Backtesting without Bookies odds

[trueblue2]

Regarding using Oddswizard to backtest how profitable Home wins / Away wins / Draws are at the moment is hard work because Oddwizard is unable to do it in the software itself. At the moment you have to export the odds into a spreadsheet, get the odds near kickoff using an automated system and then backtest at the end of the season. What would be fantastic is Oddswizard actually doing this itself, whereby it takes the Bookmakers odds and it's own odds and calculates the Profit and Loss if backing outcomes.

The problem with Oddswizard at the moment is that when it calculates it's success rates etc, it is not taking into account the Bookmakers odds and whether it's predictions make money. This needs to be addressed for Oddswizard to move forward. For example Oddwizard may predict something that has a 50% chance of winning, which therefore is saying if you back at odds of greater than 2.0 you will make money in the long run. However, in this prediction you will get selections where the Bookies odds are greater than 2.0 and less than 2.0, therefore knowing whether you'd have made a profit is impossible. I know for a fact that backing these selections at greater than 2.0 does not make a profit. The market is king !

I wonder if the new version ( shots on target etc ) will take this into consideration.

[Alex]

I am sorry for the delayed response. Trueblue2, You are not the first in your desire to make such a backtest. In order to fine tune the algorithm, we made a research backtest during just two seasons. Provided that we use the simplest Kelly stakes, we managed to get positive ROI ~ 5% for AHO in major football leagues. Other leagues are more subjected to human factors, and showed ROI about 0+/-2%. We took average odds from 188bet.

However we intentionally did not make such an instrument with Odds Wizard. There are too many bookmakers with totally different odds. Moreover, the odds are subject to change within the same bookie. Odds market is too violatile to make certain conclusions on the basis of such backtests. I intentionally used the term "research" to emphasis the complexity of this issue. It's not yet a time to implement feature you are asking for, sorry.

[edo]

the best way to back-test under such criteria is to compare to pinnacle closing line... if your model is able to have closing line value against pinnacle, you are in the right path.

[DrWho]

From my own betting, and with around 2700 games logged I have an edge of around 12-14% in the HDA markets. But the data is filtered using some criteria I'm not going to divulge but which uses nothing outside of OW data and Betfair prices. I don't bet the 1st markets so cannot comment about those. The CS market however is the most profitable but with long losing runs so you require a big bank and strong stomach....