Thursday 12 April 2012

Early look at this season v last season

There was a great comment on my recent results post from Andrew. The essence on the comment was to do with the fact that system 7-22 has had a lower number of bets this season compared to previous seasons and therefore, if looking at the betting bank suggested for 7-22 on the blog, then the ROC probably isn’t as high as one might have hoped. I’m probably being a little harsh on Andrew as his comment didn’t really say any of this but it’s given me the idea for this post, so I’ve sort of interpreted what he’s said and twisted it. I’m sure he won’t be offended! :)

OK, there are a few things that need to be discussed here.

Firstly, the betting bank’s suggested on the blog. When I started out with this blog, my main worry was that it would attract a certain type of reader who would only be interested in following the bets without actually paying too much attention to what I write and therefore, they’d need easy instructions when it comes to betting banks. I therefore made a conscious decision that all betting banks discussed would be hugely inflated on the blog as this way, it ensured that if people chose to follow blindly, if things went wrong, they’d not do too much damage since they’d be understaking.

Of course, some blog readers became subscribers to the service which in itself led to me discussing betting banks etc. with subscribers and therefore, we have a deviance from what one set of followers may be using compared to another when it comes to betting banks. No real issue as I want subscribers to maximise profits and I want blog readers to see how much profit they can make if they follow next season!

My own view on betting banks is probably in line with others who supply selections but deciding on a betting bank is down to the individual following. Yes, I’ll give advice but we all have different risk preferences and risk aversions at this game. Not only that, we all have different styles of gambling and therefore, one size will never fit all.

Last season, I basically turned over my bank every 2 weeks in terms of pts played. Hence, I had 100pts in my bank and I was looking to stake 100pts every 2 weeks. I was trying to leverage my systems to the absolute max. I’ve been much more conservative this season and being honest, I’ve staked much less than I did last season but I’ve a much higher ROI this season and due to the additional turnover on some of the systems this season, a better ROC than last season too.

The point of what I’m getting at here is that if people blindly follow my advice on the blog, they probably aren’t going to maximise their ROC. My target on the blog is not to bankrupt anyone first and foremost. Making a good ROC wasn’t even a thought that crossed my mind when discussing suggested betting banks!

Let’s look at System 7-22. Here’s the latest P&L graph and drawdown graph for this system.



As you can see, the biggest drawdown remains 16pts. Hence, I said 2.5 times this would be pretty safe for a bank. Hence, 40pts. Do I actually know anyone who’s using a 40pt bank? Well, no. Basically, the biggest drawdown this season has been 5pts. Looking back, after the 16pt drawdown, the next biggest drawdowns have been 10pts (twice). Hence, the average of the 3 biggest drawdowns in history is only 12pts. Hence, a 40pt bank suddenly looks pretty high doesn’t it?

Of course, using drawdowns is only one way of trying to devise a betting bank. You can do mathematical calculations using strike rate and average winning odds but at the end of the day, there is a flaw in all of this isn’t there? Clearly, as soon as the systems are tweaked in any way, it invalidates much of this and if we look at the systems this season compared to last season, you will see what I mean!

Here’s the proofed results on the combined systems to the end of March 2012:



Quite clearly, you can see that the number of bets on all 6 systems aren’t in line with last season. Obviously, we still have 4 weeks of the season to go but basically, system 6-21 is going to end up with around 50% more bets than last season. System 6-22 is going to have around 30% less than last season. 7-21 is going to be 20 bets less which is pretty close.

7-22,8-21 and 8-22 clearly stand out though and all 3 systems are going to have much fewer bets than last season. All 3 are going to end up with between 40%-50% less bets than last season.

Clearly, something has changed as 6-21 has more bets which would imply the systems have found more bets (they have done!) and yet, when they are filtering the bets, they are finding less stronger bets.

Am I surprised? Again, I’ll be 100% honest and say when I was making the minor tweaks to the systems in the preseason, I didn’t expect to cause such a substantial change to the number of bets on each system, so I was a little surprised at first. One problem with tweaking individual systems is that you never know the impact on the combined systems as I’m never sure what impact the tweaks have. Of course, one could say I should test the combined systems whilst making the tweaks but in a way, I think this is wrong. My intention is to get both rating algorithms to do the best they do and that consequently has a knock-on effect on the combined systems. It would be wrong to try to maximise the combined systems by tweaking the rating algorithms as that goes down the road of backfitting and quite simply, one reason my combined systems are so good is because they aren’t backfitted!

