Friday, 22 August 2014

UK Systems - Draw Indicator

After posting the last blog post, I remembered there was one other piece of analysis I wanted to carry out this Summer. It is something I said I would have a look at after the season was done but being honest, with everything else on the go, it slipped my mind.

What I wanted to look at was the performance of the UK bets and comparing the performance of when they do/don’t appear on a draw algorithm. Of course, the limitation with this analysis is that I have 4 seasons of live results on some UK systems and only one season of live results for the UK Draw systems, so I don’t think we can draw too many conclusions. Having looked at the data, I think it is very confusing and therefore, it’s something we need to park until next Summer I think.

Before looking at the data, I guess I expected to see that results would improve for my systems if the same game didn’t appear on a Draw system. If a game is on a Draw system, it isn’t going to be as profitable as bets that don’t appear on Draw systems. That was my hypothesis but then again, given how profitable the Draw systems looked before they went live, it sort of follows that this would be the case. If the Draw systems had a 20% ROI, it must be the case that the results of the outright systems would be better if the Draw systems didn’t find the same bets!

Anyway, having done the analysis, I think last season is very confusing and doesn’t fit in with the historical results (which to be fair, isn’t totally surprising).  I think once we get a few seasons Draw data behind us, we can draw more meaningful conclusions.

Here are the results anyway:

For the first 3 seasons, it is clear that if the Draw systems don’t find the bets, it is better for the systems results. No surprise given the Draw systems had such high returns. However, last season, the opposite is true. An ROI of 6.2% overall where the bets appear on Draw systems and an ROI of only 1.5% where the bets don’t appear on Draw systems. Hence, rather than reducing stakes where the Draw systems found the same game, we should have been increasing stakes!

I think we need to wait until we get more Draw data and then we can maybe draw more meaningful conclusions.

I’m done with analysis for others.  Next task for me is to decide what I do myself next season! So many ideas, so many ways to play and only 2 weeks left until the first bets!.

Next post will hopefully be a guest blog post from a TFA subscriber. Keep an eye out for it……

Euro Underlying Results

One of the outstanding pieces of work I had this Summer was to pull together a tracking sheet to monitor the underlying results for the Euro leagues. I’d sort of been doing it in the background but it was only earlier this week that I got on top of the email exchanges with subscribers and given the fact my baby hasn’t arrived yet, I thought it was a good opportunity to complete this piece of work and do a quick blog post on it.

As you will remember, it was only last season that I started tracking the underlying results of the UK leagues closely. I’ve always been acutely aware of the fact that my system results are heavily influenced by what happens to the underlying results but I didn’t ever feel the need to pay too much attention as at the end of the day, I can’t change the underlying results. However, what I found last season was that without knowing exactly how the underlying results were behaving in the short-term, it was impossible to explain some of the variance I was seeing in the results on a monthly basis.  I set up a little tracking sheet to monitor the results on a monthly basis and it really helped me with the mental side when I was experiencing some of the ups and downs.

So, I’ve now completed the same piece of work for the Euro leagues.  

Being honest, I’d looked at something like this before in early 2010 before I built the first UK algorithms as back in the day, my intention was to build Euro ratings first rather than UK ratings. However, my first few attempts at building Euro ratings weren’t too successful as I couldn’t find a way to create an edge on Aways (looking at Draws hadn’t even been a consideration) and when I looked at the underlying results, I could understand why. Basically, from 2006/07 to 2009/10, backing all Homes blindly was pretty near break-even in these top 4 Euro leagues (Italy, France, Spain & Germany). Backing Aways was heavily loss making and therefore, trying to find an edge with Aways was very tough.

Anyway, roll forward 4 seasons and now I have two seasons proofed results behind me in the Euro leagues for Homes/Aways (not the greatest results but a foundation to build from) and next season will see the first season of Euro Draw bets. Therefore, it seems like it was worth investing the necessarily time to create something similar to what I have for the UK bets.  I have data going back to 2006/07 but similar to the UK underlying results, I’m going to concentrate on the last 4 seasons. The further back you go, the less competitive the prices are and therefore, the results look a bit different to the most recent years.

So, here’s the results for the last 4 seasons:

A really interesting picture and considering I’m about to embark on new Euro Draw systems, let’s just say my confidence is a little lower having seen these underlying results!

I think the first comment is the fact that if following Homes only, you should struggle to make a loss in these Euro leagues long-term based on these results. Many of my blog readers will be aware of other tipster services that specialise in Homes in Euro leagues and looking at the Underlying results, I can see why they have gone down this approach.  

