BTW, for people who want to make money don't underestimate the power of having a good database. I know you can tail people, pick off stale lines, steam chase, etc. but you're leaving money on the table if you don't do this.
Let's look at a common prop that is offered: Will there be a score in the first inning. You can take two approaches to this: Use your player projections to determine a fair line or you can use historical data to look at when the total was X, was there a run in the first inning.
Some unreasonable assumptions you make doing this: All games with a total of 8 behave the certain way and yes you will probably make some bad bets as a result, but overall it is a profitable strategy. The full game today is preferable over the 5 inning line due to there being little range in the 5i line. Maybe if you are dedicated you could separate into 3.5 based on some cut off value (-120?), but that seems like a lot of work. The limit on these props tend to max out at $500, so why not just take easy money.
So here is what I did, I subsetted all the MLB o/u since 2014, then calculated how much you would win against Pinnacles closing line for is there a score in the first inning. For 2015, if it was May 27, 2015 it used all the data from 2014 + anything before May 27, 2015. May 27, 2016 used all of 2014, 2015, and April-May 26, 2016, etc. I assumed a $500 max bet (Pinnacle is higher, but this prop is available at a lot of sportsbooks with a $500 max, also if odds are < 0 then it's to win 500, > 0 risk 500) and that you would bet just once.
I started with four theoretical bankrolls at the beginning of each season: 1k, 5k, 10k, 25k. Then I looked at the closing line for run scored in first and if there was an edge, I bet it based on 1/3rd kelly value. At the beginning of each day the bankroll adjusted based on the previous day W/L (e.g., if our initial bankroll was 1k and on day 1 we won $100, our day 2 bankroll would be $1,100). At the beginning of each season I reset the value to the starting bankroll (say you wanted to withdraw or whatever, but this was to highlight that each year it's a profitable strategy).
Below are tables for how you would have done each year.
Profit every single year. Even a "down" year of 2016 still returned a 2% ROI. Also note this is against CLOSING lines. There are tons of bets that are also available closer to closing which get bet down and either lower their value or make them unbettable. Also this is using Pinnacle lines which are sharper than other books. So this is definitely on the low end of what is possible (And from betting this every year, it is definitely the way low end).
How long would this take to accomplish?
Maybe if you have 0 background on databases and scripting - to get a database setup 30 hours?
Scrape the data, if you never did it and need to learn 10 hours?
Then all you need to do each day is write a script to generate the proper RSIF line based on the full game total, look at the lines, and if there's value put them into a kelly calculator.
If you wanna get fancy and have access to XML feeds you can do this all automatically or you can set up a thing where you type in the line and it spits out how much to bet.
So what 15-20 minutes a day to check the lines and bet? And starting with 25k you can make 14.1k, 7.8k, 7.8k, and 13.6k over the four years? Would you really make that much more if you develop a super complex model to predict if there's a run scored in first? Maybe you'll make an extra K but is it worth it? Use the time to make money elsewhere.
And this is just one prop. HRE behaves similarly and has had better returns over the time frame.
There's money out there. Building a good database and learning how to look for meaningful trends (e.g., it makes sense that RSIF should be related to the total, same with HRE, modeling total K's with total doesn't make sense but it might be profitable by pure chance) is a good place to start.