Holding my breath...fully automated betting going live!

[CBedo]

It took me way longer than I had originally thought, but it actually happened today. Today was the first day of beta (half day roughly) of going live with fully automated, hands off handicapping and betting. :jump:

As many of you know, I've been working on a new model since June, and although it still needs lots of work, it is promising enough that I wanted to try to automate my process a bit since it was pretty onerous to get from data input to bet output. Since I was trying to automate that, I decided to see if I could automate more in the whole process.

From beginning to end, here is a summary of what I (or most all handicappers) do every day.

1) Get data (whether it be data files, pdfs, or the paper form).
2) Handicap a set of races.
3) Adjust any playable races for scratches or other changes (some might wait to do 2 until after this).
4) Wait for the playable races, and if there is value make a wager either live or at a ADW.
5) Collect some money (hopefully), and then do some accounting and record keeping.

1 & 5 take relatively little time (reviewing yoiur records is another matter). The handicapping can take a while, and waiting for the scratches/changes, and then the actual race takes forever (could be all day if you have enough tracks to play). In order to have more time to actually do some more research and other handicapping, I wanted to see if I could make the process as hands off as possible.

Here is the current automated process:

1) Automatically download the days tracks data files early in the morning, and prepare the data for modeling. Put data into a database.
2) Wait until the scratches are online (or until very close to post time), and then use information from the database to process the model which outputs possible bets and value needed to make a wager.
3) Upload conditional bets or watch toteboard and place bet with ADW near post time.
4) Update records and send some reporting to my email.

It's a little scary to "let go of the wheel" so to speak, but it was pretty exciting as well. It helped today that the first race was a 16.00 winner. Technically, I could leave right now, and it would be able to bet races every day without any human intervention (until it goes broke I guess, lol). I'm going to "oversee" it tomorrow, and then we'll see what happens!

[LottaKash]

Fascinating, the "ulitmate black box".....:eek:

Say more later, please.....:jump:

best,

[HUSKER55]

I wish you well. That would require a little more nerve than I got. Keep us posted. Sounds fascinating

[CBedo]

There's still quite a bit to do to get it where I want, but so far so good. I'm sure some unexpected weird situaitional errors will pop up. I also want to add some more error checking, make a few things more flexible, and have the reporting expanded (sending text messages and emails).

Realistically, I just wanted to see if I could do it, and then if I could, use the pieces where needed now and in the future--more efficiency means more time to find opportunities (or drink :lol:).

[Space Monkey]

So does your program make the bets for you based on probability of winning and value in the odds? Similar to another program we've discussed before? Except of course, that yours is automated.

[SchagFactorToWin]

1) Automatically download the days tracks data files early in the morning...

Is there a service that provides this?

[CBedo]

So does your program make the bets for you based on probability of winning and value in the odds? Similar to another program we've discussed before? Except of course, that yours is automated.This uses some of the things we have talked about, but comes at it from a slightly different angle. It basically tries to find the "best" horse by looking at a number of factors, and then decides what it would be willing to pay for it (mid odds). The definition of "best" is not exactly the same as what you would think of when you think best overall.

Is there a service that provides this?Not that I know of, but it's not really that hard to quickly download data from drf, equibase, bris or tsn. If you google "downloader," you will get a long list of products that will help automate downloading. There are also downloader plug-ins for Firefox that work well. You just set the criteria or filters for the links you want to download and it does the rest. Just be careful, as some sites say auto downloading is against their terms of service, as they don't want there servers getting overloaded. PM me if you need further guidance.

I'll give a "how was my day" progress report for day 1 a bit later.

[CBedo]

Trip Report, Day 1 (it is a wild ride!).

Error 1: I started off the day by waking up and seeing a nice error message on my machine....I had set the program to download my data twice (just in case the data provider's site was down for maintenance or something the first tme I checked). The problem was that I hadn't instructed my unzipping library to overwrite any existing files, so it didn't know what to do.

Error 2: Luckily, that was well before any tracks had started. The second error of the day happened uploading the first set of bets to PTC. I had a track code that I thought was valid, but it turns out that PTC didn't take the track (PHA). This wasn't a disaster as I had planned for bad codes (have you seen their crazy code list?), but I spent way too much time debugging what turned out to be not the kind of bug I was looking for.

Error 3: After that, things went pretty smoothly until my machine decided to lock up early in the afternoon not due to my program, but having the memory overloaded (my development machine is a bit older, and not exactly memory efficient. I had forgotten the cardinal rule of working in a windows environment...Reboot early and often as sloppy programming, memory leaks, and poor garbage collections will kill a machine in a hurry (and I hadn't rebooted in a few days, hehe). Luckily, the only track I missed processing didn't have any opportunities for me today anyway.

So three issues only on the first full day of live testing. I consider this way better than expectations, haha. It actually scares me a bit, as I wonder what bugs are lurking out there!

Results: As far as results, for the day, I only had 16 possible bets, and 7 of them got passed for value reasons. Of the other 9, I had an easy 3/2 winner that I would have bet at even money or better (he was 9/5 when the bet was processed). and then lost a couple before catching a 7.60, 6.60, and 10.00 dollar winner (none of which, for the record, were bet down after I put my bet in. In fact, 2 of the 3 actually moved up a point from when I bet them (including one early easy front running leader). Of course, after getting on a hot streak, and thinking what a great first day would be if I could catch one or two of the my last four bets, one went off at 1/2 and was passed (he lost), and the last three all ran terribly, one getting burned up in a speed duel, one being in the garden spot and not firing, and one got bumped a bit at the start and never ran a step. but no complaints about a very nice first day!

Tomorrow will be the first real test of "letting go of the steering wheel" as I'm driving to Las Vegas and will let the model do it's thing (it's a Tuesday, so there's not that many possibilities).

I'll check in with it, and with you guys (and girls) tomorrow and hopefully won't have any disasters to whine about!

Good luck to all.

[Tom-Oh]

This sounds awesome man, it's great to see someone experimenting with something like this. Keep us posted, I'd love to here the progress your making!

[CBedo]

Quck report about day 2. While driving to Vegas, I had 11 bets upload yesterday. Of the 11, only five met their respective odds filters. Of those only 1 won (4 2nds & 3rds, grrrr) paying 6.60. A losing day unfortunately, but the program ran smoothly and continues to run today. I'm excited for the weekend to get here to see if can handle betting many more tracks at one time (and I won't even be able to check in since I'll be on a houseboat).

Good luck everyone.

