We have the first tranche of World Cup qualifiers of course, but I can't really get too excited by these and am frankly looking forward to when the proper football fires back into life again next week.
In the meantime, however, I thought I'd endeavour to ply you with some little snippets of information, just to make sure you're still paying attention.
As bettors and traders, we're always looking for little patterns or sequences that could possibly help us with our profitability, so I thought I would look at one of the most heavily-studied patterns to see if we could squeeze any advantage out of it. Every man and his dog looks at latest form to see if it can yield some insight into the next game's result, but is there any real point in doing so?
Some people have suggested that if a team loses a match two games ago and then draws their last game, then that points to an improvement in form and perhaps they are due for a win in the next game. By the same reckoning, if another team wins their match two games ago but only manages a draw in the last game, then this could indicate a waning of form, and maybe that team will lose their next game.
Not exactly rocket science, is it? But let's try and flesh these rudimentary ideas out nonetheless. Now at this point, there may be many of you saying to yourself, "Well this is just the triad stuff that he's already discussed and provided tables for" (HERE). That's true to a certain extent, but this is slightly different as I'm not taking home and away matches into consideration. I just want to see if there are any trends between the last two matches of each team. No other factors will be taken into account. Also, the tables I provided for the triad stuff was all stolen and, to be honest, I don't know what data was used to compile them. This time, I will be mining our information from all the freely-available data.
So anyway, the questions I asked myself were of the type: "Should a home team playing a match with a Draw-Loss record beat an away team playing with a Draw-Win record?" Time for a little spreadsheet, methinks!
Okay, so I analysed 6,600(ish) matches and paired each two-match sequence together so that I would get a string like "DLDW". The first two characters are for the home team (latest match first) and the second two characters are for the away team (again, latest match first). Then I assessed each possible sequence and determined the 1X2 percentages based on these 6,600 matches. For a bit of fun, I also decided to include the Under/Over 2.5 percentages as well. I wasn't really expecting to see any patterns of figures of significance, but thought I would take a look anyway.
Ideally I would like to have run some regression on these results, but due to the nature of the data, I didn't really see how I could. I'm sure the more analytical of those amongst you would be able to manipulate the data in a superior manner to me, but to my unprofessional eye this seems okay for now.
There are 81 possible combination of WDL sequences when joining the home and away team's results (nine pairs of nine), so the downloadable spreadsheet shows the 1X2 and Over/Under results for all these combinations across the sampled 6,600 matches. I've highlighted high percentages in green and low percentages in orange, and have done so based on the following (admittedly rather arbitrary) criteria:
- Home - green above 55%
- Home - orange below 36%
- Draw - green above 31%
- Draw - orange below 20%
- Away - green above 37%
- Away - orange below 20%
You may well decide that these thresholds that I've plucked out of thin air need tweaking, and if so then be my guest and tweak away.
Conclusions:
So, are there any decent headlines from all this? Should we be looking at specific sequences for specific results? Well, first-off, as a sanity-check, the 1X2 figures look about right, with 46.77% ending in a home win, 26.53% finishing as a draw, leaving 26.70% as an away win. There are also a few obvious sequences such as WWLL giving a 70.09% chance of a home win - but those of you expecting to see confirmation that a DL or DW sequence means an impending win and loss for those respective sequences, then you'll be disappointed to know this is not really the case.
On all the home DL sequences (that's DLLL, DLLD, DLLW, DLDL, DLDW, DLWL, DLWD and DLWW), only the DLLD sequence appears to provide an increased probability for a home win (58.18%). Some of the others favour the away side. And if you had hopes for DLDW (meaning that the home side were coming in to form while the away side were losing form) then you'll be surprised to know this showed more likelihood of an away win, with H=42.31%, D=19.23%, A=38.46%. I suppose it just shows how dangerous hunches can be (Gundulf take note!).
The away DL sequences didn't fare any better I'm afraid, with more advantage being shown to the home team than the away team.
Okay, so I've told you about the unexpected or the disappointing, but is there any good news in all this? Well, if you look at the spreadsheet attached here, I suppose you could say that all those percentages that have been coloured green or orange "could" be good news, but we need to take a closer look, and we'd need to check if the given probability is better than the odds provided in reality.
Results Checking:
Using the table I've provided, I separated-out only those green and orange sequences and then placed imaginary £10 bets if the following criteria were met:
- Bet on home if sequence is one of the green home win sequences, and draw and away are not one of the green sequences, and if the implied home win odds are better than the actual odds for the match.
- Bet on draw if sequence is one of the green draw sequences, and home and away are not one of the green sequences, and if the implied draw odds are better than the actual odds for the match.
- Bet on away if sequence is one of the green away win sequences, and home and draw are not one of the green sequences, and if the implied away win odds are better than the actual odds for the match.
Due to boredom over this exercise, I only ran this against three season's worth of results. These were:
- Premier League 2010/11
- Premier League 2011/12
- Championship 2011/12
These are the results from these three seasons:
Woo-hoo, we've made a profit. After placing £4,420 worth of bets, we're £324.90 in profit. That's an overall ROI of 7.35%. However, before we start getting too excited, there are one or two things to consider. Firstly, the ROI is not high and, when combined with a relatively smallish number of bets for each season, we're subject to the vagaries of variance. These figures could just as well have shown a loss. Also, three seasons-worth of results is by no means sufficient. If we wanted to be more satisfied that looking at sequences in this way is valid, we'd certainly need to back-test against many, many more seasons.
However, on a positive note, these are results purely based on the stats. No consideration is given to teams, the importance of the match, whether teams have just got back from Europe, yada yada yada. Also, as mentioned I plucked the threshold figures out of my head, so these are more variables that could be changed and tweaked Definite room for manoeuvre.
In reality, I don't know many people who would be prepared to risk their money on such a rudimentary method as these WDL sequences - but perhaps if used in conjunction with other methods that you use, there could be some benefit in considering past win/draw/loss sequences in this manner.
You will notice that, although I included Under/Over figures in the spreadsheet, I didn't bother back-testing those. The number of sequences was small, so it could be that the variance will be even greater. But either way, this could be a nice exercise for the reader to pursue.
I'll expect all your homework to be by this time next week.
Quite a bit of analysis there SoccerDude, i think i prefer Rick Ford's method of predicting the future and lumping on short priced games which he has a hunch for.
ReplyDeleteYou have misunderstood, Al. After completing this analysis and settling on a selection, you should then bet 120% of your bank :-)
ReplyDelete