As I suggested last night, there are no midweek bets this week. That only leaves this weekend as the last set of bets this season and then that's the season completed for my ratings. I'm away when the last set of games in the Premiership are played the weekend of the 22nd and therefore, there will be no ratings produced for these games in advance of them taking place.
I'll post up any bets for this weekend on Thursday night.
Just a couple of quick things. I've had a comment from 'Your New Blog Reader' on a post a few days ago suggesting that the draw % in the leagues this season is only slightly higher than previous seasons and therefore, my numbers don't look entirely accurate.
There is a simple explanation (all my draw analysis is on my system bets and not on all the leagues) and my full reply is in the comment section but I think it does raise an important point about this blog. All of the building of the ratings and systems was done behind closed doors in the TFA forum over 12 months ago and therefore, it is difficult for blog readers to pick up the thread at times on here. In addition, all my early analysis this season when things were going great sits within the TFA forum and it was only at Christmas that I started to use the blog for any analysis. Until Christmas, the blog was purely a results tracking blog as that's the way I wanted it to be with no posts on any detailed analysis.
If anyone has any questions about anything I post, don't be afraid to ask me. I'd rather people ask me what's going on rather than just going on and not understanding what's going on.
I've also had a question from someone this morning looking for a high strike rate system that only looks at homes. I've mentioned this in the TFA forum plenty of times but it's fairly easy to create your own system using the other system bets. In addition, if people ask me to build a system to suit their particular needs, it's fairly easy for me to point out a system for them to follow.
I'll add this to my list of things to do this Summer but to give you a flavour of how easy it is......
I know the second algorithm finds much stronger home bets than the first algorithm. This can be seen by the fact that over the 5 seasons now, System 21 has an ROI of 25% on homes against an ROI of 12% on System 6.
Looking at this season only, System 21 has an ROI of 15.3% and system 6 has an ROI of 2.1% on Homes. Hence, it fits in with the backtesting although again, both systems have disappointed this season on homes (and aways!)
So, why can't we just back the homes on system 21? Well, we can do. Over the last 5 seasons, the strike rate has been 66%,64%,63%,68% and 60% this season. The profits have been 40pts,27pts,29pts,32pts and a measly 14pts this season.
This isn't doing anything fancy but isolating the home bets from the away bets on one system. It has produced 558 bets in 5 seasons and 93 bets this season in particular.
Following 100 bets in a season for strong home bets isn't everyone's cup of tea (definitely not mines!) but I can see the attraction of backing 1 or 2 teams a week on average with decent stakes at odds of around Evs.
As I keep saying on here and in the TFA forum, I'm sure that there is a system that suits everyone and it's important that people think about what sort of system suits them before next season. There is no guarantee that their system will be profitable next season (there is never a guarantee at this game that your edge remains from one season to the next) but I personally don't mind losing money following something I believe in and that's been my attitude since day one on the footie.
I hope this answers the question from the anonymous poster this morning. I think having ratings that can make a profit is one thing and being honest, it's not the hardest thing in the world to produce footie ratings to make a small profit over a large sample of bets. The hardest thing in the world is to use the ratings to maximise your profit and that's the bit I think I can do better than anyone else in this game hopefully!
Graeme - only one thing to add to what's already been said (the readers and commenters are of a higher standard than the average blog!) - you are clearly pricing the games yourself (from comments like Aldershot would have been 6/4, and they still would have been a bet at 6/4) - have you considered the Kelly Criterion? I'd suspect that 8% ROI (which I find extremely respectable at widely available prices) could be improved upon with Kelly staking.
ReplyDeleteGreat blog.
Hi Adam.
ReplyDeleteI'm not ashamed to admit I don't post all the comments on here and as many have found out, I don't post comments asking for links unless it's a good blog that is asking me! lol
It's funny, a few months ago, I was getting a lot of daft comments from people having a go at me for doing this and since my results have taken a massive dip, they've all but stopped apart from one or two idiots. I think I'm either winning people over or boring the pants off them mate!