[Hosshead]

This is exciting CBedo ! It's like putting the Starship Enterprize on automatic pilot.
Good luck Capt. :ThmbUp:

[kroebuck67]

That's really impressive, man. Imagine where you might be in a few years, if your handicapping is solid enough. You'll be getting updates on your Iphone on how much money you made today, while you globe-trot (or whatever floats your boat).

Definitely keep us updated on your progress! :ThmbUp: :ThmbUp: :ThmbUp:

[CBedo]

Just back from four days on a Lake Powell houseboat, and I haven't scrutinized the data and logs yet, but thought I'd give a quick update on the system and the results.

First off, the overall system continued to run smoothly until sometime late Saturday afternoon. When designing these things, we are always looking for bugs in the system, but what I didn't plan for were bugs "behind" the system. My girlfriend's mother-bleepin cat (who I inherited when she moved in--I did not purchase) decided to chase something behind my desk (spider probably), and somehow she managed to knock the power cord out of the wall (I do have a backup power unit, but it only lasts about an hour or so). So, I guess after doing some debugging, now I have to cat proof my office.....:bang:

As far as results, I think the last day I had reported was Tuesday, the day I drove to Vegas. Since then I've had 116 possible plays (that counts the ones that are logged--so up until the system crash. I haven't looked at the "coulda been" bets that didn't get uploaded yet). Of the 116, 51 did not qualify as having value. Of the other 65, 19 won (just under 30%) which returned 165.20 (+27.1%). I haven't looked at the results super-closely yet, but I did see a 12/1 winner that probably skewed the results to the positive side.

More info tomorrow for anyone who cares. Good night and good luck.

[JustRalph]

good stuff...........

[Triopstor]

Yes! Very Exciting! Having a complete auto-system.


2) Handicap a set of races.

Generally how do you handicap the horses?

[CBedo]

Generally how do you handicap the horses?In terms of the model that I'm using for this test, if you search the posts in the selections forum, You'll find a week of picks that I made using it and a description of it. Basically, this model looks for the "best" horse in races it thinks it can handicap effectively, and then determines what value it needs to make the bet. The definition of best doesn't necessarily equate to the traditional meaning. It more ore less looks for horses that have the best overall ability, are in OK form, and have enough pace ability to hopefully not be overly affected by bad early scenarios (too fast or too slow).

But I would also add, that this total automation is more of a test of many individual pieces that I hopefully can use in further endeavors. It would be very easy to "drop in" another handicapping model and have it do the rest.

[CBedo]

Just a quick note on what's going on since I've gotten back from vacation. Everything seems to be running smoothly since I've gotten it plugged back in, haha.

Interestingly, the days that did not get bet because of the previously mentioned issues would have been a losing set of wagers (not every day would have been losing, but overall), including one day where the model had 17 potential plays and all 17 lost (14 2nds & 3rds). Only 9 of those would have been bets--a not so nice losing streak! Including all those in the overall numbers since inception of the full automation would bring the roi down dramatically, but would still be solidly positive. Yesterday was also a very good day: 16 potential bets, 7 qualified, 4 won. Payoffs were 4.60, 6.00, 6.00, & 10.60.

Good luck.

[DeanT]

How many plays have you signalled since you started and what is the strike rate Chris?

[CBedo]

How many plays have you signalled since you started and what is the strike rate Chris?I don't have the numbers right in front of me (other machine), but I think counting the bets that did not go in, there have been about 222 bets so far in September. I recently started looking at what percentage of races I was getting a bet in and it seems to usually be in the 9-15% range each day. In general, it has been giving me about 700ish possible plays per month before value considerations. Odds considerations seems to be eliminating roughly 250 of those. The overall win rate has been in the low to mid 30s looking at all possible wagers. After value filters, it seems to drop about 4-5 points.

Sorry I don't have more specific numbers, I can post them later if you want them.

[CBedo]

Strange day today. Very few bets. I thought something was wrong with the model since it hadn't loaded any bets when I checked it this morning, but it turns out, it just didn't find anything it liked. Overall, for the day there were only 7 possible bets (by far the fewest I have seen since I started developing this back in June). Of the 7, only 3 qualified for actual bets (only 1 of the other 4 won, paying 3.20). Unfortunately, all three lost.

We'll see what happens tomorrow as there are 25 possible bets (before scratches obviously).

[CBedo]

What a day. I ended up with 30 possible bets. Of the 30, 11 failed on value considerations. Of the 19 bets that got executed, I won only 3! 3 out of 19 is a stellar 15.8% hit rate. The good news was that the three winners paid 7.50, 13.20 & 21.80 (also had a 14/1 shot just miss)! So based on $2 wagers, 38 wagered, returned 42.50 for a return of 11.8%.

Who needs a high hit rate! :lol:

[CBedo]

Besides for being a positive day (somehow), this was the perfect day illustrating to me what is still one of the biggest value of computer based handicapping. Looking at the potential bets the model was using today helped me to quickly zero in on some potentially great horses to use in exotics. With very little time spent looking for races, I zeroed in on seven that I thought might hold some great value. Of those, one turned out not to have any value (but since I didn't have much time invested, and I had plenty of other opportunities, it was easy to pass). Of the other six, only one of my horses won, paying 21.80, keying some nice exotics, but two others ran 2nd also completing some nice payoffs for me.

I sill aspire to fully automated handicapping (without worrying about it, lol), but there's no way, I could have found all of today's opportunities without the computer.

[Space Monkey]

Just want to let you know that I'm following your endeavor with great interest Chris. I don't have much to say. Just reading and hoping for success!!!

[CBedo]

Just want to let you know that I'm following your endeavor with great interest Chris. I don't have much to say. Just reading and hoping for success!!!Thanks. Everything is going pretty smoothly right now (except my horses, hehe). I'm monitoring the sytem, but it has been running smoothly without any bugs showing up lately.

As far as results go, going through the normal ups and downs of betting horses. One day I'm the smartest guy in the world, the next day I'm an idiot.

Luckily today, I'm smart, haha. 13 possible bets, 8 had enough value to bet, 4 winners--3 low priced and one at 6/1.

[CBedo]

After today, the automated model has put in over 100 bets, 106 to be exact. This doesn't count the bets that didn't execute due to having no value, and it also doesn't count any bets the model didn't load due to program errors, bad track codes, or my machine crashing (some of these bets I caught and was able to put in manually). So this only counts fully automated, no human intervention wagers.

The first 106 were interesting to say the least. The hit rate was a bit lower than I had expected, but the odds were much better. Of 106 bets, 29 were winners which is 27.3%. In testing, the hit rate had been in the 29-32%, so it was a bit lower. There were 11 days that the model put in bets (one day wasn't quite complete, but almost, and a few days were dark when my machine was down while I was gone). Of the 11 days, 6 days were winners and 5 days were losers, which is, honestly, about what I expected.