Interestingly, in my first season of Euro ratings, my Home bets were terrible and therefore, I think this highlights how weak the ratings were I built. I knew that of course and that’s why I binned the ratings and started again with new Euro ratings last season. Also probably explains why I ended up with such a high % of Home bets on my rating algorithms now compared to what I had before. If you are relying on Aways in these leagues to make a profit, you will struggle I think long-term. 

Looking at Aways, it is a tough place to make money in these leagues and we’re looking at a loss of around 3.9% over the last 4 seasons.

However, compared to the Draws, a loss of 3.9% doesn’t look too bad! The Draws have lost 6.4% over the last 4 seasons, last season lost 8.5% and the 2010/11 season lost 10.5%. Not exactly an easy place to make money!

If I look at the month by month view, we see lots of months where backing all Homes is profitable and considering this is the first time I’ve really looked at this, I find it very interesting.

20 winning months from 37 if backing all Home bets. Amazing really but then again, given the overall results, shouldn’t be too surprised but it shows how consistent this method is.

12 winning months from 37 for Draw bets shows the difficulties playing in this space. Only 2 winning months in the last 12 shows how difficult it has been recently in these leagues if backing Draws.

16 winning months from 37 for Away bets and I was surprised it was as high as this given the issues Aways have generally.

Similar to what I do with the UK underlying results, I’ll be keeping tabs on these Euro results every month and when writing monthly reviews, I’ll try to use what I have seen to hopefully understand some of the performance issues we’re seeing on a monthly basis.

So, does the above change my thinking at all about the Euro Systems and Euro Draw Systems?

In terms of the Euro systems, I think last season was a pretty easy season for backing Homes/Aways and therefore, maybe I’m not too surprised that my Euro systems made a decent profit. I guess if I go back another season, it wasn’t that much more difficult and my ratings bombed badly, so I think that highlights deficiencies in my ratings for the first season which is what I concluded myself based on the poor performance.  I said at the end of last season that I still thought my Euro systems had a lot to prove and I think the above backs up my thinking. Let’s see how they do this season.

In terms of the Euro Draw systems, I think the above has opened my eyes to how tough this place is to make money. Being honest, having seen Cassini’s results over the last two seasons, I realised it wasn’t exactly an easy place to make a profit but I’m not sure I’ve seen too many tougher groups of data to try to find a betting angle over the last few years.

Last season, backing Draws in UK leagues was very tough with an ROI of -5%. This was the worst ROI since season 2007/08 for these leagues (also the lowest strike rate since then) and therefore, I was overly excited about how my Draw systems had outperformed the underlying return. For 3 of the last 4 seasons, the Euro Leagues have had a worse ROI than this!

Unless conditions get a bit easier for Draw backers (higher strike rate), I would struggle to believe that my ratings could make a profit next season. It’s not impossible (my UK Draw systems did it last season) but experience tells me that when the underlying results are tough, making a profit from a very high number of bets is just as tough. The odd system with lower turnover can do well but the higher turnover systems may struggle if the losses are as bad as 3 of the last 4 seasons. 

Of course, as I have pointed out before, tracking underlying results is great for explaining what has happened historically but in terms of explaining what may happen next, they are about as useful as a chocolate teapot. Some will say the Euro leagues are due more draws after the last few seasons whereas other will say that the trend has changed and the market hasn’t caught up and therefore, there is no value in draw bets in these leagues unless the strike rate increases.  Let’s see what happens this season!

Monday, 18 August 2014

FTL Systems - Season 2014/15

Well, the time has come to take the wraps off my FTL systems for the upcoming season.  I’m still three weeks away from my first bets as I wait until 6 games are played in each league before getting involved but now seems a good time to start outlining my assault on Cassini’s tipster league this season.

Firstly, before I discuss the plans for the upcoming season, it’s probably worth having a quick recap of the system I entered into the FTL last season and how it performed.  The official (prices on a Thursday evening) results of the system last season were a profit of 3.4pts from 401 bets for an ROI of 0.85%. A disappointing performance and ranking it in terms of my TFA systems last season, this unique system would have finished 36th out of 53 systems last season. Clearly, this was slightly frustrating as I obviously hoped that I would end up with a system in the FTL which was better than most of my other systems but this wasn’t to be.

The biggest issue I found last season was the fact that the odds used for P&L purposes in the FTL are Pinnacle’s odds on a Friday afternoon. I’ve done various bits of analysis myself on an ad-hoc basis and I’ve seen a very good analysis from a subscriber this Summer comparing Pinnacle prices over time for my bets and as I expected, Friday afternoon is a low point when it comes to the odds of my selections.  The odds start to rebound a little from Friday and by the time kick-off approaches, there are much better odds available than on a Friday. However, the best odds available are at the time I release the bets as you would expect.

