Quick recap on value and a quick look at variable staking

When I logged on tonight to start writing this post, I had a quick look through the blogs on the right as I usually do and I saw Cassini over at http://green-all-over.blogspot.com/ had commented on variable staking. As usual, he’s spot on with his commentary and I don’t have anything more to add on the subject of variable staking in the sense of trying to turn a losing system at level stakes into a winning system.

In my opinion, if your system doesn’t have an edge at level stakes, then you can win in the short-term using some sort of variable staking strategy (usually one that involves chasing losses) but in the long-term, you will lose. That’s actually not just an opinion, it’s a fact. I’ve read lots of stuff on the internet over the years related to gambling which involve an 80%+ strike rate and involves chasing losses and for anyone without a brain, it does look like a strategy that is full-proof. Ultimately, without an edge, you won’t beat the person laying you the odds and it might takes weeks, months or even years before you lose your bank but rest assured, it will happen eventually.

Anyway, that’s not really the purpose of my post. I do like to think I share many of the same philosophies as Cassini when it comes to gambling but he’ll probably say I learned them all from reading his blog over the last few years! ;)

A couple of comments in the past few days have asked about variable staking with my footie selections which is the real reason for this post. I’ll also take this opportunity to explain a little more about the ratings and how they are good at determining value.

Cassini touched on it in his post tonight but long-term, you ideally want a system that is capable of finding value but for the Holy Grail, you ideally want the system to identify the better bets to stake most money on. i.e. the ones with the highest edge. This was one of the key things I had in my mind when I started work on this footie project 16 months ago and I think I can say that this has been a success this season. Not sure much else has been a success this season though after what’s happened since Christmas!

A quick summary of the systems again then for anyone who has missed this or not read the backlog of posts. There are two sets of ratings produced for every game I rate. One set is algorithm A and one set is algorithm B. System 6 has all the bets from algorithm A that are defined as value and system 21 has all the bets from algorithm B defined as value.

System 7/8/9 then narrow down the bets on system 6. As you go higher up the systems, the more selective the bets are and the higher the edge there is on each bet. Likewise, system 22 and 23 take the bets on system 21 and narrow them down to get the best bets.

Historically, the results have been in line with this idea most seasons on each system and overall, it is definitely the case that system 8 is better than system 7 and system 7 is better than system 6 and so on. What has happened this season?

Well, if you look back at my post a few days ago, it shows the results so far this season on all the systems and helps to illustrate my point. http://the-football-analyst.blogspot.com/2011/04/delusions-of-grandeur.html

I don’t want to spell it out too much but system 8 has an ROI of 8.8% against system 7 with an ROI of 4% and 6 with an ROI of 2.8%. Likewise, system 22 has an ROI of 9.1% and system 21 has an ROI of 4.5%.

The eagle eyed amongst you will be wondering why I have left out system 9 and system 23 from my analysis. Well, if you were to look back at the historical results tab and study the returns from these systems in past season, you will see that now and again, these systems do end up with lower returns than the less selective systems which goes against my idea of filtering the bets to find value which doesn’t happen on the other systems. What causes this I hear you ask? Simply, it is a lack of bets.

System 9 has only had 45 bets this season and system 23 has only had 69 bets this season. As you will be aware, over a small sample of bets, it isn’t necessarily the case that a significant edge can show through and quite often, if a couple of games go against you (as happened this year on both these systems I guess), it can turn a winning system into a losing system.

I knew from my backtesting that these systems probably wouldn’t have enough bets in a season to be a meaningful system but I kept them in here for this season. Hence, I don’t want to read too much into these systems and like systems 10,24 and 25, they’ll probably be consigned to the shelve due to a lack of volume of bets after this season.

If you look at the results at Christmas time in that table above, you can see the same effect where the higher numbered systems have a higher return. Again though, the small sample of bets can do funny things to this relationship between systems and system 9 had only an 11.4% return then when system 8 had an 18.5% return. Interestingly, system 23 had a return of 67.3% of Christmas from 20 bets and 49 bets later, it is making a small loss now! It shows that anything can happen in the short-term at this game.

