29th December, 2018.
Email: [email protected]
Following the end of the 2018 season, I performed some detailed betting analysis on the players highlighted as representing greater than the historically used mark of 4% implied percentage value between model and Pinnacle prices, using the formula (100/model price) – (100/Pinnacle price).
I have always advocated using Pinnacle as an accurate assessment of profitability, as opposed to ‘soft’ bookmakers. These ‘soft’ bookmakers may offer better prices on occasion, but will also be likely to create issues both for getting substantial stakes on, as well as eventual account closures or restrictions.
With such a long-term sample of data – over 2,000 non-round one market discrepancies on the ATP Tour were highlighted – the sample size is rather robust and you can download the full P&L spreadsheet, sorted by model vs market edge percentage, below. In addition, after some requests by users, I have added 1/4 Kelly staking (based on a sample £5k bank), P&L and ROI to the spreadsheet, in addition to the traditional ‘to win x’ staking method. While 1/4 Kelly is still a little on the aggressive side for a staking strategy (a high value proposition is around 15% of bankroll), it does provide some meaningful comparison from an ROI perspective for Kelly staking compared to the ‘to win x’ staking method.
An assessment of the data indicated that there were a number of sweet spots where my model has historically outperformed the market to a strong extent, and these include the following areas:-
Higher edge scenarios hugely outperformed smaller edge scenarios, indicating that huge value prices do not tend to be ‘too good to be true’.
When model to market edge was greater than or equal to 12%, ROI was 5.50% from 328 bets (to win x staking) and 6.39% based on Kelly staking. The higher the edge, the better results were generated from Kelly staking, which is completely logical.
There was a further strong spot where model to market edge was slightly lower (10.00-11.99%) and value player was priced as a favourite by the model.
This generated a further 124 bets which generated 3.68% ROI (to win x staking) and 4.58% based on Kelly staking.
Favourites and slight underdogs considerably outperformed heavy underdogs. This is nothing new, and is in line with my previous analysis across both the ATP and WTA Tours, and is not a surprise given that bookmakers tend to load a bigger profit margin on underdogs.
Highlighted price discrepancies when market price was 3.00 or below generated 3.09% ROI from 1291 bets (to win x staking) and 5.02% based on Kelly staking.
Considering the above, I will be making the following adaptations to the highlighted value spots on the daily spreadsheets for the coming 2019 season:-
* Players with a model to market edge of greater than or equal to 12%, or greater than or equal to 10% when value player is priced as a favourite by the model, will be highlighted as a strong value spot (purple highlight on the daily spreadsheets).
* Players with 4.00-9.99% model to market edge will still be highlighted in red as previously, but only when priced 3.00 or lower. In addition, players with 10.00-11.99% with price 2.00-2.99 will also be highlighted in red, as previously. The red highlight will now indicate a lower value spot.
* Price discrepancies where market price is above 3.00 will no longer be highlighted.
* As has been the case in previous years, betting in round one, where player fitness and motivation is often questionable, is not recommended. A few people asked about round-robin matches in Tour Finals, which is absolutely fine given the large financial and ranking point rewards in these events.
If you have any questions or would like any further information, please feel free to get in touch via email ([email protected]) or via Twitter (@tennisratings).
