Tuesday, 1 October 2019

MLB 2019 Season Summary and Playoffs Preview

The 2019 MLB Regular Season is in the books, 2429 matches instead of the expected 2430, which nevertheless makes it by far the best popular sport for data analysis. By way of comparison, the NFL is much studied, but analysis can be problematic with its low sample size each season as well as lop-sided schedules and the impact of losing one key player etc. 

So how did our baseball systems do in 2019? 

The well established T-Bone System relies on the market underestimating a team coming off a loss - aka recency bias, and there are no signs that this edge has been eroded. 

The Money Line and Run Line bets combined for 42.8 points of profit, the best season since 2012, although once again as expected, September wasn't profitable

July was the "hottest" month with 18 winners from 20, and the steady increase in selections continued with a new high of 180 matches. 

The average line was also a new record, the shortest ever at -177.3 (1.564) and the strategy of not backing the Run Line bet when it is an American League day game continued to benefit the bottom line, although that is getting into a little too much detail.

The Home Improvement System from 2016 had a strange season, with the Money Line up by 23.45 points (an ROI of 19.3%) but the Run Line down by 4.2 points (an ROI of -4.7%).
As has been mentioned several times, backing hot favourites (defined as -200 and shorter) continues to be profitable:
For the Totals Systems, the Overs System started the season including 9.5 run games and finished with a 56.1% record. 

After seeing an issue with the 9.5 run total, I adjusted the system after the All-Star Break to exclude these, and the season total for the new Overs system finished with a 65.9% record.

To be fully transparent, the numbers before the All-Star Break were 98-65-8 (60.1%) and after the break were 72-49-4 (59.5%) for a total of 170-114-12 (59.9%) which are very strong numbers for such a simple system.

The Unders System wasn't modified during the season, and finished with a record of 83-60-4 (58.0%). 

Here is a summary using the Overs System we started the season with, as we stay fully transparent:

Using the new Overs parameters for the whole season would have been wonderful, but hindsight isn't something most of us possess. 

At the risk of turning this post into something of an eye chart, here are the results of the Extra Innings System:

The Doubleheader System failed miserably this year, with only three wins from the six opportunities, but ROI percentages of 23.9% on the Money Line and 21.8% on the Run Line aren't going to be discarded after just six bets. At least the more profitable away selections had a winning record. 

Plenty to look at in the close season, and whether you followed one or all of the systems and ideas from this blog, you would have struggled to lose money this season. 

Results for all the systems combined in 2019 are below:

310 points profit, systems all verifiable, and all disclosed in this blog.

Meanwhile the season goes on with the playoffs starting tonight with the National League Wild Card game - Milwaukee Brewers @ Washington Nationals.

The Wild Card games started in 2012 and so it's a small sample size, but for what it's worth the road team in National League games have the edge in these games. It's a winner take all game so worth watching if you enjoy watching meaningful sports. 

The database goes back to 2004 and contains 503 games. Overall the table below might help you identify where the market weaknesses have historically occurred in playoff games:

The 12% / 8.1% ROIs on Home 'Dogs is what jumps out, and over the past five years the returns hold up at 11.4% and 5.5% but the low number of selections means any one season may well end in a loss, although only 2016 has done so since 2010, but don't go crazy and lose too many of those 310 points.

Should any playoff games go to extra innings, forget the Home 'Dog next time out, and go with the favourite.


MP said...

Hey, nice work!

Any chance you could post links to descriptions of the current versions of your systems? I've tried to hunt them down in the archives, but to no avail.

n00bmind said...

Hi Mr. Cassini,
Not sure if you read blog comments but, about your statement that "hot favourites continue to be profitable", I've started calculating my own stats for these and my numbers seem to be pretty different to yours..

Season Bets Points
2019 118 -4.60
2018 423 8.24
2017 304 1.31
2016 278 -7.85
2015 151 8.84
2014 115 13.13

I only have stats for 2019 up to about mid-June, but even ignoring that, our number of selections seems to be somewhat different, pretty weird given the easy selection criteria (unless I'm missing something here).
Also, given how low the point-profit per bet is, your numbers suggest one should get quite a huge strike rate, which is just not what I'm seeing.

May I ask whether you compile odds yourself or you use any public dataset, and whether your stats are using Pinnacle's closing prices?

Love your blog btw, have been following you for about a decade now, please keep it up!


bumby said...

Let me start out by saying I really enjoy your blog.

I am having trouble replicating your numbers for the T-Bone system. I am probably doing something wrong but I was hoping you could help clarify.

The system is simply road favorites -140 or > (ie -150 -160) who are on a losing streak of 1 or more games. Is this correct?

Is there an additional filter I am not seeing?

From Betlabs this system would have produced a loss of -7.97 units for a -1.2% ROI since 2015. Perhaps its the betting system they employ that is different.

I think they use a flat bet (risk) of 1 unit on all situations and you are betting to win 1 (risking more than 1 unit) when it is the favorite. Perhaps that accounts for the difference.