If you read between the lines of the systems, there is in effect a staking method such as Kelly built into the systems. I don't intend to share this sort of info at the moment but as I've shown throughout this season, the higher up the system you go, the more value there is to be had. Hence, a bet on system 8 is a larger value bet than a bet on system 7 and system 6. Likewise, when you combine the two algorithms, a bet on system 8-22 is better than a bet on system 6-22 or 7-22 and so on.
Hence, by staking 1pt on each system bet that appears, you are in effect having more on the better bets and less on the weaker bets.
That's the sort of strategy I use although I don't bet on every team that appears on every system as that would involve staking 1pt on some teams and 13pts on other teams!
One missing ingredient then would be how much value each team needs to be to appear on each system 6-9,21-23 but again, I'll keep that to myself at the moment. The second ingredient then is when two teams appear on the same system, they clearly won't always be the same value. Hence, you could break it down further into more systems.
I think at the end of the day, the system approach works better than looking at each game individually and the value attached to it. I'm working with models that I think give me an edge but I'm dreaming if I think I can tell the difference between a team that may represent 10% value and a team that represents 20% value. Yeah, the ratings tell me that but I don't believe it!
I'm not sure if you are familiar with Fink Tank ratings but my ratings produce the same numbers. They give me a probability of a Home/Away win and the % left is the draw. I only highlight games to bet on where the % chance to win (Home/Away) is greater than the odds on offer, subject to boundaries of minimum odds and maximum odds.
Hence, I do get a % value for every game rated where value is available but as I say, I don't tend to base too much weight on this number. Instead, I like to think that every team that appears in a range of X and Y provides the same value and should sit on the same system.
If you think about the Kelly Criterion, what it would say is that you should stake more on the higher systems and less on the lower systems. This is in effect what I tell people to do or go one step further and only bet on the teams representing the highest value as you won't go wrong too often with that approach as long as you capture enough games to bet on.
I hope that makes sense.
I'm enjoying the interaction on the blog with the people who are commenting. I always try to answer each comment as fully as I can in the hope others will comment and enjoy reading the comments as much as the blog!
Cheers,
Graeme
Ran out of room for my PS.....
ReplyDeletePS. Aldershot were a one-off I think. It was all based on the fact Lincoln needed the win. I had Aldershot down as 45% likely to win that game with a 38% chance of a draw. Hence, I had Lincoln at 17% to win and they were a shade of odds on with the bookies. Usually when that happens, I go and check my sheet as there is a formula error or something! Port Vale weren't too far behind though and they did lose 1-0 to Barnet but Port Vale weren't in the same form as Aldershot and Barnet are better than Lincoln and Port Vale were a shorter price than Aldershot. Hence, it was easy to pick between Aldershot and Port Vale even though both appeared on system 9 as the highest value bets I find. Aldershot were the second highest value team I've found this season. The highest one lost though!
If you are going to have a one-off though, it would be last weekend, because of the soft factors occurring. I had a whole lot of big bets compared to the rest of the post-xmas period last month and last weekend, because of the overbet (IMO) motivational situations.
ReplyDeleteI take your point about the "Implied" kelly if you will but I was more thinking if just considering the strongest systems that have made 9%-12% ROI on level stakes stand-alone; I would expect Kelly to add 3-4% to that if not more.
When I look over my results for the last 6 years, if I mess around with overlays on my kelly formula, I do find that with the homes especially i am getting a 12% ROI when i put in a 10% overlay, a 17% ROI when I put in a 15% overlay, etc, and as such I am fairly confident that I am pricing home teams exceptionally well. Things are not so smooth on the aways but I don't price the draw to the effect that I could bet my draw "value" bets that are few and far between for anything other than breakeven, so something is still slightly wrong there. I think you are well on the road to being able to be confident that you can tell the difference between an evens shot that should be 4/5 and an evens shot that should be 4/6 or even 1/2, since they do occur (IMO).
The problem (or one of the problems) with 100 strong bets, with an expectation of even 20% ROI, is that your observed results could be maybe 2-40% ROI over the course of one season - there's definitely some safety in numbers and I prefer 5% ROI over 2000 bets to 20-25% over 200 bets, even if the marginal 1800 bets are by definition only running at around 4% ROI.
Thanks for the replies.
Hi Adam.