There were some not so nice losing streaks including a 0 for 7 day, and a a day when after hitting 3 of the last four bets, it then lost 11 of the last 12, but it also has some fun small winning streaks as well such as today when after starting one for two, it lost five in a row, but then hit four in a row, that included a 9/1 and a 5/1 horse within minutes of each other at For Erie & Thistle.

Offsetting the lower than expected win percentage was a much higher than expected average odds which came in at just over 7/2. This resulted in a slightly higher than 20% return so far on those 106 bets.

This is clearly above expectation, but I'll take it while I can get it! We'll see what happens in the next couple weeks.

[gm10]

After today, the automated model has put in over 100 bets, 106 to be exact. This doesn't count the bets that didn't execute due to having no value, and it also doesn't count any bets the model didn't load due to program errors, bad track codes, or my machine crashing (some of these bets I caught and was able to put in manually). So this only counts fully automated, no human intervention wagers.

The first 106 were interesting to say the least. The hit rate was a bit lower than I had expected, but the odds were much better. Of 106 bets, 29 were winners which is 27.3%. In testing, the hit rate had been in the 29-32%, so it was a bit lower. There were 11 days that the model put in bets (one day wasn't quite complete, but almost, and a few days were dark when my machine was down while I was gone). Of the 11 days, 6 days were winners and 5 days were losers, which is, honestly, about what I expected.

There were some not so nice losing streaks including a 0 for 7 day, and a a day when after hitting 3 of the last four bets, it then lost 11 of the last 12, but it also has some fun small winning streaks as well such as today when after starting one for two, it lost five in a row, but then hit four in a row, that included a 9/1 and a 5/1 horse within minutes of each other at For Erie & Thistle.

Offsetting the lower than expected win percentage was a much higher than expected average odds which came in at just over 7/2. This resulted in a slightly higher than 20% return so far on those 106 bets.

This is clearly above expectation, but I'll take it while I can get it! We'll see what happens in the next couple weeks.

Interesting stuff ... a few questions if I may

a) how do you place your bets without human intervention (Betfair has a very good API but I'm not familiar with American versions).
b) how do you determine whether something is value when you don't really know what the final odds will be? DO you only use odds close to post time and then build in a threshold?

[CBedo]

As a walking/talking adult I'd avoid hehe & haha when typing.I use hehe, haha, lol too much I know, hehe, since I don't like to take myself too seriously, haha.

LMFAO :lol:

[CBedo]

Interesting stuff ... a few questions if I may

a) how do you place your bets without human intervention (Betfair has a very good API but I'm not familiar with American versions).
b) how do you determine whether something is value when you don't really know what the final odds will be? DO you only use odds close to post time and then build in a threshold?a) There are two ways that I get bets uploaded without having to do anything. The first way is to use file upload features at some ADWs (like PTC). After the model decides on a bet, it creates a text file which it then uploads to the ADW. The second way is for ADWs that don't have file upload features. It uses some windows automation to basically go through the same steps you or I would to put in a bet, except it does it all in the background.

b) Getting the correct value is currently in phase I, and hopfeully soon will be moving to phase II. In the first phase, I'm using the odds at 0 minutes to post to judge the value of a bet. I have some data that has given me some idea of how much higher I need to set the miniumm to get what I want (only on some of the lower priced horses), and I either build that into a conditional bet or have the program watch the odds if the ADW doesn't have conditional betting. So, I guess you could say, I'm making an educated guess. Phase II will have a little less guesswork. I'm collecting data on odds changes after zero minutes to post, and hopefully this will allow me to do two things. 1) Find out how long after 0 minutes to post I can put in bets at specific tracks. Some seem to run quickly, and some seem to take 3 minutes on average after. Waiting til the last minute will help some, but we all know that there will still be odds changes. 2) I'm doing some work to project final odds changes based on odds movement, track, and even a few other fundamental pieces of information. If I can do that, then I can change my odds thresholds to where they need to be. Also, more simplistically, since I'm collecting odds at the time of bet, I can backtest my model on the "when bet" odds instead of the "going off" odds to see if I need to change the gap I'm using.

Hope this answers your questions.

[CapperLou]

Chris:

What you have achieved here is truly remarkable. Kudos to you. I'm still in the 20th Century the way I'm placing wagers although Conditional play is somewhat new.

[CBedo]

Chris:

What you have achieved here is truly remarkable. Kudos to you. I'm still in the 20th Century the way I'm placing wagers although Conditional play is somewhat new.Thanks. Of course as is usual with any large project, the more I do, the more things I see that need to be done (or I would like to do). The reality of this was that it was more a proof of principle project than anything. Now I have many tools to use whenever I need a specific piece from the handicapping process from end to end.

The fact that it currently is running profitable is bonus! :jump:

[gm10]

a) There are two ways that I get bets uploaded without having to do anything. The first way is to use file upload features at some ADWs (like PTC). After the model decides on a bet, it creates a text file which it then uploads to the ADW. The second way is for ADWs that don't have file upload features. It uses some windows automation to basically go through the same steps you or I would to put in a bet, except it does it all in the background.

b) Getting the correct value is currently in phase I, and hopfeully soon will be moving to phase II. In the first phase, I'm using the odds at 0 minutes to post to judge the value of a bet. I have some data that has given me some idea of how much higher I need to set the miniumm to get what I want (only on some of the lower priced horses), and I either build that into a conditional bet or have the program watch the odds if the ADW doesn't have conditional betting. So, I guess you could say, I'm making an educated guess. Phase II will have a little less guesswork. I'm collecting data on odds changes after zero minutes to post, and hopefully this will allow me to do two things. 1) Find out how long after 0 minutes to post I can put in bets at specific tracks. Some seem to run quickly, and some seem to take 3 minutes on average after. Waiting til the last minute will help some, but we all know that there will still be odds changes. 2) I'm doing some work to project final odds changes based on odds movement, track, and even a few other fundamental pieces of information. If I can do that, then I can change my odds thresholds to where they need to be. Also, more simplistically, since I'm collecting odds at the time of bet, I can backtest my model on the "when bet" odds instead of the "going off" odds to see if I need to change the gap I'm using.

Hope this answers your questions.

Hi ... very interesting stuff. It certainly sounds like you are doing things properly and logically. Just a little idea for you ... with Betfair you get fixed odds, so you can eliminate the guessing for whether the final odds will be 'value'. I also opens up arbitraging possibilities. (Send me a PM if you want to discuss this further).

Thank you for your reply, and good luck.

