Friday 18 July 2014

Kelly In Action

Prompted by a Cassini subscriber who wants to look at the 2013-14 Cassini Value Selections in more detail, I went through all my old emails and recorded the edges in a spreadsheet which is something I should have done all along. I expunged four matches from the records (three winners and one loser) because they were from the early days and were lays or draw no bet selections which left me with 117 pure selections and profit of 23.69 points. A small sample, but it was interesting that when sorted by edge percentage, the top five, and six of the top seven, were losers. There seems to be some truth in the saying that if something appears too good to be true, then it probably is. Selections with a supposed edge in excess of 30% all lost:

My calculated price is in column E with Pinnacle's official price in F and the edge in G.

At the other end of the range, there were 19 selections which, at the time the email was sent, had an edge of over 10%, but were less than that in the recorded Pinnacle prices. Those 19 selections were still profitable, generating 5.05 points.
One embarrassing and, in hindsight, avoidable loss was the selection of Bayern Munich to win at Augsburg in April. A price of 1.4 at the time the list was compiled looked good enough, with a 20.4% edge, to justify inclusion, although I did caution in my email that:
One comment is that Bayern Munich have clinched the league and their motivation may not be what it was.
In the event, Bayern fielded "an experimental side" and models don't work too well with "experimental sides". The motivation factor also cost me a few points at the end of last season. Once April was over, the results were poor:
One of the benefits of having a precise number for your edge is that if you are confident in that number, you can maximise your returns by using the mathematically proven and optical  Kelly Criterion.

I think the subscriber who prompted this review may be thinking in terms of this, and using a £1,000 bank, full Kelly would have ended the season with the bank at £2,837.58 (using the prices at email time) or £2,782.37 using the recorded Pinnacle prices.

The more conservative half-Kelly results would have been £2,524.34 and £2,490.85 respectively, and the quarter-Kelly  numbers are £1,750.52 and £1,736.14. Interestingly the maximum bet here would have been around £248 on the aforementioned Bayern Munich loss.

One other statistic I looked at was the price at email time versus the Pinnacle recorded price. (Worth mentioning that the Pinnacle recorded price is not necessarily the closing (kick-off) price but is the price recorded by Football Data on a Friday afternoon).

Anyway, of the small sample of 117 matches, 60 shortened in price, 20 remained unchanged, and 37 drifted.

The 60 that shortened by a total of 2.27 points made an official profit of 13.13 points. The unchanged made a 2.85 points profit, while the drifters added 7.71 points despite drifting by a total of 1.07 points. What was interesting was that the bigger drifters (those of 0.03 points or more) were net losers, while those selections shortening by 0.03 points or more were net winners. Follow the market.

Some lessons to be learned as the new season fast approaches - less than a month away in fact.

No comments: