Sharing my system - looking for feedback

[allen29]

Hello all, For the past 2 months I have developed and tested my system and I decided that I want to share it that I get some opinions and maybe improve it. This is how it works: The systems starts after a number of games in a league have been played. I use 13 played games. In excel, thanks to Froment, I have some forumlas that calculate number of matches played by home team until the current game, number of matches played by away team until current game, numbers of goals scored by teach team, average numbers of goals scored, number of Home wins, Home Draws, Home Loses, Away wins, Away draws, Away Loses, both in absolute terms and in percentages, very important, UNTIL the current game. Then, I construct 3 probabilities for 1, X, and 2 as follows: -HOME WIN: Average between Home Wins % and Home Loses % -Draw: Average between Home Draws % and Away Draws % -Away Win: Average between Away Wins % and Home Loses % Then I compare this to the bookmaker odds (I use average odds from Betexplorer, or Oddsportal). Now the interesting part kicks in: I have 3 new variables, I will explain them here: -Diff HW: The difference between my HOME % and Bookmaker % -Diff D: The difference between my DRAW % and Bookmaker % -Diff AW: The difference between my AW% and Bookmaker % Now, some may see where this is going. As I thought in the beginning if your % estimate is higher than the bookmaker, than you there is value there and you should bet. But it's not. It doesn't work that way. My main assumption of this strategy is that all available information is included in the bookmakers odds and they can estimate far more better than my method (all in all, my probability estimates are based on simple averages). This strategy I like to think that is based on regression to the mean concept. For example, if a team has a low wining probability, as estimated by the method describe above (average between Home team home winning % and Away team away losing %), for example 30%, which implies an odd of 3.33 and a bookmaker says that Home Win prob is 45%, meaning odds of 2.22, than this bet it is a good BACK HOME WIN candidate for my strategy, or LAY AWAY. You will see for example that some BACK home wins candidates are teams that are in the middle half at the table playing a team with a better table position. Ok, moving on. The main criteria for the selection of my bets are the variables described above. For example: Diff HW: [-15%, -100%], the bets that fall between this interval are candidates for backing home wins Diff D: [-10%, -100%], the bets that fall between this interval are candidates for backing draws DIFF AW: [+8%-100%], lay away wins You see where this is going, when the difference is negative and within the interval, you back, and when positive, you LAY. You should keep in mind that when laying, the odds are higher than average odds (which are for backing), from my research, with about 13%, plus deudcting a profit tax to your winnings. I have tested 35 leagues, from past 3 years and I had mixed results. The thing that works for me is to tinker with every league until I get positive returns for 2-3 years in row using THE SAME thresholds(the differences that I talked above). With some leagues, I simply could not get consistent results. Therefore, I chose some leagues that worked for me. Another assumption of this strategy is that: If it worked the past 2-3 seasons using SAME thresholds, it is likely to work now. I only got good results in Backing HOME WINS and LAYING AWAY wins. Also, some good results in backing draws from few leagues, especially Italy Second League and French second League. Also, it helps to look at leagues that have a high percentage of home wins consistently over the years. Another critical element for the success of this tactic is odd selection. Normally for backing home wins, I will chose only odds above 1.7-1.8, and not higher than 2.4-2.5. For laying I wouldn't go above 4.5-5. Getting matched at better rates that the average odds, helps a lot. So does lower profit tax. As for the staking plan, I use a fixed % of a bankroll, which I haven't yet decided, but will be in the region of 2%-3%. I know that this is not a purely scientific method and it is somewhat flexible and depends on your tinkering with different leagues and particularities, but after a lot of tests I think it has potential. I attached you a spreadsheet model with a league that is among my picks for my lay selections, the Norwegian first division. You can find it here: http://www.speedyshare.com/25sEC/Norway1.xlsx Also, a screenshot of my biggest test for this system, a uniform selection, regardless of the league, just using the same threshold for every league. About 10k games, from which the lay away selections landed a 15.94% yield from 650 selections with odds ranging from 3.5-4.5. Starting bank was 500 units, each bet was 3%(both flat and compound simulation). I am looking forward for your feedback.

[froment]

Re: Sharing my system - looking for feedback Hi, Allen. Your strategy seems to me very similar to the one that BetVirus has on their site: http://betvirus.com/evals/ (click on + sign on the right to see the calculation) or http://betvirus.com/bvrates/ Actually, it is the same if both home team and away team have the same number of matches played (because you calculate percentages for each team, and then average them, and they first average matches and then calculate percentages), otherwise there will be a slight difference, so it could be considered basically the same system. My concern is, though, that systems based purely on statistics are likely to fail, sooner or later. I'm "statsman", and I rely mainly on statistics, but it let me down many times; there are too many other factors that influence actual result vs. one expected by the stats. If you obtain your selections statistically, then you are statistically doomed to face long winning or losing runs; the question is which one will come first, and how long will it last. If it's not viable to inolve "external" factors into calculation (such as missing players, fatigue due to cup or international matches or long distance between venues, weather, etc.), I'd recommend involving at least teams' form, to confirm or reject the bet - say, if your calc says "back home team", I'd check last 3 or 4 matches, and if home team has, say, 2 points and away team 7 points from those 4 mtaches, I wouldn't dare backing home team, and vice versa - if home team is in better recent form, go for it.

Another assumption of this strategy is that: If it worked the past 2-3 seasons using SAME thresholds' date=' it is likely to work now.[/quote'] Well, this is pretty lousy assumption. Of course, each league has its own pecularities that were established over time; new players are bound to fit into such a profile, so leagues' pecularities change slowly. But still, they change, and you cannot predict if it will change sufficiently to affect your results. Take as an example Norway 1, that you enclosed above. Over previous five season, percentage of away wins was fairly consistent, about 25%, except 2009, when it was 23%. But amount of draws and home wins oscilated for as much as 8%: Home wins 2013 / 2012 / 2011 / 2010 / 2009 / 2008 / 2007: 48.8% / 50.0% / 52.9% / 50.0% / 46.7% / 49.5% / 53.4% Draws 2013 / 2012 / 2011 / 2010 / 2009 / 2008 / 2007: 25.4% / 24.6% / 22.9% / 27.1% / 27.9% / 24.7% / 20.3% Away wins 2013 / 2012 / 2011 / 2010 / 2009 / 2008 / 2007: 25.8% / 25.4% / 24.2% / 22.9% / 25.4% / 25.8% / 25.8% So, the question is why away wins were low in 2010, when 3 seasons both before and after that were higher? At this moment though, season 2014 has away wins above the average, at 28.3%; that obviously is not good for both your back home teams and lay away teams. So, the question is, will the rest of the season remain the same, thus setting new high for away wins (and is your system robust enough to survive that increase), or will decline back to the average? Nice simple system, so if you feel it's good, just go on and try it! :ok