ReplyDeleteYeah, I take on board your points about Kelly staking. I personally prefer to play variables stakes on all the teams selected, with higher stakes on the so called higher value bets but nothing as sophisticated as Kelly!
I think we're from the same school of thought when it comes to the footie. I'd rather have a 5% return over 2000 bets rather than 20% over 200 bets. It's much more difficult to erode an edge of 5% across 2,000 bets than it is to erode 20% over 200 bets!
However, I think it's fair to say that not everyone has the inclination to play this way and many people do prefer the 'less is more approach' when it comes to the footie and at the end of the day, everyone has a different risk aversion at this game, so it probably suits their risk aversion better to play this way.
I had all of this in my mind when I started out on this project and that partly explains why there are so many systems. It would have been easy to stop at system 6 and system 21 and just highlight all the value bets thrown up by my models but IMO, that does me a massive disservice!
Anyway, I'm sure you'll have some more questions or observations as the summer progresses as I'm planning on using the blog a fair amount this summer when I make the next step on this journey.
Cheers,
Graeme
Graeme,
ReplyDeleteI will be watching and reading with interest. I do all my footy research work over the summer whilst betting the scandi leagues as well and any decent tournaments.
For readers who do prefer the less is more, they could do well to look at confidence intervals, and what they can mean over smaller samples - or look at football-elite's results over this season. No doubt in my mind that there is an edge there for F-E but the small number of bets means that one season can be +20% and the next can be 0, quite easily.
I totally agree with you again mate and when I wrote my last comment, I had FE in mind as it's a service I know pretty well.
ReplyDeleteWhen I first started looking at the footie, I set out with the intention to build something to rival Matt at FE. I was really keen on looking at strong home bets on the European Leagues and interestingly, I also looked at Scandinavia too as the same theory appeared to hold true in these leagues also!
Anyway, I did OK with my backtesting but I was finding the draw was a real pain in the French and Italian leagues and with the top English/Spanish clubs very strong at home anyway, my sample of bets was fairly small and therefore, I decided to move onto the English lower leagues and then brought in away games too!
I did pass on some of my workings to Iain who started a blog tracking how these bets would do in the European leagues. (Football Formbook blog) I know he struggles for time to write much on the blog due to work commitments but I think he too probably has an edge at what he's doing and he's did similar to Matt this season (or slightly better I suspect).
I'm a great fan of Matt at FE (he'll think I'm stalking him if he reads this blog as I said that on the SBC forum a few months back too!) and I think guys like Matt, Skeeve, Stewboss etc. are respectable guys who give the tipster game a good name and I've great respect for all these guys and the knowledge they have of the game and how to make a profit.
Hoping to get myself into that illustrious league one day! ;)
Graeme
PS. Do you have a list of bets in Excel from your blog you could ping me across to my email and I'll do some cross referencing with my bets and share the analysis with you (either privately or publicly!)
I've found when my two rating algorithms agree, the results are very good as I've found this season.
Hence, when I build my 3rd rating algorithm this summer, I expect that when my 3 sets of ratings agree on a value bet, it will produce even better returns although I'm limiting the sample size too much possibly but I'll not know until I do it this summer.
I'd be interested in looking at some external sources though and checking how the results compare when we agree on value bets.
Do you have a list of teams in Excel that you could share for a bit of an experiment?
Graeme,
ReplyDeleteI do - I do things in a similar way to you, in that I have several different ways of generating (profitable) selections - and indeed, i find when they agree independently of each other, they are more profitable.
However, my wife and I had a baby a couple of months back, and my records need SERIOUS updating, I have been very lax - but when I am in a position to do so, once the dust has settled on the season(s), I am willing to share a sheet of selections.
No probs mate.
ReplyDeleteI'm lucky in the sense that I do a lot of this in my spare time at work (after work, before work or at lunchtimes!) as my other half works in the same place as me, so I'm often waiting around on her!
If you do get some time this summer to update things, it would be good to cross check how our bets do when we agree. I'm always looking for another way to improve my returns! The moment you standstill at this game I think, your edge vanishes.
Congratulations on the new baby by the way.
Cheers
Graeme