I spent a lot of time at the start of October
writing a blog post discussing the Underlying results for September. Given
trends haven’t really changed too much, I won’t spend too much time on this
post. The additional info in this post that wasn’t there last month is for the
Euro bets, so I will touch on these results so far this season too.
In terms of the UK results, below is the table
which summarises the results so far.
October’s loss on Aways wasn’t as bad as
September’s loss but even so, a loss of 11% if backing all Aways is a difficult
place to start from for Away betting. Homes had a good month again with a 5.8%
ROI. Draws ended the month with a profit also of 3.5% which is a big
improvement on September’s results.
I would therefore expect my Home system bets to
have been profitable in October, I would expect my Aways to be loss making
(again!) and I would have hoped for a decent profit from Draws. Of course, my results aren’t 100% correlated
to the Underlying results but there is enough of a correlation to help
understand some of the trends we see on a monthly basis.
In terms of terms of the season to date, a
nightmare season so far for Aways with a 15.4% loss. This has been the toughest start to any
season for Aways since I started playing this game 5 seasons ago. It’s also one
of the toughest runs ever although it’s not the worst run by any means. We have
had periods of 4 or 5 months in a row where Aways have been loss making but I
guess the difference this season is that we are seeing it at the start of the
season which always seems to hurt a little more!
The best algorithm so far this season has been
algorithm 1 (system 6) and it is showing a profit of 2.9% which is a tremendous
result given the difficulties faced so far. Algorithms 2 & 3 are not far
away from break-even and again, given the difficulties so far, I think this is
a respectable effort.
I guess my big concern is algorithm 4. It was
my weakest algorithm coming into the season and after 2 months, I’ve not seen
too much to change my mind. I’ve said this before but the strength of an
algorithm is always seen during a tough period rather than an easy period as
all algorithms make profits during an easy period. A loss of 8.5% is not a
disaster given the underlying results and there is a big away bias on this
algorithm which is not helping things this season at all. Hopefully once things
start to get easier for Away betting, algorithm 4 can bounce back but at the
moment, it has questions to answer.
I think that’s a brief overview of the UK
results so far. Keep these results in mind when I review the UK systems.
Below are the results for the Euro games I’ve
rated so far:
The first thing that strikes me looking at this
is that I only wish the UK leagues had this pattern of results this season! :(
A big bias towards Aways, a small bias towards
Homes which means that Draws have been in short supply. Basically, it’s been
very easy to make profits backing Homes/Aways (Euro Systems) and it has been a
nightmare for anyone trying to back Draws (Euro Draw systems).
I spent a little time looking at the overall
results this season for Draws and not just the games I’ve rated and the results
aren’t too bad overall for Draws. There was a bit of a bias towards Draws in
the first 6 games played in each league and what we are seeing is something of
a correction I suspect. Unfortunately, my Euro Draw systems in their first live
season went live at the start of the correction and I’m already thinking
break-even for these systems would be an amazing result this season. I’d
probably take a small loss!
In a way, I’m really glad I wrote the blog post
on the Underlying Euro results before the season started as it really opened my
eyes to how difficult it can be making a profit from backing Draws in the Euro
leagues and after 5 weeks, I’m seeing this myself! The Home ROI is very steady,
so what we have seen is a number of Aways teams winning when they should really
be drawing and this has caused the Euro Draw systems to get off to a terrible
start.
The good news is that as my table shows, I’ll
rate over 2,000 games a season for the Euro leagues and therefore, we are only
just over 10% of the way through our season. The first weekend of November was
another nightmare for Euro Draw backers and therefore, the overall ROI for
backing all Draws in games I’ve rated is now -20%. We are now due a correction
the other way I hope!
One of the best methods to analyse a system is using statistical tests.
ReplyDeleteThis can be a daunting task if you are afraid of numbers!
Joseph Buchdahl, the owner of the much respected Football-Data, is a statistician and he has written an article called “Luck Versus Skill in Sports Betting: Analysing a Betting Record”.
