Friday, 24 July 2020


A new web site appeared on my radar a couple of days ago when Toni sent me an email:

hope you are doing well. Recently, I came across your blog which I like very much, lots of useful information. My name is Toni and I run a small website on betting strategies and automation. Your "Bundeslayga" method caught my attention and I would like to ask how you adjust the lay price for the market overround and what staking you use?
The reason why I am asking is that I've tried to replicate your approach at with some Python code but unfortunately my results differ from the ones published on
After clarifying the method and the staking strategy (risk one unit on each bet), Toni responded with:
Many thanks, Cassini. Your clarification was very helpful. After adjusting my script I obtained pretty much the same result - still a small offset, probably due to rounding / lower precision floating point calculation on my side.
He then asked me for my rationale behind adjusting the odds to produce a baseline:
I am struggling a bit to understand your motivation behind using the adjusted odds as baseline to report on profit/loss figures. I mean as a Pinnacle user, I wouldn't be able to place bets at the adjusted price to achieve the reported profit/loss, would I?
Since I typically adjust the prices to a 103% over-round, and the Pinnacle over-round is (for the EPL at least) usually below this number, then certainly you could expect to beat the results shown, but the reason is primarily to be able to compare results for a strategy from one season to another, even where the over-rounds may have been 112.5% in one season (EPL 2000-01) and 102% in others (EPL 2014-17).

Clearly using the raw prices doesn't allow this. The idea and process isn't perfect of course, but as I told Toni:
...the purpose is not to give a precise P/L but rather to show where the market isn't efficient and let anyone who is interested work out how best to take advantage of that.
As I've written in the blog before, if 100 people applied any system in practice, you would get 100 different results.

No comments: