Shelbyrr11's MLB Gambling Thread

[shelbyrr11]

I'm cutting to the chase:

For the last two years, I have been fine-tuning and perfecting an MLB prediction model that predicts the winner of games and by which margin. While it was not a "live" run, I did simulate most all of the games in 2018 with this prediction model. I was able to pick the correct winner in 57.4% of games, having only excluded doubleheaders and the few games in which the expected margin of victory for either team was 0.00 runs (a very rare event in itself). 

Now, that 57.4% accuracy rate has several layers. There are quite a few games that I can predict with about a 65% accuracy rate, depending on the factors going into the game. Conversely, I have a bundle of games with about a 52-54% accuracy rate, primarily if I do not foresee a particularly strong advantage by one team over another. This is all from a 2300ish game sample size.



Now, for the fun part.



I was able to cross-reference all my simulated 2018 picks with historical betting lines for those games. I made sure to pick games that I had an assurance would be a profitable wager for me, which was centered around my accuracy percentage and the betting lines that I was up against. Although none of my simulated picks involved actual money, I was able to "turn" $1,000 into $2,500 over the course of the season. Only on the first day of the season did my bankroll fall below $1,000.



Of course, to many of you, I am but a pesky Shelby fan. Why should any of you take my word on this?



You shouldn't. Not yet, at least. I hit some unexpected delays in preparing the MLB prediction model for 2019, but I am going to be ready to launch it in the next seven days. From there on, I will be publishing my picks until I either move to Las Vegas or Ohio legalizes sports gambling. I am posting very early in order to get some thoughts / questions before I start showing off my work here. I will begin to track my results on here as well, simply so you guys can get an early inside look at what I hope this thing could hopefully become.

[shelbyrr11]

I have been posting my results on Twitter.

My overall picking record: 16-9 (64.0%)
My record against the best money line: 14-5 (73.7%)

The twitter is @scooperman_betr

[shelbyrr11]

Update going into the All-Star Break.

I haven't yet finished testing this model for 2019. In 2018, I was able to predict 57.4% of all games correctly. The 2019 model has a few calculation changes which will drive more accuracy, so all I have to work with is that Jul 4 - Jul 7 data.


Overall Predictions: 34-21 (61.8%)

My Recommended Picks (against the money line): 23-11 (67.6%)

[shelbyrr11]

After 7 full days of games predicted:

Overall Accuracy: 63-38 (62.38%)

Accuracy in games I recommend to bet on: 44-27 (61.97%)


Using Vegas lines offered on the day of games and deploying funds in pretty uniform fashion, my simulated portfolio has gone from $5000 to roughly $12,800 in these seven days.

I will be going to Las Vegas to bet on games during July 25-28 so I can put my money where my mouth is.

[heresjim]

Jul 15, 2019 8:05:43 GMT -5 shelbyrr11 said:

After 7 full days of games predicted:

Overall Accuracy: 63-38 (62.38%)

Accuracy in games I recommend to bet on: 44-27 (61.97%)


Using Vegas lines offered on the day of games and deploying funds in pretty uniform fashion, my simulated portfolio has gone from $5000 to roughly $12,800 in these seven days.

I will be going to Las Vegas to bet on games during July 25-28 so I can put my money where my mouth is.

So... How is the newly amassed fortune?

[shelbyrr11]

Aug 4, 2019 7:34:31 GMT -5 heresjim said:

Jul 15, 2019 8:05:43 GMT -5 shelbyrr11 said:

After 7 full days of games predicted:

Overall Accuracy: 63-38 (62.38%)

Accuracy in games I recommend to bet on: 44-27 (61.97%)


Using Vegas lines offered on the day of games and deploying funds in pretty uniform fashion, my simulated portfolio has gone from $5000 to roughly $12,800 in these seven days.

I will be going to Las Vegas to bet on games during July 25-28 so I can put my money where my mouth is.

So... How is the newly amassed fortune?

Things are going well.

I had expanded wager amounts than the $1000 example in the attachment, so I was able to do well enough in Vegas to pay for the entire trip.

I am arranging many more plans to go out there for the duration of the season.

Attachments:

[heresjim]

Aug 4, 2019 9:08:25 GMT -5 shelbyrr11 said:

Aug 4, 2019 7:34:31 GMT -5 heresjim said:

So... How is the newly amassed fortune?

Things are going well.

I had expanded wager amounts than the $1000 example in the attachment, so I was able to do well enough in Vegas to pay for the entire trip.

I am arranging many more plans to go out there for the duration of the season.

That's not too bad. How well does this system perform compared to other people's algorithms at this point?

Also, considering Pennsylvania legalized sports betting, would you just consider driving to Erie instead rather than travelling to Vegas? Might be more economical.

[shelbyrr11]

I'm not sure how I stack up to others. I am presently hitting about 56% accuracy (I was 57.4% in my 2018 simulation). I am comfortable saying I'm 56-57% accurate.

All I know is that 56-57% is higher than, say, card-counters can aspire to hit (they usually top out at 54%). I don't know what sports handicappers usually hit, specifically baseball. I think I am in a normal range, although I have several ways to amplify my accuracy once I get time during the off-season. For example, I tend to perform notably worse on travel days, and I have the beginnings of a calculation that does a fine job of predicting output by a rookie based on his age when he comes up to the big leagues.



Pennsylvania sports gambling is not a bad shake, in all honesty. My strategy involves using multiple casinos (4-6 casinos) in order to hunt for the best lines and to minimize my presence at any individual casino. In Vegas, it is very easy to set-up an entire network of mobile betting in one day (which I did). When I started this, PA was having fluctuations in their mobile sports betting, so I stayed away from it. Your question did inspire me to look into PA again, and I think it is actually viable. I'll need to road trip to Philly, Erie, and probably reside in Pittsburgh full-time, but the state does offer a pretty satisfactory situation from what I am seeing.

[shelbyrr11]

Alright I'm back with more updates.

I quit my job to pursue this baseball gambling full-time in Pittsburgh.

My model has been batting 58.3% since I started picking games in July. My daily Return on Investment percentage is over 5% gain. That's daily. At that rate, I am able to double my money about every 25 days or so. Which I have managed to do. 

If any of you are familiar with FiveThirtyEight, they are an analytics venture started at ESPN by Nate Silver. They eventually got bought out by ABC, and then Disney. They have their own prediction systems for all sorts of sports. Either way, my model has consistently beat theirs since I began tracking their results. 

I will happily answer any questions.

I post all my picks and results on my twitter, @scooperman_betr

[heresjim]

Sept 11, 2019 17:09:32 GMT -5 shelbyrr11 said:

Alright I'm back with more updates.

I quit my job to pursue this baseball gambling full-time in Pittsburgh.

My model has been batting 58.3% since I started picking games in July. My daily Return on Investment percentage is over 5% gain. That's daily. At that rate, I am able to double my money about every 25 days or so. Which I have managed to do. 

If any of you are familiar with FiveThirtyEight, they are an analytics venture started at ESPN by Nate Silver. They eventually got bought out by ABC, and then Disney. They have their own prediction systems for all sorts of sports. Either way, my model has consistently beat theirs since I began tracking their results. 

I will happily answer any questions.

I post all my picks and results on my twitter, @scooperman_betr

View Attachment

I'm surprised you keep posting your results in public. Are you concerned that you'll start influencing spreads once people catch onto what you are doing?

Besides that I have a few questions. What software do you use to make these algorithms and interpret data? And how viable is it to predict other sports games (such as pro football, basketball, or hockey respectively)?

[shelbyrr11]

There is some concern, but if that many people follow my picks, the plan would be to take my picks behind a paywall and monetize access to the picks. Right now, I am trying to drum up interest in what I am doing. I think I would be under less stress if I have stable income through a subscription model. The better term is diversification of revenue streams. Even if I influence the spreads to some extent, you'd be surprised how often the casino spreads are dramatically out of touch with what I project.

I like to keep my software usage private. One of the reasons why I was emboldened to pursue this was I put together snippets from Q&As and interviews of other professional sports gamblers. I don't want to make the same mistake. The Oakland A's greatest mistake was allowing MoneyBall to be published as a book. Once they did, everyone had all their playcalls. I don't intend on showing my hand. This is technically intellectual property.

I think there is viability in other sports. My system is extremely-tailored to handle baseball situations. There is no aspect of my system that can be replicated to the other sports. That isn't to say the theories and logic couldn't be applicable, but the actual math itself is wholesomely restricted to baseball. If I were to ever go into another sport, I believe NCAA basketball is going to be the next pursuit. This is simply because of the volume of games. From my perspective, baseball is great because there are numerous games every single day. If I ever have a bad day of gambling, I can bounce back the very next day and wipe my conscious clean. If I were doing football gambling and I have a bad Sunday, I would have to sit there on the couch on Monday...Tuesday...Wednesday...self-doubt settles in...Thursday...Friday...oh god I hope I don't have another bad Sunday...

That ain't what we call a good program for mental health, haha. And I don't believe many systems can push accuracy much higher than 60%. I'd rather play baseball with 2400 games than the NFL with much less than that.

NCAA Hoop? 300+ teams, games every day, and a lot of un-knowledgeable folks pushing lines around on random games. I think with other analytics out that like KenPom that there are teams that display a signature style in their statistics that would make picking winners fairly lucrative. I won't fry that fish until I perfect my baseball operation, which I intend on doing during the offseason.

Keep asking questions if you have them! I don't mind.

[heresjim]

Hmm... Well could you tell me if the software is free or not lol(if it doesn't give away what it is) and if it has capabilities that make it superior to R(which I'm learning at the moment)? I can do the necessary digging to figure things out, I just don't want to dedicate time and find out that I will not be able to afford it. And also... how long did it take to produce that model for MLB (in terms of hours, including learning the software)?

[dirkadirka]

I'm sure there are ton of variables in the model, but I'm curious which variable you use has the most explanation for the result of a game. Is it a calculated one?

Interesting stuff.