Thinking back, the changes weren’t significant on any system. I tightened up the criteria for Premiership Away bets as these were loss making in 2 of the 4 seasons of backtesting and during the first live season, the results weren’t great, so I had to do something there with Prem Aways specifically. The biggest change on all systems was the odds restrictions I applied to the rating algorithms and I basically dropped all Away bets at less than 2.75 (apart from a few that are deemed massive value) and I’ve dropped all Home bets less than 1.67 this season as they don’t produce a high enough ROI. I’ve also tightened up the criteria on Home bets less than 1.80, so there are fewer of these this season.

To offset these changes, I’ve tweaked the max odds restrictions on the algorithms for the Away bets as although these were highly profitable during backtesting, they were causing too much variance. e.g. losing runs were causing the need for bigger betting banks! Hence, we now get more bets as value at big odds than last season but importantly, when filtering the bets, the systems don’t use the value rating on these bets at all as I’d end up with a lot of massive prices on systems 7,8,22 which is not what I want!

Now I write down all the minimal tweaks, you can actually see I’ve tweaked a helluva lot last summer even though I didn’t think I had done! Small tweaks here and there soon add up!

I’ve ended up with more system bets at big prices, no short priced homes or aways and much less Premiership Away bets. Importantly, when you filter the bets down to get on the better systems, they don’t get tricked into using the massive value % thrown up on these aways.

The question people are no doubt waiting for an answer for then is…..has it been worthwhile?

Well, I can’t say categorically for 100% certainty that everything I’ve done has worked this season as it would involve a helluva lot of work but as you would expect, I’ve paid particular attention through the season to this.

A quick summary would be:

- Dropping all Home bets at low odds has worked tremendously this season. They have bounced back well in the second half of the season but I’d have probably dropped them by then anyway as they were terrible in the first half of the season (as all Home bets were). Definitely a good decision.
- Dropping all Away bets at low odds has been an OK decision but ultimately, if I’d left these in, I’d have started the season really poorly but again, these have bounced back and have actually bounced back to profitability I suspect. However, the ROI on low priced Aways will always be lower than I want on my systems, so there will be no going back here.
- The best decision I’ve made this season was to try to adjust the systems to cope better with Premiership Aways rather than drop them all. It would have been so easy to drop all Premiership Aways after last season’s performance and the backtesting results but I decided to do a lot of work and boy, has it paid off! Up to the end of last season, Prem Aways had made a 40pt profit on the Established systems from 727 games. This season, Prem Aways have made a profit of 140pts from 139 bets. WOW! A 100% ROI. :)
- Looking at the results, anyone following the selective systems (6-22 to 8-22) is more likely to get a higher ROI than before but anyone following 6-21 is likely to get more bets for a similar level of ROI.

Interestingly, if I look at the overall results by season for the combined systems, they are remarkably close to the end of March. Hence, anyone following these 6 systems probably doesn’t know what all the fuss is about! I guess the main difference is the lower strike rate as I’ve taken out low odds bets that may have been profitable and replaced them with higher odds bets.



Last but not least, the key thing for me personally (and those following) is that this season has been a much steadier season in terms of profitability and we haven’t seen the same volatility on a daily, weekly or monthly basis which I think vindicates my work last Summer. Yeah, it may look like this season is much in line with last season but considering I tripled my bank last season and then lost a 1/3 of it, I’d much rather have a steady return every month rather than amazing returns and amazing losses!

I can say that this Summer, I will …….NOT be touching any of these 2 rating algorithms I discuss above. I will be tweaking the 3rd rating algorithm (31-33), building a new rating algorithm for the UK leagues (41-42) and of course, building 2 new algorithms for the top 4 Euro leagues. However, I feel like I’ve got these two algorithms exactly where I want them.

I hope the above adds a little insight into the systems and the work I undertook last summer. Can’t wait to get started on the work this summer. :)

2 comments:

  1. Fascinating reading as always.Followed the 7+22 selectiions this year on the blog but can’t wait to join the service next season and follow more systems.

    The chart for 7+22 is as good as Ive seen for any betting system that has actually been tested live and not just some backtested system that never works in real life.

    Astrid

    ReplyDelete
  2. Hi Astrid.

    Thanks for the comment. Good to hear you’ve done OK following system 7-22 mate.

    Keep reading the blog for details on next season’s plans.

    Cheers,

    Graeme

    ReplyDelete