[CBedo]

Hi ... very interesting stuff. It certainly sounds like you are doing things properly and logically. Just a little idea for you ... with Betfair you get fixed odds, so you can eliminate the guessing for whether the final odds will be 'value'. I also opens up arbitraging possibilities. (Send me a PM if you want to discuss this further).

Thank you for your reply, and good luck.I can only dream of using Betfair, since I'm in the U.S.......

[CBedo]

A fun (and profitable) week for the automated model. From Monday to Sunday, the model made 82 bets and returned 37% on those bets which is obviously significantly better than anything I could have hoped for.

There were 4 profitable days and 3 negative ones, but the negative days were pretty tame, with the worst being down about 4 bets for the day. There were two spectacular days that really drove the results, and both were very different and interesting. The first was Tuesday, where the model made 15 bets, many more than usual on a Tuesday. Usually, about 2/3 of possible bets actually get bet due to value, but on Tuesday, 15 of the 16 total bets executed, and a few nice priced horses carried the day. The second day that drove the results was Saturday. I had 38 possible bets for the day, 20 of them ended up executing, and it got off to a fast start, hitting some nice prices early (and the longest win streak so far--4), which was good, since it then proceeded to go on a 9 race losing streak (interestingly, I think the 9 race losing streak is WAY more likely than the 4 race winning streak). Even after enduring that losing streak, it still ended up 9 bets for the day, so I couldn't complain...well not as much as I would have anyway!

I'm not going to change anything until the end of the month (if even then), but if the model continues to produce higher than expected returns, then I'm thinking that I'll probably have to change my value proposition to reduce the required overlay to make a bet. A high return% is great, but it's about total dollars returned, not percentages, and it would seem that I might be leaving some on the table by being too selective. Any thoughts from you guys on this subject?

Good luck in the upcoming week to everyone.

[Niko]

:ThmbUp: :ThmbUp: :ThmbUp: :ThmbUp:

[markgoldie]

Isn't it possible to run an ROI on the selected-but-not-bet horses to see what you might or might not be leaving on the table? If you can do so, then I'd let those results tell me whether or how much to loosen.

[CBedo]

Isn't it possible to run an ROI on the selected-but-not-bet horses to see what you might or might not be leaving on the table? If you can do so, then I'd let those results tell me whether or how much to loosen.Yes, and the overall nonbet horses are negative, so I have to figure out which (if any, I'm assuming there are) subsection of the nonbets to use. The issue for now revolves around the late odds changes, and not yet having the right data (working on it) to make an accurate assessment.

[CBedo]

Just a quick update as the model went over its 200th bet yesterday. After 223 bets. the model remains postive with a total return on invested capital of just over 20%. I still think this seems too high to me, but I'm not complaining! It also achieved it's first ever 5 day winning streak (why do I have a perverse feeling of impending doom? :bang:). None of the days in the streak were big up days, in fact, one day it made 20 cents, but it continues to grind away.

Quick update for today: After ending yesterday with 5 winners of the last 6 races, it started today 0/7, but then proceeded to catch a 7/1 horse at NP and a couple others to put it back in the black for the day...so far.

[DeadHeat]

Hey CB,

Keep up the good work. I am following this thread (and PA's) with great interest.

I am curious what language you are using to program with. Especially how it can be done "behind the scene". I can see how an Auto Mouse program might do it with another programming language but that would have to be done "on top".

Thanks,

DH

[CBedo]

Hey CB,

Keep up the good work. I am following this thread (and PA's) with great interest.

I am curious what language you are using to program with. Especially how it can be done "behind the scene". I can see how an Auto Mouse program might do it with another programming language but that would have to be done "on top".

Thanks,

DHThe "upfront" stuff is pretty forward with windows automation, whether you are using macro program like iMacros in Firefox or using a scripting language like AutoHotKey (highly recommend for some simple automation tasks and if anyone cares, I can point you to a nice tutorial).

I'm doing almost everything "behind the scenes" as you say, so I never have to lose control of what's going on on my screen. For this I personally think there are really two options, php or Ruby. I went with Ruby primarily for two reasons (I'm sure php guys have plenty as well). The first reason is that Ruby (and it's Rails web application framework) have a really slick and easy to use library for working with databases that uses ORM (object relational mapping) to make working with your database (no matter what db it is) pretty easy. The second reason that I decided to learn Ruby was that one of my good friends is a Ruby on Rails programmer, so I had someone that I could lean on when I couldn't find the answers anywhere else. Even if when I bothered him with questions, his answer was "google it," it was still nice to have him around.

Hope this answers your questions. Feel free to ask me anything. I won't have the answers to everything, as I'm still learning as well, but I'll try to answer as best (and as honestly) as I can.

Maybe PA can chime in with what he language or framework he is using?

[DeanT]

Good job Chris! I love it when a plan comes together.

Hold it, who said that line? I have a mental block.

Anyway, hope it keeps rolling.

[CBedo]

Good job Chris! I love it when a plan comes together.

Hold it, who said that line? I have a mental block.

Anyway, hope it keeps rolling.Hannibal (George Peppard) from the A-Team. :lol:

Thanks!

[PaceAdvantage]

I'm using Visual Basic...old school...not .net...actual visual basic....lol

[CBedo]

I'm using Visual Basic...old school...not .net...actual visual basic....lolGo with what works! I guess I should have mentioned .net as a possibility for doing this stuff, but i know so little about it, I just block it out of my mind.

[CBedo]

Quick update for today: After ending yesterday with 5 winners of the last 6 races, it started today 0/7, but then proceeded to catch a 7/1 horse at NP and a couple others to put it back in the black for the day...so far.Should have kept my mouth shut! Today was a terrible day that somehow was still somewhat encouraging. How is that possible? 20 bets today, and the model only won 3 of them (YUCK). It also started the day with 7 straight losses and then ended the day with an 8 race losing streak (double YUCK). Sounds pretty disheartening until you look at the bottom line and see that it somehow only lost 4 bets for the day.

Live to fight another day.

[Dick Schmidt]

CBedo,


I'm following your progress with great interest. For the past four months I've been involved in creating an automatic program to trade Foreign Exchange. It's tough to find an automatic procedure that covers all bases and is flexible enough to survive. I have a winning procedure that I thought would be easy to automate, but I find that I provide more insight and fuzzy logic into the trades than I thought. Still hoping to pull it off, as the FOREX market runs for 120 straight hours each week.

Anyway, I'm encouraged that you are doing so well. Horse racing is easily as difficult as investing, maybe harder.

Dick

Beware of the man who won't be bothered with details.

- William Feather Sr.