You can find it here: http://www.football-data.co.uk/blog/luck_skill_sports_betting.php
In the article you will find this nice little piece of software: http://www.football-data.co.uk/blog/P-value_calculator.xlsx
Using this program is about as simple as placing a bet at best odds at a bookmaker!
If you would like to try it on real data here are the numbers required for two of the systems in this post.
System 6: Number of bets: 2823, yield: 0.0403, average odds: 3.28
System 41: Number of bets: 2021, yield: 0.0244, average odds: 3.34
For both systems determine the actual p-value and estimate how many bets are needed to get a p-value of 0.01 under the assumption of unchanged yield and odds.
Please comment in this thread if you have the answers and/or want to ask questions.
Dmitri
Hi Dmitri.
ReplyDeleteIf you drop me an email, I don’t mind adding the P value for all my system results. Just need you to tell me what formulae to use as over the years, I’ve had conflicting answers on the blog from Tage and others.
Cheers,
Graeme
Graeme,
DeleteYou can see the formulae in Joe’s p-calculator.
The test is a so called ‘One sided t-test’.
I think it would be a bad idea to add the p-values to your system results.
Very few members would know how to use the numbers, and if misused disaster awaits!
I wanted to see if there is any interest for statistical thinking among your subscribers.
If not it is a waste of my time to write more comments on such issues.
Dmitri
Hi Dmitri.
ReplyDeleteIt’s this sort of response which slightly frustrates me if I’m honest and I’ve been here with Tage before.
You’re trying to generate a discussion about statistical measures on the blog but the truth is, those that are really interested in this don’t need to read your comments about the fact that the systems aren’t statistically significant as they’ll know this themselves. Those that are really interested in stuff like this tend to know what they are doing I find.
I think people like myself, who aren’t a massive fan of statistics but always enjoy reading this stuff and can keep an open mind, are the subscribers who would appreciate your comments and input but as always, the comments from statisticians don’t really help us decide what to follow. It’s easy to point out that system 6 needs thousands more bets before becoming statistically significant but that’s not really the point. I don’t know any subscriber who follows system 6 as a chosen system to follow. Likewise, following every bet on system 41 in isolation probably isn’t a good idea to make money long-term but again, I don’t know anyone doing this.
I may not have agreed with everything Tage wrote but I think the best thing he did was give me and the subscribers the idea of the Max 4 system. Even then, he simply used it as a way to show how silly it is to stake more on bets appearing on multiple systems but simple people like me want systems and methods to follow and therefore, I adopted the Max 4 system as part of my portfolio this season with a variable staking plan linked to value. Again, I doubt this method is statistically significant and is probably flawed in some way but like I do every season, I will follow the bets, make a profit (touch wood!) and then do something completely different the season after. That’s part of the fun for me, following different methods and systems every season.
I know it drove Tage mad, it will drive you mad too but I simply want to make a profit each season from my chosen betting method/portfolio. That’s it. Not too fussed how I do it as long as I make a profit, don’t go broke and get a smooth P&L journey.
The fact that I have 41 live UK systems proofed and 39 of these are currently showing a profit is for me the best indicator of my future success. My target is the same every season and it’s that every system is profitable. I personally don’t care that none have tiny p-values as long as they make profits every season. Likewise, I suspect most subscribers only want the systems to make a profit every season. The fact that they are not statistically significant can’t bother them all too much. There are a whole host of football systems and tipsters in the market with p-values of 0.0000001 which make subscribers fortunes every year and charge £200 a season but I suspect these are probably full and explains why subscribers end up following my rubbish systems!
Most of the above is a bit tongue in cheek Dmitri but as I said in my opening line, it is borne out of frustration.
“Give a man a fish, and you feed him for a day; show him how to catch fish, and you feed him for a lifetime.”
I remember saying something similar during our blog comment exchanges last season but feel free to tell us how we should be using the systems to make more profits than we make at the moment with less risk please. You would definitely get people commenting on the blog I’m sure and we may even get a discussion going. I’d even post your comments on the blog itself and get away from the comments box!
Graeme
Graeme,
Delete“Give a man a fish, and you feed him for a day; show him how to catch fish, and you feed him for a lifetime.”
