|
JCapper Message Board General Discussion
|
| Home |
Register
Log In |
| By | Expected Value - When is it there and how do you know? | |||
| jeff 10/30/2008 11:47:59 AM | Post in progress - check back later for more... From another thread:
Ralph, please understand that I'm not picking on you. This post is going to grow each time I come back to it. By the time I'm done I'm sure it'll be almost book like in length. My intent isn't to argue or disprove anything you've said. Rather, I'm trying to post a little "hands on" instruction in how the Live Play Module can point JCapper users towards good bets and help them avoid bad bets. I think there's something truly important embedded between the lines in the above quote. Screenshot(s) of Live Play Module: http://www.jcapper.com/images/haw10292008r9.jpg Take a look at the above screenshot. Forget for a second my UDMs which are used to emphasize things I discovered in the Data Window. Forget for a second my own UPR. Even without my own personal adaptations to the program Flurtini is straight out of the Reading The Reports Help Doc and/or Finding An Edge... only she isn't. If you must know she wired her field on the turf at HAW about an hour ago at 20-1. A few more: Ball Four http://www.jcapper.com/images/kee10252008r7.jpg Cure For Sale http://www.jcapper.com/images/kee10172008r6.jpg Whatsat http://www.jcapper.com/images/haw10182008r8.jpg Desert Code http://www.jcapper.com/images/osa10252008r2.jpg Horses like this/these don't happen every day. But you CAN find several each week worth playing. One of the keys to using JCapper (or any program or methodology) is discipline. Once you've done the R&D and identified a valid play type - and made the decision to venture into the arena of live play - you have to let the game come to you. Unless your UDMs are geared around super-chalk (which mine aren't) you can't make marginal bets (the 9/5, the 5/2, the 3/1, the 7/2) and expect to come out ahead. Years of experience have convinced me that the following holds true for every viable UDM or methodology: If you break things down and make a careful analysis using the Data Window you'll find that roughly 20 pct of your plays produce about 80 pct of your profits. The better the job you do of identifying and eliminating the 80 pct of your plays that fail to produce profits - the stronger the plays that you do make will be. IOW, separate the wheat from the chaff. You HAVE to pre-identify and wait for the good plays only. Q. How do you define a good play? A. That's easy. A good play is one where the player has positive expected value. Everything else is a bad play. In my opinion, for the majority of players out there, an extremely tiny pct of the overall plays made actually fits my definition of a good play. As players we crave action. We're wired to jump all over our selections even when they go off at at odds too low to create positive expected value... In my early years you wouldn't believe how hard it was not to bet my selections at 2/1, 3/1, or 7/2... even though a careful analysis with the Data Window (broken out by incremental odds ranges) would show those plays to produce a negative roi. Getting beyond the "I'm here to gamble. I just wanna bet horses!" urge was really hard for me. But get beyond it I did. And get beyond it you must... IF you want to win. 99.9 (or some similarly high pct) of all players fail to get beyond it at all. One of the things I practice (and try to teach) is the idea that it's ok to sit on your hands until a situation with positive expected value comes along. Then (and only then) is it time to bet. And when you do bet, go for the throat. When you have a strong opinion on a horse: back it. Take a swing and try to hit one out of the park. Play exotics when value is there. But make sure that at least 2/3rds of whatever money you put into any race goes into the win pool. Understand that they don't all win. In fact (for horses like the ones I'm showing) less than 1 in 5 wins its race. Your job as a player isn't to win races. Your job is actually pretty simple. It's all about expected value. Your job is to make good plays while avoiding the bad plays. Do that often enough and the winning and losing part pretty much takes care of itself. Let's talk about the Live Play Module readouts for a minute... In the above screenshot(s) take a peek at the Expected Value column for Bettor's Tote Prob. ALL of them are well over the cutoffs shown in the BettingInstructions field for my UDMs. For Flurtini, Expected Value Bettor's Tote Prob at post time was 2.73. The lowest EB cutoff determined by my own Data Window R&D and recorded in the BettingInstructions fields of my UDMs is 1. 