[Handiman]

Dick,

I took a spin at forex using several experts as they are called. And by experts I am not talking about forex touts but automated trading systems. One thing that I found to be interesting was that most I used made nice profit with test accounts or play money accounts or whatever you want to call them.

But once switched over to live trading they did not perform as well or even nearly the same. And added to that the fact that the spreads were liquid based on trading times depending on what market was active made it even tougher to get an auto trading system to be profitable.

I'd be interested in your results as they come to fruition.

Handi

[CBedo]

One thing that I found to be interesting was that most I used made nice profit with test accounts or play money accounts or whatever you want to call them.

But once switched over to live trading they did not perform as well or even nearly the same. And added to that the fact that the spreads were liquid based on trading times depending on what market was active made it even tougher to get an auto trading system to be profitable.I was afraid of the same thing when going from paper to real wagers with the ponies. So far anyway, the preparation for the differences between paper and real trading have made a difference, and to date, the testing results have held up in the real world.

[CBedo]

I'm following your progress with great interest. For the past four months I've been involved in creating an automatic program to trade Foreign Exchange. It's tough to find an automatic procedure that covers all bases and is flexible enough to survive. I have a winning procedure that I thought would be easy to automate, but I find that I provide more insight and fuzzy logic into the trades than I thought. Still hoping to pull it off, as the FOREX market runs for 120 straight hours each week.

Anyway, I'm encouraged that you are doing so well. Horse racing is easily as difficult as investing, maybe harder. Thanks Dick. That means a lot coming from you. The never ending layers of complexity that you can add when wagering (or trading) can bury a model for sure which is why I have try to start from the simple and work towards the more complex, in planned, structured steps, working in multiple branches as I go. I still have a huge laundry list of things I want to "improve." The biggest surprise, so far, has been that I didn't have to dig down into as many "complexity" nodes as I thought I would before I came up with something that worked (or at least seems to be working currently).

[DeadHeat]

...AutoHotKey (highly recommend for some simple automation tasks and if anyone cares, I can point you to a nice tutorial).

Found the website but if there is another tutorial somewhere I for one would like to see it.

Thanks

[CBedo]

Found the website but if there is another tutorial somewhere I for one would like to see it.

ThanksThere's not really any great documentation that makes it easy, but here is a tutorial that someone put together that at least will give you some decent direction.

http://www.autohotkey.com/forum/topic47791.html

[formula_2002]

cb, using excel 2000, and brisbet. In trying to import the wps pools all i seem to get is the 1st or second page of the web site.

I click the yellow arrow on the page I want, but do get that page..

any help here?

thanks
joe m

[CBedo]

cb, using excel 2000, and brisbet. In trying to import the wps pools all i seem to get is the 1st or second page of the web site.

I click the yellow arrow on the page I want, but do get that page..

any help here?

thanks
joe mI'll see if I can find my old copies of Excel to check it out. I'll take this discussion back to the Excel thread when I get a chance to try to duplicate what you did (probably tomorrow).

My first recommendation is to get Excel 2007. It's not that expensive, and you can download a 60 day trial for free.

[CBedo]

I've gotten a dose of humility (if not reality) this week as I had a seven day losing streak (which finally got broken today). A few of those days, it lost less than 1 bet for the day, but however you look at it, it has been ugly.

With the ending of a number of summer meets, the number of bets per day dropped this week quite a bit (especially early in the week). So it took 8 days to get from the 200 to the 300 bet mark. At the end of today, the model has now made 317 bets, and is cumulatively up 10.6% over those bets.

If you back out the returns since I last reported, there have been 94 bets which lost 11.6%. I think we have all gone through periods like this where you seem to lose every bob, have horses throw jockeys, get dq'd, and just not run a step. The win rate and the average mutuel suffered.

As a poker player, sometimes it's easier than in horse racing to see where you were "unlucky," and I've had a few thoughts on that, and how to possibly measure it to some extent in horse racing. Look for a thread in the handicapping forum sometime soon with more on this topic., but I will say that it does seem that this week I've been "less lucky" than in past weeks--which week is the baseline (if any) is still the question.

Lastly, with regards to the mechanics of the model itself, it continues to work as planned although Equibase changed its scratch page this week, so that caused some consternation in the beginning of the week. I love the changes, and think they are good for all of us (look for an article about this from me soon ;) ). I also am happy that Equibase was smart enough to deploy it on one of the slowest days of racing in the last few months. Some non-thinking types would have tried to do this on a busy day, and had us (and themselves probably) cursing.

Good luck to everyone, and here's to hoping for a better week!

[Warren Henry]

I've gotten a dose of humility (if not reality) this week as I had a seven day losing streak (which finally got broken today). A few of those days, it lost less than 1 bet for the day, but however you look at it, it has been ugly.

With the ending of a number of summer meets, the number of bets per day dropped this week quite a bit (especially early in the week). So it took 8 days to get from the 200 to the 300 bet mark. At the end of today, the model has now made 317 bets, and is cumulatively up 10.6% over those bets.

If you back out the returns since I last reported, there have been 94 bets which lost 11.6%. I think we have all gone through periods like this where you seem to lose every bob, have horses throw jockeys, get dq'd, and just not run a step. The win rate and the average mutuel suffered.

As a poker player, sometimes it's easier than in horse racing to see where you were "unlucky," and I've had a few thoughts on that, and how to possibly measure it to some extent in horse racing. Look for a thread in the handicapping forum sometime soon with more on this topic., but I will say that it does seem that this week I've been "less lucky" than in past weeks--which week is the baseline (if any) is still the question.

Lastly, with regards to the mechanics of the model itself, it continues to work as planned although Equibase changed its scratch page this week, so that caused some consternation in the beginning of the week. I love the changes, and think they are good for all of us (look for an article about this from me soon ;) ). I also am happy that Equibase was smart enough to deploy it on one of the slowest days of racing in the last few months. Some non-thinking types would have tried to do this on a busy day, and had us (and themselves probably) cursing.

Good luck to everyone, and here's to hoping for a better week!
Chris,

I think what you have accomplished is quite remarkable. Having a stable platform that can gather the data, make decisions on the bets, then place the bets all automatically is impressive.

Personally, I feel that your betting results are quite good too. Overall, you are in the black by a very good percentage over 300 bets. While the most recent period is no doubt a disappointment, the cumulative result is still quite good. Even this most recent period, you have beaten track take.

Hopefully, at the end of 600 bets, you will still be solidly in the black.

Keeping my fingers crossed for you.

[DeanT]

Excellent thread CB. Keep this going. As for losing streaks, well I am the King of them, so any advice just ask :D


Keeping my fingers crossed for you.