How true this is!
But I’m afraid you don’t understand it.
Martin’s guest post was great, probably the best I have read on this blog.
It ended with a “fish”. A system based on value.
I’m sure you loved this, but the really great thing about his post was the statistical reasoning he showed.
He showed you “how to catch fish”.
Tage likewise showed in his article to members how “to catch fish”.
And for your and other members benefit he gave you a fish.
I have written on fora for more than 10 years, and the most common response is: “We don’t understand what you are saying, and actually we don’t care. Give us a fish please!”
Dmitri
PS This is my last comment on your blog.
Hi Dmitri.
ReplyDeleteThanks again for the comment. All my discussions with statisticians seem to end in this way with you getting frustrated with my comments and then refusing to communicate with me again!
I think you misunderstood my reference to the fish quote. I agree with you that Martin’s blog post was great and I also think Tage’s Max4 system idea is great. As you say, both great fishes. :) I have literally seen hundreds of ways and ideas to play TFA over the last few seasons from subscribers but the vast majority are not backed up by p-values or statistical tests. Hence, this was my challenge to you.
I really look forward to someone showing us how we use statistical tests and p-values in order to make money from the bets. Clearly, if all the statisticians thought there was no edge with my ratings, they wouldn’t be communicating with me. I can’t imagine you all just sit and track systems and methods, waiting for a magical p-value to appear (might never appear!) and then you follow that system or that method. You surely must follow a system or method you believe that long-term, may have a significant edge.
If I take season one when I had no live bets, everyone following the bets made a profit as the systems all made big profits. Same scenario for season two I think. Clearly, following a system with 0 bets must send shivers down the spine of statisticians but that’s part of the game for me. Getting in first and making a profit before the edge is even known is better than waiting on the edge appearing and then everyone lumping on the same bets I have tended to find over the last few years. I change strategy every season for this very reason!
What I really want is someone to come to us (me and the subscribers) and say here’s a statistically significant way to make a profit from these ratings and systems. Every summer, I see various ideas and methods but they aren’t built from statisticians, they are built from people like me with a little Excel skills but no stats background and I’m sure half of the ideas are badly flawed and backfitted.
I will happily hand over my blog to anyone who wants to share how we should follow the bets in a better way. I did it with Tage via some PDFs I shared with the members, I did it with Martin via the blog. I offered everyone the same chance over the Summer to write a blog post and only Martin took the time and effort to do so which I appreciated.
I don’t expect you to reply to this, so no worries but if any other statisticians are reading this and want to educate others about how to make a profit, please get in touch. My only caveat is that it needs to say do this, rather than don’t do this! Telling people what they are doing wrong is easy, telling people how to do it right is much more difficult but that’s my challenge.
Form a queue please……….
Graeme
Dmitri,
ReplyDeleteI've read you're posts so far with much interest and you clearly now a lot about statistics (more then I do at least), but you can't blame Graeme for his systems not being 0.01 significant. As said, im no expert on statistics and my view is that you should use a mix of statistics and common sense to make decisions regarding gambling/investing.
When evaluating TFA I saw a tipster who has been profitable for 4 seasons on well over 2000 unique picks, who beats the closing line consistently. He might not be 0.01 significant, but when I use my common sense mixed in with some knowledge about variance/sample size, it's enough for me to invest.
When waiting for a system to be 0.01 significant it might take someone 10 years, after such a period he might lose his interest in tipping and quit, he could also lose his edge because of changing market conditions....
To give you an example from the poker world, I know guys who have been profitable for 5 million hands (thats way more then 0.01 significant), still they can't beat the current games....What im saying, the world is changing, not only the poker world is, but so is the sportsbetting world. There is just not enough time to reach sample sizes needed for <0.01 significance, so what you need is to use some common sense as well, at least that's what I try to do.
I hope you keep on posting here, because I like to hear you're input on things and just like you I'm not a big fan of the portfolio approach as mentioned in my blog post, but there are so many other ways to use the picks as has been debated before.
Martin