Friday, 12 August 2016

Best Of Five, Certainly

Smithlondon had a comment on my findings regarding backing the Draw in the EPL, and has gone further by conducting further research looking back further than four seasons.

To clarify, the reason I chose the past four seasons is that these are the only seasons for which Pinnacle’s data is available, and now with the added bonus of the closing prices.

The reasons Pinnacle are the sportsbook of choice are that they have the most favourable over-round at ~102%, and they do not restrict or close winning accounts.

Here is Smithlondon’s comment:

"Backing the draw in the (conveniently exactly) 1,000 matches where the implied probability of the Draw was 0.25 or greater at kick-off would have boosted your account by 49.22 points, and you can all work out the ROI% on that. (+50.33 from 1,017 selections if you bet early)."
As always your blog is an interesting read. The above caught my attention. I don't currently bet on football but I'm always looking for new systems to add to my portfolio. I believe diversity is important, so naturally those numbers piqued my interest, and with the season starting I decided it was worth a few hours of my time to have a look at. Even if it was more in hope than expectation. So I data dumped the sheets from football-data (excellent website) and went to work. I took the odds of five of the bookies listed (B365, BWIN, Lads, Hills, VC) for the last 10 seasons. I used those as they seemed to be the only ones listed for the complete 10 years. Averaging the odds and then applying your criteria (Back the Draw when implied prob > 0.25) gave me a return of 7.73 points from 1134 for the last four seasons, but a loss over the 10 seasons of -96.16 from 2937. I got closer to your four year results when I adjusted the average odds to represent a 100% book (my proxy for exchange odds). In that case I got a 39.81 point profit off 1005 (while applying a 5% commission) but I still registered a loss of -26.35 over the 10 seasons(2689). While I'm not suggesting your approach or numbers are wrong, a longer term view dampens the enthusiasm somewhat although I'm still searching for your filter.
The over-rounds on B365, BWIN, Lads, Hills and VC are all higher - VC (selected at random) is 103.22% - so you are always up against it with these books, and if you look like you have anything resembling a clue as to what you are doing, your account will be closed or restricted to peanuts anyway. Bottom line is that these books don't bring a lot to the party for the serious bettor so I try to ignore them.

Just how bad are the non-Pinnacle books?  Of the 1,520 EPL matches over the past four seasons, Pinnacle (outnumbered 5 to 1 remember) were beaten for best price on the Draw 604 times, less than 40% of the time. Of course, if you can get on with the sportsbook offering the top price, you should do so - using Pinnacle's prices here are for record keeping, and returns given are always the least you can expect.

So I’m not sure that AVERAGING the (already poor) odds on these five books makes any sense. Pinnacle may not always be the best price, but they are rarely too far off, so to get anything like a meaningful comparison going back an additional six seasons, a better approach would be to use the best (MAXIMUM) available price. 

Curiosity, of course, got the better of me, and I tried this approach using Smithlondon’s choice of books for the past four seasons, and here are my findings:

2012-16 Best of Five 1,069 bets, +47.65, ROI 4.46%

2012-16 Pinnacle Early 1,017 bets, +47.18, ROI  4.64%

2012-16 Pinnacle Closing 1,000 bets, +49.22, ROI 4.92%

Unfortunately I don't have the time to go back six more seasons and run the numbers, but the Best of Five are comparable to Pinnacle's profits, and the Draw hit rate between 2012-16 was 25.4%, while for the six seasons 2006-12 it was 26.1%. Perhaps Smithlondon can re-run the numbers using the best prices?

It's quite possible that the six seasons up to 2012 were not profitable with this strategy, because both the EPL and the markets are constantly evolving. How many seasons to look at when evaluating a strategy or a system is up to the individual. I personally think results from ten years ago in the EPL are totally meaningless, although some 'experts' have no problem using data from past match-ups, despite the fact that the players, managers and possibly venue have all changed since previous encounters, e.g. "Twelve of the last thirteen EPL meetings between these two sides have seen at least two goals", as if a match six or seven seasons ago has any relevance today.

I will spare the below preview "writer's" blushes by not linking to his site, but count the 'certainly' references in just two sentences!  
Someone is certainly certain...
Nothing is certain in betting, and it’s a paradox of profitable betting that if you wait until you have a statistically proven edge, the edge will be gone. In my opinion, the trick is to spot a trend (and an edge) early, act on it while maintaining a healthy scepticism that the trend may be fleeting and evaporate at any moment, and stop when you determine the edge is no longer there. Right now, the Draw in the EPL appears to offer value in matches of a specific profile. Was value there in 1985? Will it be there in 2020?

Mixing metaphors I know, but no one should be a one-trick pony, nor have all their eggs in one basket. Smithlondon mentions he has a portfolio, and this is the best approach. Track the performance of each system appropriately, constantly develop new systems, and add / remove them from your portfolio as the results dictate. The benefit of a portfolio with diverse systems is that the variance of any one individual system is moderated. It’s an approach that works well with financial investments too. Putting all your money in one stock is a recipe for disaster – use mutual funds, preferably passive Index Funds!

1 comment:

Smithlondon said...

Fair points about the bookies selected. Finding the right price is obviously the key to all/any strategy. As requested if you take the best of those 5 prices over the ten years, you get an 8.65 profit from 2808 selections.

06/07 -22.65 316
07/08 -7.4 301
08/09 -28.85 303
09/10 -1.5 264
10/11 30.7 289
11/12 -9.3 266
12/13 55.5 269
13/14 -33.85 251
14/15 7.15 271
15/16 18.85 278
8.65 2808

Profit results from some big swing years (10/11, 12/13 & 15/16) all of which had (unsurprisingly) above avg draw % in those respective years.

06/07 47.89% 25.79% 26.32% 100.00%
07/08 46.32% 26.32% 27.37% 100.00%
08/09 45.53% 25.53% 28.95% 100.00%
09/10 50.79% 25.26% 23.95% 100.00%
10/11 47.11% 29.21% 23.68% 100.00%
11/12 45.00% 24.47% 30.53% 100.00%
12/13 43.68% 28.42% 27.89% 100.00%
13/14 47.11% 20.53% 32.37% 100.00%
14/15 45.26% 24.47% 30.26% 100.00%
15/16 41.32% 28.16% 30.53% 100.00%