Some posters we know are telling the truth when they say that, and you are one of them Warren. Very nice!

[Hosshead]

CB the fact that you have been able to completely automate the betting process is just amazing. All the best to you.

Your concern about profit/loss, losing streaks etc. just brings home the reality that handicapping and betting are two seperate things for all of us.

I wonder if your + ROI to date is mainly due to a few big hits? Might help to find out.

Maybe after all the time you've put into automating the betting process, you will have time now to fine tune the handicapping "rules".

Keep it going, and us updated.:ThmbUp:

[CBedo]

CB the fact that you have been able to completely automate the betting process is just amazing. All the best to you.

Your concern about profit/loss, losing streaks etc. just brings home the reality that handicapping and betting are two seperate things for all of us.

I wonder if your + ROI to date is mainly due to a few big hits? Might help to find out.

Maybe after all the time you've put into automating the betting process, you will have time now to fine tune the handicapping "rules".

Keep it going, and us updated.:ThmbUp:I have definitely spent more time lately on process as opposed to handicapping, and the list of things that I'd like to look at is piling up. Hopefully, I'll get to some of it soon.

As far as odds group profitability. Although the current 300 bet trial by fire is not as smooth, if you add that to about 1000 bets before it (about 200 of which were real, and the rest were on paper), the model is pretty consistent. Above a minimum odds range (between 6/5-8/5 depending on a few other factors), all odds ranges are profitable. The return does become more profitable as measured by return on investment percentage as you move to higher odds ranges, but as you would expect, the number of plays drops off dramatically as you move out the same range, so the actual highest profitability in total dollars is in the lower end of the range.

Thanks for the encouragement and your interest.

[newtothegame]

Cbedo,
Would love to hear more about the results and how this has been going for you.

[CBedo]

Cbedo,
Would love to hear more about the results and how this has been going for you.I'll continue to report the results after each 100 additional bets, at least for the first 500 or so. I'm about 30 bets from the 400 mark, so more than likely I'll hit that mark on Saturday (possibly tomorrow, depending on scratches). Feel free to ask if you have any specific questions about the model or the process. I'll try to answer as best I can.

Today was an interesting day to say the least. After starting the day with a nice 8 race losing streak with 5 seconds (4 to a favorite or 2nd favorite that each paid from 20-36 for the exacta--future research), I got a head bob at Turfway Park and a 16.20 winner that gave me a little momentum that continued with a couple few priced winners before getting seconditis again to end the day. If I was actually watching each race, I'm sure I'd have had some big emotional swings today, but since I only looked at the summary, the early losing streak, the subsequent winners, and the late losing streak are just statistics in the database.

[newtothegame]

I'll continue to report the results after each 100 additional bets, at least for the first 500 or so. I'm about 30 bets from the 400 mark, so more than likely I'll hit that mark on Saturday (possibly tomorrow, depending on scratches). Feel free to ask if you have any specific questions about the model or the process. I'll try to answer as best I can.

Today was an interesting day to say the least. After starting the day with a nice 8 race losing streak with 5 seconds (4 to a favorite or 2nd favorite that each paid from 20-36 for the exacta--future research), I got a head bob at Turfway Park and a 16.20 winner that gave me a little momentum that continued with a couple few priced winners before getting seconditis again to end the day. If I was actually watching each race, I'm sure I'd have had some big emotional swings today, but since I only looked at the summary, the early losing streak, the subsequent winners, and the late losing streak are just statistics in the database.

Well I for one can say without hesitation, that I am impressed. Although I am unable to ask much about the programming (as honestly I wouldnt understand even what I was asking much less your response). Don't get me wrong, I would love to learn more about it but, at this point I am still learning my own shortcoming with the ponies. To quote my father some years ago, " I am like a kid in a candly store".
Although I would like to hope that as I gain more knowledge and understanding that I could in the future ask questions of ya. Till then....I wish you all the best with your endeavors.

[CBedo]

It took a day longer to get over the 400 bet mark than I thought it would. Yesterday was a strange day as the model only executed 14 bets, well below the normal Saturday. Before value considerations, there were 43 possible bets, but only 14 met odds requirements. This was an abnormally high number of failing bets.

The way the model is performing recently, I wish all the bets had failed to execute! 101 bets have been made since the last update which have had a negative return of -7.9%. I think every losing horse player (like poker players) have a book full of excuses as to why they are losing, and I'm no different. Looking at the last week's records, it seems that the model has a severe case of seconditis, seemingly losing every head bob. Favorites are killing the performance. When the model's horse is the favorite, it runs 2nd, and when the model's horse isn't the favorite, the favorites are winning (maybe I've found a way to judge the legitimacy of favorites? :bang: ).

Overall, the model has now made 418 bets at 32 tracks and is hanging on to a 6.1% overall return. Here's to beter tomorrows.

[BillW]

If -7.9% is the worst you do in a 101 bet stretch, it's money in the bank! Keep plugging!

[Warren Henry]

It took a day longer to get over the 400 bet mark than I thought it would. Yesterday was a strange day as the model only executed 14 bets, well below the normal Saturday. Before value considerations, there were 43 possible bets, but only 14 met odds requirements. This was an abnormally high number of failing bets.

The way the model is performing recently, I wish all the bets had failed to execute! 101 bets have been made since the last update which have had a negative return of -7.9%. I think every losing horse player (like poker players) have a book full of excuses as to why they are losing, and I'm no different. Looking at the last week's records, it seems that the model has a severe case of seconditis, seemingly losing every head bob. Favorites are killing the performance. When the model's horse is the favorite, it runs 2nd, and when the model's horse isn't the favorite, the favorites are winning (maybe I've found a way to judge the legitimacy of favorites? :bang: ).

Overall, the model has now made 418 bets at 32 tracks and is hanging on to a 6.1% overall return. Here's to beter tomorrows.
Chris,

Still looks good from my perspective.

It is probably still too early to look for meaningful patterns in your bet performances. However, I assume (always dangerous) that you have given thought to attempting to isolate subsets of your selections that produce higher or lower ROI success than the set as a whole?

To quote an old, old friend "I think you have found a bird's nest on the ground".

This is good stuff. Thanks for sharing.

[CBedo]

It is probably still too early to look for meaningful patterns in your bet performances. However, I assume (always dangerous) that you have given thought to attempting to isolate subsets of your selections that produce higher or lower ROI success than the set as a whole?I have a good number of ideas for subset analysis, but as you know, small sample sizes are always a challenge. So I've (for the moment) decided to leave things alone until I get at least 1000 or so bets under the belt.

