stats, jockey stats, finishing position, etc), often refined using things like exponentially recency-weighted averages, to assess each horses true probability of winning. And while some of them have indeed gone bust trying to get their systems to work, many have gone on to launch what are widely considered to be the most successful professional gambling ventures of all time (even Benter and Woods lost their intitial $100k stake before going on to make $100’s of millions apiece).
Miff - I don't know what “charters” are and I’m not a computer guy, but for me - and I think I agree with Rick on this - the “edge” comes from knowing whether or not you’re going to get the value you think you are on the horses you’re wagering on. If you’re looking at keying a horse at 3MTP because he’s an overlay at 6-1 but you know he’s going to go off around 3-1 (and thus be an underlay), that constitutes a nice edge to your bottom line IMO.
There’s also a huge arbitrage opportunity in the Betfair pools, but the liquidity isn’t there yet in early prerace wagering.
RickB - I understand your skepticism, I shared it myself when I first looked at this stuff (btw, when I say “near-perfect,” I’m referring to overall long-term results, not all individual cases).
I first became aware of the merit of using parlays to estimate multi-race payouts (and upcoming race final odds) back in 1993 when Steve Crist gave a presentation at a DRF Expo on the topic with respect to Pick 6 payouts. One point he made is that an adjustment needs to be made to account for the difference in takeout between a win bet and a multi-race bet, along with (and more importantly) the fact that you only get hit with the takeout once in a multi-race wager. Because of this, a multi-race wager should generally pay a good deal more than the straight win parlay.
For instance, at the SoCal circuit, where the win takeout is 15.43% and the Pick 6 takeout is 23.68% with a 30% payout to consolations, the Pick 6 should pay (1-.2368)x(1-.30)/(1-.1543)^6 -1 = 46% more than the straight win parlay on noncarryover days.
Similarly, at NYRA, which has a win takeout of 16% and a Pick 6 takeout of 15% with a 25% payout to consolations, the Pick 6 should pay (1-.15)x(1-.25)/(1-.16)^6 -1 = 81.5% more than the straight win parlay on noncarryover days.
For the record, here’s the Pick 6 results of the 1993 winter Aqueduct meet Crist examined, along with the other meets I’ve tracked:
Aqu’93 (39 noncarryover days): Estimated average payout was $15,368 (49.1% more than straight win parlay); Actual average payout was $14,265.
Dmr’97 (33days): Estimate: $25,498; Actual: $25,959
Hol’97 (29 days): Estimate: $23,760; Actual: $20,289
Bel’98 (38 days): Estimate: $13,524; Actual: $10,196
Dmr’98 (32 days): Estimate: $39,570; Actual: $31,804
Sar’98 (25 days): Estimate: $25,656; Actual: $20,212
Sar’12 (27 days): Estimate: $18,275; Actual: $14,641
The correlation is even better when looking at the average payouts for Pick 3’s, Pick 4’s and Pick 5’s:
SA/Hol’12 (80 days): Pick 3 Estimate: $75; Pick 4 Actual: $72
SA/Hol’12 (80 days): Pick 4 Estimate: $123; Pick 4 Actual: $114
SA/Hol’12 (80 days): Pick 5 Estimate: $516; Pick 5 Actual: $509
The same concept applies to the Daily Doubles, which at NYRA should pay on average 15.5% more than the equivalent win parlay (at SoCal they should pay 8.11% more).
The correlation between the odds based on the doubles pool and those of the win pool allows me to estimate final odds for horses in every race that has a daily double associated with it. There’s no rocket science involved – it’s just a matter of plugging the correct equations defining the mathematical relationship into an Excel spreadsheet. It’s a closed-form solution, but you do need to look at all of the combinations in the pool (at least for those involving the first few favorites from the first leg anyway), because the doubles-based odds are interdependent. If you base final odds estimates on the win pool odds from the first leg rather than those of the doubles, the results will be less accurate, especially when dealing with a long-priced winner in the first leg.
I don’t keep records on their precision, but after doing it for thousands of races now, I’ve found that the majority of the time the estimates are very close to the final odds. Of course any one individual horse’s odds can vary a bit (for instance a horse that is 8-1 in the doubles may go off at 10-1 in the win pool) but when looked at on an overall basis over thousands of races, the correlation is excellent. You see more variation at high odds, but that’s simply because the difference between the winning probability for a horse at say 35-1 and one at 45-1 is very small.
I have toyed with the idea of tweeting the final odds estimates at the beginning of each race to help Serling and company with their prerace prattle about this and that horse’s value based on odds that are nowhere near where they’re going to end up, but it’s a lot of effort on my part and I doubt they’d pay any attention to it anyway – LOL. I wouldn’t mind doing it for a day’s card though if anyone’s interested in seeing how they stack up to the true final win odds. Or if you’re ever up at the Spa seminars, just ask me and I’d be happy to tell you more about it.
Btw, you can find a good deal of the academic literature providing supporting evidence for the efficiency of wagering pools in general amongst the papers presented in an anthology called, “Efficiency of Racetrack Betting Markets:”
http://www.amazon.com/Efficiency-Racetrack-Betting-Markets-Edition/dp/9812819185/ref=sr_1_1?ie=UTF8&qid=1376494947&sr=8-1&keywords=efficiency+of+racetrack+betting+markets
Best,
Rocky