2.73 is well above my min cutoff of 1. The decision to play was an easy one. Had Expected Value Bettor's Tote Prob at post time been less than my cutoff the decision would have been equally easy: PASS. Thats' how it works. Q. Do I play every horse where Expected Value Bettor's Tote Prob at post time is greater than one? A. Not only no but hell no. I do this for UDM horses only. If you try this with all horses everywhere you are going to get killed. Why? In Finding An Edge I introduced a concept which I called hidden positives. Simply put, in an ALL button Data Window query broken out by Expected Value BettorsToteProb you'll see a lot of horses where Expected Value BettorsToteProb is greater than one. The reality is that very few of those horses have hidden positives in their records. As a result, if your sample has any size to it at all, your query results are very likely to show a negative roi. I try to build my UDMs in such a way that they only identify horses with hidden positives in their records. If I do a good job of that I find that when I run my UDMs through the Data Window broken out by the various Expected Value numbers in the program I can always see cutoff points that produce profitable play so long as I stay above those cutoff points. Simply put, horses with strong hidden positives in their records usually show positive roi in Data Window queries broken out by the program's expected value numbers. Once I find expected value cutoff points in the Data Window, I then record them in the BettingInstructions field for each of my active UDMs. During live play I stick to my cutoffs. It's the easiest way I know of for separating the good plays from the bad plays at crunch time. Sorry for being so long winded guys. But putting an accurate a description of what I do to words isn't always easy. I hope I explained things in a way that makes sense. -jp . ~Edited by: jeff on: 10/29/2008 at: 7:00:37 PM~ ~Edited by: jeff on: 10/30/2008 at: 11:37:45 AM~ ~Edited by: jeff on: 10/30/2008 at: 11:47:59 AM~
| |||
| JustRalph 10/31/2008 3:35:41 AM | you keep posting, I will keep reading Thanks
| |||
| jeff 10/31/2008 6:06:52 AM |
A. I evaluate each according to the expected value that it represents. I don't care how many UDM selections I have in a given race. If the selection itself has an expected value that is higher than my cutoff as determined by Data Window R&D I'll bet that selection every time. If a selection has expected value below my cutoff point I'll pass that selection every time. When multiple selections are present at at or above my cutoff points I prefer dutching them as opposed to betting equal amounts on them. Remeber, the lower the odds (generally) the higher the probability of a win. I much prefer proportional win betting for consistent bankroll growth to the roller coaster ride that comes from betting equal amounts on two horses in the same race... where one is 20 pct likely to win and the other is 5 pct likely to win. In the Breeder's Cup Contest thread I posted my UDM selections and proportional win bet amounts ahead of time for Saturday's Breeder's Cup Card. I didn't know what final odds would be. The win bet amounts I used in that thread were based on dutch win betting UDM selections I thought would produce positive expected value at the odds I thought they would end up going off at. I was off by a long ways on estimated final odds for several of those selections. But the win bet amounts I posted produced a profitable day. In real life my strategy is not all that different from what I did in that thread. But there are a few differences. The thread called for me to bet $50.00 in every race. The reality is that I generally find one maybe two horses worth betting (think positive expected value) on any given race card. So I end up passing almost every race that I look at. The other difference is that I generally wait until there's maybe 2-3 minutes to post and base dutch win bet amounts for multiple contenders when I have them on what I think final odds will be from what I see at that point. Understandably the closer you get to post time the more accurate your final odds assessment can be. But it's never exact. With very few exceptions horses 4/1 and below tend to drift lower in the final minute of betting. How much they drop is strongly tied to how pretty they look in the flesh and on paper. We've all had horses go into the gate at 9/2 or 5/1 only to watch them win and pay $5.00 or $6.00. It's really frustrating to make all the correct decisions only to see your odds get cut in half during the actual running of the race. But it happens the other way too. Despite much of what I see posted around the web, many horses at 10/1 and higher actually go UP. Cure For Sale (posted at the top of this thread) almost doubled in odds in the final flash while they were on the backstretch. This in spite of the fact that the horse broke on top and wired the field. I firmly believe estimating final odds is a skill that can be learned. Understand that you'll never be exact. But you CAN consistently be close enough to generate profits. And in reality that's all you need. -jp . ~Edited by: jeff on: 10/31/2008 at: 6:06:52 AM~
| |||
| DeanT 10/31/2008 12:47:56 PM | Jeff, what do you do with an apparent no go horse. Is there any data that shows (I think you mentioned before that you had some) where it shows we are obviously wrong? For example, the first race at CD and this happens often. ML 7-2 horse, logical, good numbers. It all points to a likely winner. But he is 10-1 on the board. CDX R1 7.0f DIRT 10-31-2008 DAILY DOUBLE / EXACTA / TRIFECTA / SUPERFECTA / PICK 3 (RACES 1-2-3)PICK 4 (RACES 1-2-3-4) 6 +EDIT THIS Shipper_UDM +_Yellow_Hi 9 1.79 .1787(2) 2.79 .2788(1) 1.52 1.61 5.6 .1523(4) 76(1) 3.5 2.57 1.59 2.2 .3099(1) 1.8 bettors tote, 76JPR etc. I dont touch those horses after some experience with them. Should I? Would you have bet this horse? Do you have any numbers that you can share in your big DB that show you that a horse like this (i.e. a near ML fave going off at high odds) is a good bet or a bad bet?
