Gilward Kukel Posted August 21, 2003 Share Posted August 21, 2003 Maybe I will try here some simple systems, but don't expect too much from them. ;) Here's the first one: System 1 Rule: Bet on home win if the following conditions are true: -) It is a game in the English Premier League. -) The home team has played at least 6 games. -) The home team has won at least 3 of the last 6 games. -) The average odds for the home win are >1.5 and <2. Backtesting (using average odds): Season Bets Strike rate Yield 2000-2001 56 66,07% 13,09% 2001-2002 35 74,29% 28,03% 2002-2003 51 72,55% 22,78% Total 142 70,42% 20,25% Quote Link to comment Share on other sites More sharing options...
Guest madmick Posted August 21, 2003 Share Posted August 21, 2003 simple systems Nice idea Gilward, have you tried it on other leagues? what kind of fall-off is there if the team has won less than 3 of the last 6, or the odds are slightly outside this range? Quote Link to comment Share on other sites More sharing options...
Gilward Kukel Posted August 21, 2003 Author Share Posted August 21, 2003 Re: simple systems Nice idea Gilward, Thank you. have you tried it on other leagues? The results for other leagues are not so good. For instance: Germany 1: 117 bets; Strike rate 58,12%; Yield -1,91% France 1: 167 bets; Strike rate 48,50%; Yield -16,50% Italy 1: 63 bets; Strike rate 50,79%; Yield -12,57% Spain 1: 136 bets; Strike rate 54,41%; Yield -6,29% The results for the other English leagues are bad as well. A bit better: Scotland Premier League: 57 bets; Strike rate 59,65%; Yield 5,89% what kind of fall-off is there if the team has won less than 3 of the last 6 A massive one! 177 bets; Strike rate 50,85%; Yield -10,46% or the odds are slightly outside this range? odds =2 and Quote Link to comment Share on other sites More sharing options...
Gilward Kukel Posted August 21, 2003 Author Share Posted August 21, 2003 Re: simple systems Here's another system: System 2 Rule: Bet on home win if the following conditions are true: -) It is a game in the French Division 1. -) The home team has played at least 6 games. -) The home team has won at most 2 of the last 6 games. -) The average odds for the home win are >2 and Season Bets Strike rate Yield 2000-2001 65 50,77% 15,15%        2001-2002 71 53,52% 20,37%        2002-2003 77 54,55% 24,73%        Total 213 53,05% 20,35% Quote Link to comment Share on other sites More sharing options...
Gilward Kukel Posted August 21, 2003 Author Share Posted August 21, 2003 Re: simple systems Yet another system: System 3 Rule: Bet on home win if the following conditions are true: -) It is a game in the German Bundesliga 1. -) The home team has played at least 1 game. -) The home team did not win the last game. -) The average odds for the home win are >=2.3 and Season Bets Strike rate Yield 2000-2001 33 54,55% 35,18% 2001-2002 29 44,83% 11,69% 2002-2003 46 52,17% 24,87% Total 108 50,93% 24,48% Quote Link to comment Share on other sites More sharing options...
Joe Posted August 21, 2003 Share Posted August 21, 2003 Re: simple systems Sorry to rain on your parade, but this all looks very much like the effects of data mining to me, and I suspect that the high yields for specific sets of matches as identified by the mining process are the result of extraraneous influences as yet unaccounted for, which will result in future prediction failure. In my forthcoming book (still with the publishers Mick, not heard from them for a couple of weeks), I discuss these effects with reference to a draws predictor I've mentioned on the PL before (strike rate 32.7%). I still have misgivings about this system and doubt I will put it into practice like I thought I might earlier this year. Good luck anyway, G.K. Quote Link to comment Share on other sites More sharing options...
Gilward Kukel Posted August 21, 2003 Author Share Posted August 21, 2003 Re: simple systems Yes, I know it's not safe to believe that these systems will work forever. Quote Link to comment Share on other sites More sharing options...
Osesame Posted August 21, 2003 Share Posted August 21, 2003 Re: simple systems There was me believing that data mining is a useful tool! The info used to arrive at these results is a concensus of many bookmakers . The hypothesis seems to be that there may be weak areas in their strategy. I note with interest that this ties in with several mentions of "fingerprints" for different leagues. IMO no system can work for ever unless it is constantly updated in similar fashion to DIMB perhaps. Quote Link to comment Share on other sites More sharing options...