Even though I understood the potential issues with the odds in the FTL, I didn’t really think the impact would be as big as it was last season. The official results in the FTL showed a loss of 10.72pts from 401 bets for an ROI of -2.67%. A 14.12pts swing across 129 winning bets. Given my average odds were 3.11, losing over 0.1pts per winner is a big swing (c. 5%) and ultimately, for football systems such are mine which are price sensitive, I think it shows the barriers my FTL system will have to overcome to do well in the FTL.

So, last season was a bit of a nightmare and at the end of the day, it cost me £125 in bounty prizes for other contestants who finished in front of my FTL system.  Given I’ve put up another bounty of £25 for everyone who finishes in front of me this season again, I’ll be trying my best to do better this season!

FTL Bounty System

One of my regrets from last season was not entering a TFA system into the league! I thought it was a good opportunity to try out a different system and test something live but it flopped badly. Therefore, lesson learnt and this season, my FTL bounty system will be one of my 61 live TFA systems this season.

Given what I learnt last season about the odds movements, I think I have to steer clear of my main systems this season which most people are going to be following. I would love to put a system such as system 31 into the FTL but the odds movements would mean I would lose a helluva lot of the edge this system has at Friday afternoon prices. Over the last 3 live seasons, system 31 has achieved an average profit of 52.5pts per season.  With Friday afternoon prices, I expect this profit would be circa. 30pts lower. This is a massive swing and therefore, my concern is that if system 31 has an average season (might be due one!), this system will make a loss in the FTL.

I think many of my subscribers probably assumed system 31 would be the FTL system this season but I’m worried I jinx the system! If system 31 has a poor season this season, most of us will be licking our wounds and therefore, I’d rather not take the risk. :)

So, my 2014/15 FTL bounty system will be …….. System D1-D6.

Yip, after only one live season of results, I’m going with a UK Draw system as my FTL system this season.

It was a straightforward decision if I’m honest. I know that Pinnacle offer the most competitive draw prices and I know from last season that my Draw systems don’t move the market much at all (maybe a little bit this season as some more people will be following them) and therefore, a Draw system gives me the best chance of competing in the FTL.

In terms of deciding on what UK Draw system to follow, here are the results from the Draw systems last season:

Well, quite quickly, we can rule out systems D1, D3, D6, D3-D6 and D3-D7. All 5 systems didn’t do that great last season and I’d not have much confidence in any of these systems next season.

Looking at the systems that were left, it’s not easy to choose the ‘right’ system. I think the thing that made me choose D1-D6 was simply the turnover. The system had 390 bets last season and I think about 400 bets is a good number of bets for a FTL system.  Last season, a profit of 44pts was achieved and if we look at the FTL over the past 3 seasons, this system would have walked the league each season if a profit of 44pts was achieved. Importantly, I did a quick check of this system based on the same odds as the FTL use (shows I’m trying my best this season!) and this system achieved around 43pts profit last season. No real impact from the prices quoted on a Friday.

The aim is simple with this system. 400 bets and an ROI of 10%. All it has to do it repeat the performance from last season and it will surely finish high up in the FTL this season. 

Before we all get too excited, the backtested results of the FTL system I used last season were better than this and the system flopped badly when it went live. I guess the difference for this season is I have a season’s proofed results underpinning this system and therefore, I have to be fairly hopeful I have a system which can win the FTL. That’s the aim!

FTL Other System

I think the other thing I learnt last season was not to put all my eggs in one basket! Given there is a really good prize on offer in the FTL, I think it is worth having a back-up system in the league.  To be clear, this back-up system has no bounty attached to it and therefore, it is going to be a very high risk system. Similar to my other system in the FTL, it will be a live TFA system but this system has no live results behind it and season 2014/15 will be the first live bets for this system.

Well, anyone with the service will know which type of bets this system will have. Given what I discussed above about odds movements, it’s no real surprise that I’ve gone down the Draw system route again.

Season 2014/15 sees my first foray into the European Leagues with Draw systems and therefore, this will be direct competition to Cassini’s own Draw system for these Euro leagues.  I’m sure Cassini doesn’t mind a little bit of competition. :)

Given I have no live results to go on, picking which system to follow from the European Draws is a bit of a punt.

Similar to what I concluded with the UK Draw systems, I think I’m forced to go for system DE1-DE6. The system is expected to have around 300-350 bets a season and I guess my hope would be that it can achieve an ROI of around 10%. Given I don’t expect any odds movements with these Draw bets (very few subscribers will be following these new bets), I don’t need to worry about the odds quoting impacting the returns too much.