| |||
| jeff 11/1/2008 10:46:40 AM | Dean, that's a very interesting question. My whole approach is geared towards exploiting situations where public opinion and my opinion are drastically different. IMHO, the more different public opinion is from my opinion the better my results roi-wise. Sure, horses like this obviously have something wrong... maybe barn money isn't being bet because they know something the public doesn't, etc. Despite that every once in a while the horse surprises and wins anyway. And when that happens the inflated win mutuel makes situations like that historically profitable. Here are Data Window results for one of my stronger tight model UDMs where play is restricted to odds of 7/1 or higher. The results are broken out by morning line odds:
Ok. That's a tight model UDM which btw excludes horses with really low morning lines. Here's what a loose model UDM looks like in the Data Window when play is restricted to odds of 7-1 or higher with the results broken out by morning line:
And here, with the loose model UDM, the data clearly shows the effect that you're talking about. From that it appears that the strength of the individual UDM plays a hand in determining how you should treat situations like this. I'd recommend verifying each of your own UDMs in the Data Window under such circumstances before reaching any conclusions. -jp . ~Edited by: jeff on: 11/1/2008 at: 10:46:40 AM~
| |||
| DeanT 11/1/2008 1:33:53 PM | Thanks Jeff. Kudos on those UDM's, that is some great work. You can make money at this game yknow? :) I have been messing around with some of that stuff, and have been modeling some of the UDM's based on the new factors. Playing in the live play does help a lot too. That second UDM is pretty amazing. I mean who woulda thunk that? In the UK players on betfair that know what they are doing can make money laying (i.e. betting against) the "drifters" - lower ML horses who drift upwards to 8 or 10 or 12-1 with no apparent black marks on the PP's. I think we do have our own drifters here, and finding those as a part of my live play is a staple to what I do.
| |||
| jeff 11/1/2008 2:29:46 PM | Dean, IMHO info that doesn't make it into standard past performances and/or data files can really help a guy stay on top of drifters. A couple of years ago when I was playing head to head horse matchups at Pinny I noticed their house linesmaker was really good at taking horses that were VERY pretty on paper (but had suffered some injury) and putting them in a matchup against a sound horse that wasn't nearly as pretty on paper. Based on the way the lines moved in those situations I'd say a LOT of unsuspecting bettors were regularly getting fleeced by betting the pretty on paper horse. In almost every matchup of that type the pretty on paper horse would finsih up the track. I can only imagine how much money their guy made for the house in those situations. Line movement told me the betting wasn't even close to being balanced on both sides. One thing I started doing after I noticed that trend was visiting the Equibase site each day and recording the names and dates of horses listed as vet scratches. Inevitably some of those would wind up as the pretty on paper part of a Pinny matchup. Needless to say, having the names of recent vet scratched horses was a big help in those matchups. -jp .
| |||
| DeanT 11/1/2008 3:23:25 PM | I used to watch those as well Jeff. It was amazing! I have put together a little bit of this on paper, and I am trying some things in your program. So far with a couple of things I watch with low ML horses who are dead are working well. Being able to toss a 5-2ML shot that is 5-1 out of exotics and everything is something I think can be profitable.
|
| Reply |
|
Copyright © 2018 JCapper Software back to the JCapper Message Board www.JCapper.com |