Guest Weststander Posted August 21, 2003 Share Posted August 21, 2003 Re: simple systems Looks a bit like "backfitting" writ large,especially as it doesn't work for alot of other leagues. The most likely explaination is that the system's hit upon a profitable run of results thru nothing more than short term luck. W. Quote Link to comment Share on other sites More sharing options...
Osesame Posted August 21, 2003 Share Posted August 21, 2003 Re: simple systems Weststander Why would you expect it to work for all leagues? Quote Link to comment Share on other sites More sharing options...
Guest Weststander Posted August 21, 2003 Share Posted August 21, 2003 Re: simple systems Hi Osesame, to turn the burden of proof around,why should it not work for the 1st div but work for the Prem? If you carn't find a credible reason why the 1st is fundamentally different from the Prem(goals scored/conceded by sides,distribution of draws etc) then the most likely explaination is hitting a luck seam whilst data mining. You can be slightly more confident if AFTER you've identified a seemingly profitable strategy,then that strategy perssts.But you carn't really use your initial findings as part of the proof. The systems haven't been tested with an out of sample group & they don't work for the same sport played by teams of very similar differences & characteristics. The only conclusion imo is that they're suspect at best. W. Quote Link to comment Share on other sites More sharing options...
Osesame Posted August 21, 2003 Share Posted August 21, 2003 Re: simple systems Why do you think div1 and premier leagues are so similar? Quote Link to comment Share on other sites More sharing options...
Osesame Posted August 21, 2003 Share Posted August 21, 2003 Re: simple systems The credible reason is that the bookies price matches differently in these 2 leagues,and as this is an odds based system it is unlikely that the same odds range will give positive results. It may well be that a different set of criteria could be found for each division. Quote Link to comment Share on other sites More sharing options...
Guest Weststander Posted August 21, 2003 Share Posted August 21, 2003 Re: simple systems Interesting stuff,Osesame. In which way do the bookies price the Prem up differently to Div 1.Specifically which price bands in either league differ greatest from the actual outcomes. What's the cause of this descrepancy. Home wins,away wins,draws,0-0 draws,average goals by home team,average goals by away teams,number of players,rules etc seem pretty consistent across the two leagues. Even when you can come up with major inter league differences,the odds usually reflect this(eg draws in France). Still contend that you carn't use your initial data pool to "prove" the case. W. Quote Link to comment Share on other sites More sharing options...
Osesame Posted August 22, 2003 Share Posted August 22, 2003 Re: simple systems So you are saying in effect that back testing is pointless Quote Link to comment Share on other sites More sharing options...
Osesame Posted August 22, 2003 Share Posted August 22, 2003 Re: simple systems 2002/2003 average odds and range premier league/div1 home 2.25/2.11.....1.18-6.27/1.34-4.11 draw 3.36/3.26......3.05-4.53/2.99-4.17 away 3.88/3.52.....1.48-11.37/1.76-7.58 Quote Link to comment Share on other sites More sharing options...
Guest madmick Posted August 22, 2003 Share Posted August 22, 2003 backtesting I don't think that Westander is saying that backtesting is pointless, rather that when you back test, you should reserve some data for 'cross-validation'. Westander is well able to answer for himself though, and no doubt he will have an elegant response in due course ;) It is the 'cornflakes and car-crashes' scenario. If you took a sample of 20 car-crash victims, you might find that 90% of them had had corn-flakes for breakfast and conclude that corn-flake consumption was a useful predictor of accident occurrence. To adequately test your theory, you'd want to keep some data to compare against that which you used to develop your theory. No matter how big a sample is, it is nearly always possible to find some commanilty between the data and to establish a correlation with some outcome. If that correlation doesn't hold with other data drawn from the same population, then there is likely a problem. A simple way to test these theories (to a limited extent) would be to split the original samples in two, at random, test on each half. If the samples are of reasonable size and the theory is a good 'un, then the results of each 'half-sample' should be largely similar. If they aren't then what you are getting is probably a 'luck seam' as Westander put it so well Quote Link to comment Share on other sites More sharing options...