Given my Euro systems flopped in their first season, I wouldn’t be totally surprised to see the Euro Draw systems flop next season. Of course, I hope this isn’t the case and the backtested results are very positive but I’m always wary of the Euro leagues after what happened to my Euro bets in their first season. What I would say is I managed to rebuild the Euro ratings based on what I learnt in the first season and the second season was much more profitable, so I’m hopeful I have a better handle on the Euro leagues than I had when the Euro systems first went live. We shall see.

So, to summarise, the TFA FTL Bounty system next season will be system D1-D6. The other FTL system will be DE1-DE6.

Let battle commence……..

Portfolio Probabilities

One of the most common ways to play TFA each season has been to follow a number of systems in a portfolio. Until last season, it had been a fairly simple way to make a profit each season and given the large % of systems profitable each season, it was quite difficult to make a loss over the first 3 seasons if using this approach.

Anyone who was reading Steve’s Daily 25 blog will have seen the issues Steve had last season with following a portfolio approach and ultimately, Steve followed 13 systems last season and somehow managed to hit 7 of his 13 systems to be loss making. Given that 35 of the 52 TFA systems live last season were profitable (67%), the odds were obviously in Steve’s favour in terms of randomly selecting a portfolio of systems to follow and being able to make a profit. 

I thought it was worth a very quick blog post to show the probability of making a profit for each season based on the number of systems in your portfolio. Of course, the key thing with this analysis is that the portfolio selection is 100% random and I’m not aware of any TFA subscribers who randomly select a portfolio of systems to follow each Summer. Given these results below, I think it does show that if you follow a small number of systems in a portfolio, you could do a lot worse than randomly selecting the portfolio. Based on the losses that some subscribers suffered last season from following correlated systems in a portfolio, a random selection approach is a much better method than following correlated systems!

2010/11 Season

There were 11 live systems in the first season. All 11 systems were profitable with an average profit of 23.9pts.

Here’s table showing your expected profit and expected probability of your portfolio making a profit based on the number of systems you followed:

Hardly anything too surprising here. No matter how many systems you followed, you were guaranteed to make a profit given every system was profitable.

2011/12 Season

There were 22 live systems and 21 systems made a profit (system 33 made a loss). An average profit of 21.8pts was achieved across all 22 systems.

Here’s the same table as above:

If following 1 system, you had a 95.3% chance of making a profit. If anyone followed system 33 in isolation during this season, they deserve to make a loss! Basically, by the time you get up to following 2 or more systems, you were pretty much guaranteed to make a profit on the season.

2012/13 Season

There were 41 live systems and 8 systems were loss making. All UK systems were profitable but the Euro systems were very poor in their first season. An average profit of 19.6pts was achieved across all 41 systems.

Here’s the same table again for this season:

Starts to get more interesting here and if you only followed one system, you had an 80.5% chance of hitting a profitable system. You needed to be following at least 4 systems to achieve a 90%+ probability of your portfolio being profitable.

2013/14 Season

There were 52 live systems last season and 35 of them made a profit. An average profit of 11pts was achieved across all 52 systems.

Here’s the table again:

If following 1 system, you had a 67% of hitting a profitable system. Even if following 3 systems, you only had an 86.7% chance of hitting a profit.

If we take the example of Steve (Daily 25 blog) from last season, Steve followed 13 systems and based on the numbers above, the chance of Steve making a loss following 13 RANDOM systems last season was less than 1%. However, as I mentioned at the start, Steve’s portfolio was anything but random and was actually heavily correlated. Hence, when one of Steve’s bets lost, it ended up pulling down lots of systems and therefore, his portfolio wasn’t random and therefore, the chances of making a loss were much bigger than outlined in my analysis.

I guess the other trend to pick up is that as I have grown the number of systems season on season, the number of losing systems in a season has grown and therefore, if you were randomly selecting one system to follow, you have more and more chance of hitting a losing season. I saw that last season as I know someone who has followed 7-22 since day one and after 3 profitable seasons, they hit a losing season last season.

However, if you have followed 5 random systems each season since inception, you would have had a 90%+ chance of making a profit each season.

What will 2014/15 season hold for us using a portfolio approach?  Ideally, I’d like to see all 61 systems make a profit next season but then again, I know it’s highly unlikely for this to happen but we’ll see what happens.

Here’s a summary of all 4 seasons below in one table.  Will be interesting to see how season 5 compares next Summer……