[CBedo]

94 bets only this week with an extremely slow Sunday due to weather, cancellations at CT & BRD, and lots of underlays that didn't get bet. Of the 94, 28 managed to cross the line first (29.8%). Luckily, we managed to break our two week losing streak with a positive return of 9.7%.

Cumulatively, the model has made 512 bets and has a positive return of 6.8%.

Good luck this week to all!

[PaceAdvantage]

Congrats on a positive 500 race run!

[Handiman]

Very few can post that accomplishment. My hats off to you. Can't wait to see how the next 500 go.

Handi :ThmbUp:

[formula_2002]

94 bets only this week with an extremely slow Sunday due to weather, cancellations at CT & BRD, and lots of underlays that didn't get bet. Of the 94, 28 managed to cross the line first (29.8%). Luckily, we managed to break our two week losing streak with a positive return of 9.7%.

Cumulatively, the model has made 512 bets and has a positive return of 6.8%.

Good luck this week to all!
that should surpass the another two last fellows I recall posting 500 picks with a profit.
One was Steve W.... cant recall his last name the other is me.

Where are you posting your picks!!

continued Good Luck

[CBedo]

that should surpass the another two last fellows I recall posting 500 picks with a profit.
One was Steve W.... cant recall his last name the other is me.

Where are you posting your picks!!

continued Good LuckI haven't been posting them, so I can't take the award from you guys who have. I haven't posted all of them since one of my original test weeks.

[Bochall]

Very interesting stuff ya'll...I am a follower of Steve Crists' blog at DRF.com and one poster was commenting on the fact that some of his fatter tri and P3/P4 combos are paying less these days because, in his opinion, a computer is being used by someone who plays in his pools to snuff out this value. Is this tool also recommending multi race bets?... tris? I have read a few articles about players who make massive bets based on this type of info (they are always in Hong Kong or Singapore!!!)...feeling a bit powerless...makes one think twice before diving into some of these mega pool jackpot P6's.

[CBedo]

This specific model is only recommending single race win bets, but there's no question that I'm doing research on how to best use it in multi-race exotics. Currently, it doesn't pick every race. It picks the races it thinks it has a higher probability of winning and getting value.

[CBedo]

What an interesting week this was! Just looking at the stats doesn't tell the story. First the stats, then the story...

For the week, the model made 114 bets and won a higher than normal 32.5% of them. Unfortunately, most of the winners were low priced overlays and the return for the model for the week was a modest 6.7% gain.

What was bizarre was the number of near misses for the week. If you look at the horses whose closing odds were 4/1 or greater, there were 33 bets, and unfortunately, we only won 4 of them which in of itself was frustrating, but to rub a little salt in the wounds, 13 of the 33 ran second, including 4 that were between 10/1 and 20/1! :bang: Ahhh, the week that could have been....lol. It's amazing to me that you can test and test and test, but each and every week brings some new strange (and sometimes beautiful) dynamic in the results.

Overall now, the model has made 626 bets at 38 tracks and has a positive return of 6.8%.

Unless enough people really want me to keep posting this, I'm probably done for a while (maybe until it hits 1000 bets). From the technology standpoint, everything is running pretty smoothly, and now when I post, it just feels like I'm redboarding when it's a good week and whining when it's a bad one. I have lots of things I'm working on, and hopefully I'll have plenty of new material to talk about soon.

Good luck to everyone!

[the_fat_man]

Overall now, the model has made 626 bets at 38 tracks and has a positive return of 6.8%.



I suspect this is probably as good as it gets. Add a nice rebate and you're way ahead. Especially given that you're not physically handicapping and betting these races. There's much to be said for this.

[RichieP]

Overall now, the model has made 626 bets at 38 tracks and has a positive return of 6.8%.


Morning "C"
I am working with Ted Craven's RDSS software package on an auto paceline/contender template then manually entering conditionals into TS interface.

2 questions if I may:
1) How many days did it take you to reach the 626 bets? I read that they are single horse win bets so that is clear.

2) How many days were profitable/losers? Asking this strictly from a consistency and day to day "look" at things.

Thanks in advance for any help and you are doing great work man! Keep it going strong :ThmbUp:

[Warren Henry]

Unless enough people really want me to keep posting this, I'm probably done for a while (maybe until it hits 1000 bets). From the technology standpoint, everything is running pretty smoothly, and now when I post, it just feels like I'm redboarding when it's a good week and whining when it's a bad one. I have lots of things I'm working on, and hopefully I'll have plenty of new material to talk about soon.



I never tire of hearing about the success of a fellow player - especially if it involves something I am also attempting to do. Thanks for sharing.

[LottaKash]

Just curious Cbedo, have you been tracking the profitability of place & show slots.....It seems that with all of the near misses that you have reported, that perhaps, there may well be something interesting occuring, when looking in the place & show arena...

Would very much like to see a 5,000 race run.....I am enjoying this....:jump:

best,

[CBedo]

I suspect this is probably as good as it gets. Add a nice rebate and you're way ahead. Especially given that you're not physically handicapping and betting these races. There's much to be said for this.Just wondering why you suspect that is "as good as it gets." Is it just that you don't believe an automated process can hold up over time, or do you have some other insights?

Overall, given the simplicity of this model, I originally was hoping it would beat the takeout rate, coming in somewhere between down 0 to down 10%, but it was clear to me fairly early that it might be a bit better. I was still hoping -5 to +5, which is about where it has been before value considerations. After filtering for value, it has been somewhat better.

[CBedo]

Morning "C"
I am working with Ted Craven's RDSS software package on an auto paceline/contender template then manually entering conditionals into TS interface.

2 questions if I may:
1) How many days did it take you to reach the 626 bets? I read that they are single horse win bets so that is clear.

2) How many days were profitable/losers? Asking this strictly from a consistency and day to day "look" at things.

Thanks in advance for any help and you are doing great work man! Keep it going strong :ThmbUp:1) Overall, I have been getting about 100ish executed bets per week. It was more than this during the original testing and implementation, but that was due to more tracks running during the summer period. There are just fewer races for me to look at now, although it is offset a bit by less races on the turf (or taken off the turf) which my model will take a look at (it is a dirt model). The inital model usually comes up with a list of races to look at which is normally about 1/3 of races run daily. Then after further filtering and analysis, it tends to come up with about 12-15% of races daily as possible bets. Then after value considerations, this number drops to around 9-11% of races.

2) It has held pretty consistently overall at about winning 4 days out of 7, although I have had some winning and losing streaks. For the most part, this fits with what I saw in testing and have predicted that over a string of 20 bets, it should be postive just under 60% of the time. FYI, It's probably not exactly right to look at daily results as the number of bets on Monday & Tuesday are usually lots less than Friday & Saturday.

