Friday, 4 October 2019

London Losers and Wednesday Winners

Week 5 of the NFL features the first games in London for the season, and Axios points out the interesting curiosity that none of the 25 games in the capital so far have been between two teams with winning records. 

Contrast this with the games played in Mexico City, where one of the two was between winning teams, and the third scheduled game last season was to be between two 9-1 teams except that the field wasn't fit to play on, and the game moved to Los Angeles. Ouch.

After the Carolina Panthers play in London next week, only one NFL team will have never played in London, and unless both the Panthers and Tampa Bay Buccaneers win this weekend, the streak will move to 26 games.

Regular season games played outside of the USA are fairly rare and consequently have a small sample size, but of the 32 matches played to date, the favourite has covered 67.7% of the time and 85.7% of the time when the 'dog is coming off a win. The logic here is that a combination of the recent win and the effect of a neutral site is given too much credit by the betting public, and the line is lower than it should be.

Based on current lines, we should have seven Small Road 'Dogs this weekend, looking to maintain their 73.9% record so far this season.

We should have about the same number of selections in the College version of the game looks likely as we hope to improve on the current 17-12-2 record (left). 

As a reminder, this market inefficiency has persisted with more winners than losers in every season throughout the third millennium.

Regarding the end of regular season baseball summary and some of the systems I mentioned therein, Spot On Parts had this to say:
I've covered this before, but having a logical premise or rationale for why a market might be inefficient and then looking at data to either confirm or disprove the theory is quite different to data mining, an example of which is this one from a few minutes ago:
'Less' should of course be 'fewer' but regardless, one should never be too hasty in saying that such a finding has no predictive value - just because you don't recognise a cause, doesn't mean that there isn't one (Sumo Round 15 anyone?) - but it seems highly unlikely that a combination of fouls, being on the road, and a recent win at such specific values and only for the Detroit Pistons, and it all started after January 15th 2015, has any predictive value at all. It's random nonsense, the product of some great data mining (change the 20 fouls to 19 for example and 11 wins vanish) and it's not helpful nor all that interesting. There must be thousands of such nuggets (not Denver, the golden kind) hidden in the database, but the key is there has to be some rationale behind it. 

Spot On Parts then tweeted the below before rushing to the hospital to have his sides stitched back together:
He's referring to the fact that in Wednesday games, the New York Yankees have a Money Line ROI of 8.4% over the 408 games in the database. 

Now this is an example of something that could be chance, but in baseball, because of the way games are scheduled, Wednesdays (and Sundays for that matter) see more games of a certain category than other days, so Wednesday results would be expected to differ from say those on a Monday. 

The Yankees ROI is actually 90.9% on the road in day games against teams they have more wins than, and following an extra inning game. 

On the flip side, the Los Angeles Dodgers have a -12.8% ROI on Wednesday games and they are scheduled to play the Washington Nationals next Wednesday if a deciding game is required.  

No comments: