# End of Season MLB Algorithm Analysis

I did one of these in August for an in-season assessment of the MLB algorithmic action. Now that the season is over, amazing how fast it went by, I wanted to look back on the performance of the daily MLB algorithms. My goal with this assessment is to then setup a tentative plan for next MLB season. As some of you may know, the daily MLB algorithms (see below) that I post to Twitter are all algorithms that have been designed from scratch in 2019. I have other 1st 5 innings and full game algorithms and used them for years. However the two below were designed in 2019. Looking back on the season, I am quite pleased that the first season ever doing these algorithms has been very profitable. In the end, the algorithms (all posted free daily to Twitter) were 58.5% and +40.2 units! If you maintain the 100 unit bankroll I advise, these algorithms brought you a 40.2% ROI for the season, not bad for free and their rookie year!

Let’s break the season in half and see how it went. Here we go…

At the All-Star break, the daily MLB algorithms had the following records…

1st Inn O/U version 1.0 (39-24 +12.9 units)

1st Inn O/U version 2.0 (35-22 +13.2u)

1st 5 Inn O/U (39-25 +12.4u)

Full Game O/U (24-15 +8.0u)

Overall MLB: 137-85 (61.7%) +46.5 units

At the end of the season, the daily MLB algorithms have the following records…

1st Inning O/U version 1.0 (47-38 +3.7u)  -9.2 units since All-Star Break

1st Inning O/U version 2.0 (38-23 +15.0u)  +1.8 units since All-Star Break

1st 5 Inning (46-28 +16.0u) +3.6 units since All-Star Break

Full Game (31-24 +5.7u) -2.3 units since All-Star Break

Overall MLB: 161-114 (58.5%) +40.2 units

Obviously these algorithms were monsters prior to the All-Star break. For the first half of the season they won 46.5 units and hit at a rate of 61.7%. The reason for this is often simple, the public is betting based on last years data for a good portion of the start of the season. It allows you to find some serious value, not just the MLB markets but most sports early in the season. Then there is usually an adjustment. Bettors now base decisions on the current season’s data. The books tighten up their lines and those tremendous values are no longer found. Then there are changing factors like temperatures getting warmer, humidity rising or falling, pitchers fatigue as the season goes and as the heat turns up, etc. When you combine those various factors, all the UNDER wins we achieved in the cool April and May temperatures give way to the balls flying a little further and pitcher fatigue setting in.

Looking back on the season the algorithms won 46.5 units for the first half of the season and then lost 6.3 units in the second half of the season. It is very interesting to see such strong performance in something and then on a dime, a lack of performance. The big problem in the second half of the season was the 1st Inning v1.0 algorithm. The 1st Inning v1.0 gave back 9.2 units. The next worst performer was the full game algorithm which dropped 2.3 units. Not really a big loss for the full game algorithm. Comes down to 1 or 2 games not going our way in the second half of the season. However v1.0 and the full game algorithm combined for -11.5 units of loss in the second half of the season… OUCH! On the positive side, v2.0 for the 1st Inning and the 1st 5 Innings algorithms both ground out +5.4 combined units in the second half of the season.

In analyzing this performance, the game plan for next season may be to play all four algorithms through the All-Star break. Obviously they were strong and no reason to do anything other than follow. After the All-Star break I plan to continue playing v2.0 and the 1st 5 Innings algorithm. At the same time I will monitor (but not bet on) v1.0 for the 1st Inning and the full game algorithm. I will monitor and regularly analyze all the algorithms after the All-Star break to see if the performance this past season was a fluke or a consistent pattern. You never really know until you build up data. It could just have been a bad second half of the year for v1.0 and the full game. Maybe next year will be good beginning to end. Only time and analysis will tell. However regardless, it was an amazing year from the algorithms  and I look forward to firing up their engines again next spring!