[CBedo]

Just curious Cbedo, have you been tracking the profitability of place & show slots.....It seems that with all of the near misses that you have reported, that perhaps, there may well be something interesting occuring, when looking in the place & show arena...

Would very much like to see a 5,000 race run.....I am enjoying this....:jump:

best,I have done a little research so far on place & show wagering as well as exacta possibilities There does seem to be something promising in there, but it is stratified by odds range. Overall, place betting would not have been profitable to date, and show betting would have been horribly bad, but there does appear to be some scenarios where each might make sense.

[the_fat_man]

Just wondering why you suspect that is "as good as it gets." Is it just that you don't believe an automated process can hold up over time, or do you have some other insights?

I'm not disparaging. I'm impressed. If you were to continue this rate of return, I'm sure there'd be lines of handicappers looking to buy the program from you.

I don't know whether your method will hold up over time as I don't know what factors you've modeled. Certainly 500+ plays is a good indication that it will, though, I suspect there'll be peaks and valleys, and it's a question of how low you go before coming back up and what the growth rate is.

I think (significantly) higher ROI's are attainable but I don't believe these methods of play can be automated. Nor would they offer as many plays a week. (Probably half or so.)

That's what I meant by 'as good as it gets'. Anyone would take the lower ROI if it meant doing significantly less work and putting much more money though the windows.

[RichieP]

1) Overall, I have been getting about 100ish executed bets per week. It was more than this during the original testing and implementation, but that was due to more tracks running during the summer period. There are just fewer races for me to look at now, although it is offset a bit by less races on the turf (or taken off the turf) which my model will take a look at (it is a dirt model). The inital model usually comes up with a list of races to look at which is normally about 1/3 of races run daily. Then after further filtering and analysis, it tends to come up with about 12-15% of races daily as possible bets. Then after value considerations, this number drops to around 9-11% of races.

2) It has held pretty consistently overall at about winning 4 days out of 7, although I have had some winning and losing streaks. For the most part, this fits with what I saw in testing and have predicted that over a string of 20 bets, it should be postive just under 60% of the time. FYI, It's probably not exactly right to look at daily results as the number of bets on Monday & Tuesday are usually lots less than Friday & Saturday.

:ThmbUp::ThmbUp:
Thank you so much for the feedback "C". It is greatly appreciated.
http://www.youtube.com/watch?v=FMl124VvFRg&feature=PlayList&p=80DF3FC49C19DCF0&index=2

[Niko]

Please keep us updated at 1000 bets, then 2000 etc are whatever. Feel free to share the highs and lows. Fun to follow

[Warren Henry]

Still running? Still doing as well as at last report?

[teddy]

Wondering how the model held up for you. Have you had any strange occurrances you didnt plan on?

[CBedo]

Wondering how the model held up for you. Have you had any strange occurrances you didnt plan on?As you would expect, there are always some hiccups along the way. The biggest unexpected occurrences didn't really have to do with the racing itself, but instead had to do with the underlying cost of ownership fo the technology.

The first was that instead of putting it up on an Amazon micro instance of EC2, I originally was hosting it myself on an old workstation. The problem with that arose when my giflriend decided she wanted to put her iTunes on that machine and crashed it while I was out of town for a week. :bang:

The second issue (not totally unexpected, but a pain none the less) was that when you are reliant on outside web services (BRIS/TSN for data, ADWs for odds data & betting), you are relying on them not changing their sites so your program can use them. You are also reliant on their quality of service and uptime. None of these have true web service APIs so you are kind of hacking around to get it to work. And of course, as you should expect, minor changes in their sites or access required some reprogramming from my side.

I think most people underestimate cost of ownership issues related to these types of things, and although they aren't super popular currently in horse racing software, this is why web services in other areas are exploding.

The last issue was model related and had to do with my train & test sets of data. When I set it up, I didn't have a fully developed database set as I have now, and even though I set up multiple test sets, they didn't cover the full set of tracks or types of races (especially time of year). What I found was that the model did not like winter racing for the most part, lol. As I started to see aberrations, I shut it down quickly until I could figure out why. I thought it was the normal "things look good until you put real money on it" phenomena, but as I continued to work on it, I found that the model again performed well the next year. Interestingly, as I was able to look at more back data as well, the overall model showed definite seasonal patterns that I never would have forecast. This has led me to incorporate it in a broader handicapping methodology, using it where I am more comfortable it works, and always looking to do something similar but "better." :)

[Kelso]

What I found was that the model did not like winter racing for the most part, lol.

CB,
Is the problem with cold-weather (AQU, PHA, BEU, etc.) or warm-weather (TAM, FGX, TUP, etc.) tracks in winter ... or both?.

[CBedo]

CB,
Is the problem with cold-weather (AQU, PHA, BEU, etc.) or warm-weather (TAM, FGX, TUP, etc.) tracks in winter ... or both?.In general, all were worse, but the warm weather tracks in winter were brutally bad for me.

[curious]

Hello,
I don't claim to be any kind of expert at this.

What I found for place and show that if I bet X on the win, 2X on the place, and 4X on the show I do better.

I don't know if you have any way to take what happened to date and see what this would have yielded. But, it may help.

People yell at me for betting low odds favorites this way, but if they don't win I don't lose very much, they usually come in 2nd or 3rd.

It might be true that only doing the X, 2X, 4X thing on horses > some higher odds value might be better, I don't know. It seems that you have the data to see what the result would have been if you had tried this.

[misscashalot]

Hello,
I don't claim to be any kind of expert at this.

What I found for place and show that if I bet X on the win, 2X on the place, and 4X on the show I do better.

I don't know if you have any way to take what happened to date and see what this would have yielded. But, it may help.

People yell at me for betting low odds favorites this way, but if they don't win I don't lose very much, they usually come in 2nd or 3rd.

It might be true that only doing the X, 2X, 4X thing on horses > some higher odds value might be better, I don't know. It seems that you have the data to see what the result would have been if you had tried this.

What is the roi of each of the WPS positions?

[CBedo]

Interestingly, the horses that provided the majority of the value for the model in the win slot, provided big losses in the place and show spots. I compared the models win, place & show percentages with the overall win, place & show percentages for similar odds ranges & win rates of races in general, and apparently, the way my model worked was that it tried to find horses that were ready and on the improve to get some value, and when it was wrong, it was usually very wrong, lol, meaning that I saw much lower place & show percentages that what you would have predicted if just given a hit rate or a certain going off odds.

On the flip side, the lowest price horses place and show roi was close to (but not quite) break even. These horses provided mostly no value in the win slot, but did present some potential possibly underneath.