Osesame Posted August 22, 2003 Share Posted August 22, 2003 Re: backtesting Well it is already split into 3 seasons is that not enough? Quote Link to comment Share on other sites More sharing options...
Guest Weststander Posted August 22, 2003 Share Posted August 22, 2003 Re: backtesting just a quick response, I'm not saying that backfitting is pointless,more that it's fraught with potential pitfalls. Say your marvellous,profitable sys produces horse racing selections that make a handsome profit. Say your selections,by chance include 4 horses that are being given performance inhancing substances(obviously unknown to you) & these selections are responsible for the bulk of the profits.Obviously this is hypothetical & bears no relation to our extremely honest racing game. You betting with real money coincides with the Jockey Club having a quite word. Profits plummet,never to return. In short the chances that any sys includes every factor that realises profits is small(especially if it only contains a few rules) & if you don't know all the factors that contribute you won't know when your selection doesn't meet that particular factor.And that might be the one that makes all the difference. Just trying to be realistic. W. Quote Link to comment Share on other sites More sharing options...
Osesame Posted August 22, 2003 Share Posted August 22, 2003 Re: backtesting I understand all the points put forward,and the rules used to make selections are very simple,but before these selection criteria are used, a huge amount of selection has already been made by around 60 different bookies to arrive at the average odds values. I have also been studying patterns that occur using average odds,for particular leagues and even without any rules added there have been some interesting findings. It should also be noted that the yields quoted are at average odds and therefore by using best odds they would be increased by approx. 10% . Quote Link to comment Share on other sites More sharing options...
Gilward Kukel Posted August 23, 2003 Author Share Posted August 23, 2003 Bets Bet for System 3 Eintracht Frankfurt - Hertha Berlin SC : 1 @ 2.90 (Interwetten) Quote Link to comment Share on other sites More sharing options...
Gilward Kukel Posted August 24, 2003 Author Share Posted August 24, 2003 Re: Bets Eintracht Frankfurt - Hertha Berlin SC : 1 @ 2.90 (Interwetten); 0-0 :( Quote Link to comment Share on other sites More sharing options...
Gilward Kukel Posted September 20, 2003 Author Share Posted September 20, 2003 Re: Bets Bets for System 2 Ajaccio - Stade Rennes: 1 @ 2.30 (Admiral) En Avant Guingamp - Paris Saint Germain: 1 @ 2.70 (Worldbet) Toulouse FC - Metz FC: 1 @ 2.20 (Bets4all) Bets for System 3 Freiburg SC - Schalke 04: 1 @ 2.60 (Expekt) 1.FC Cologne - Vfl Wolfsburg: 1 @ 2.60 (SportsTAB) Quote Link to comment Share on other sites More sharing options...
Gilward Kukel Posted September 21, 2003 Author Share Posted September 21, 2003 Re: Bets Bets for System 2 Ajaccio - Stade Rennes: 1 @ 2.30 (Admiral); 1-0 :) En Avant Guingamp - Paris Saint Germain: 1 @ 2.70 (Worldbet); 0-2 :( Toulouse FC - Metz FC: 1 @ 2.20 (Bets4all); 1-0 :) Bets for System 3 Freiburg SC - Schalke 04: 1 @ 2.60 (Expekt); 2-1 :) 1.FC Cologne - Vfl Wolfsburg: 1 @ 2.60 (SportsTAB); 2-3 :( System 2: 3 bets 2 won +1.50 Profit +50% Yield System 3: 3 bets 1 won -0.40 Profit -13.33% Yield Quote Link to comment Share on other sites More sharing options...
Gilward Kukel Posted September 27, 2003 Author Share Posted September 27, 2003 Re: Bets Bets for System 1 Arsenal - Newcastle United: 1 @ 1.60 (Jokerbets) Southampton FC - Middlesbrough: 1 @ 1.85 (Unibet) Manchester City - Tottenham Hotspur: 1 @ 1.85 (Interwetten) Bet for System 3 1860 Munich TSV - VfB Stuttgart: 1 @ 2.70 (Expekt) Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.