The systems I’ve not touched on then are the systems where teams appear on both algorithms. Hence, when the first set of ratings pick a team and then the second set of ratings pick the same team. As discussed on the blog before, this is something that I didn’t trial before this season (I simply didn’t think about it last summer!) but I quickly picked up on the fact the results were very good when teams were picked on the first set of systems and the second set of systems and started to track it.

Looking at the results so far this season again, you can see that system 8-22 beats system 7-22 and 7-22 beats system 6-22. The returns of 28%,26% and 14.6% are exceptional on these systems. Likewise, interestingly, 8-21 is a little behind 7-21 this season but 7-21 beats 6-21. Again, this is likely to be simply down to a smaller sample of bets more than anything. If you look at the historical results, you’ll see that 8-21 is definitely a better system than 7-21.

Overall then, one thing that is clear this season is that my systems do their job at finding out which bets are the best value bets.

What does all of this mean? Well, it means that you don’t need to follow all of the value bets that appear as you can cherry pick the best bets or if you like (which is what I prefer personally), you can follow all of the bets with a staking plan whereby you stake each team as it appears. Hence, a bet that appears on system 8 is staked more heavily than a team that only appears on system 6 since the system 8 team appears on 6,7 and 8. You may not want to stake 3 times as much on it but simply, you know you are maximising your return if you stake more on it.

On the other hand, others get attracted to higher ROI’s and the less number of bets and would prefer to only follow bets that appear on system 8-22 for example, ignoring all the other value bets that appear on the other systems. With the systems, everyone can choose a strategy that suits their risk aversion but at the end of the day, you need the ratings to have an edge. If there is no edge, you will make a loss regardless of what system you play!

OK, so that’s a quick summary of the systems for those who weren’t aware of how they worked. The next question then is the one I set out answering, can variable staking increase the returns of the systems?

I’ll be honest and say that I’m from the school of thought that level staking works best but I do like to maximise my edge. Simple, I like to stake more on a team that appears on system 8 than a team that only appears on system 6 since I know the teams on system 8 have a bigger edge. Likewise, I’d be a fool to stake the same on a team that only appears on system 6 as one that appears on system 8-22.

However, this isn’t really answering the question that the commenter asked. That was around picking a system and playing variable staking on a system and comparing that to level staking.

I always find the best way to look at this is through an example. I’ll look at two examples. System 8 is the system I recommended everyone play this season at the outset as it was the one I had most confidence in. System 8-22 is the other one that many people have played and until the last 8 weeks, it was an amazing system with very high returns but it has gone off the rails recently.

At the start of this season for system 8, I suggested a bank of 35pts would be sufficient for this system. How has level staking compared to variable staking?

Well, level staking has produced a profit of 13.1pts and an ROI of 8.8%. A disappointing return considering it made a return of 20% in all the backtested and backfitted seasons. The return on capital (profit/bank) would be 37.4%. Again, very disappointing as I was hoping this system would be able to double the betting bank this season.

What would have happened if I had been playing variable staking? For this, I will assume daily compounding. Hence, I increase stakes every day by readjusting my bank at the end of the day. There are lots of different ways to do this (every week, every month etc.) but I’ll stick with the easiest one to track in Excel. I will start with a 35pt bank and play 1/35 of the bank on every bet this season. What does the return look like?

Well, on the 18th March, the bank reached a peak of 64pts but as at today, the bank would be at 46.1pts. This is a profit of 11.1pts this season. I would have staked 206pts this season and therefore, the ROI would be 5.4%. Therefore, a lower return than level staking.

How does system 8-22 look? Again, I suggested a 35pt bank and therefore, I’ll assume 1/35 stake on each bet at the start of every day.

Well, on Feruary 1st, the bank reached a peak of 91pts (a profit of 56pts amazingly) but as of today, the bank would be 65pts. A profit of 30pts this season. The stakes would have been 170pts, so the ROI would be 17.6%. Therefore, a lower return than level staking.

Clearly, this is only a brief look at variable staking and I could have picked on any system to show this but simply, why is it that variable staking isn’t beating level staking this season on my systems (assuming the other systems follow the same pattern)?

It’s simply the way the returns have been achieved this season. Due to the fact that the first half of the season was great and the second half has been poor, the bank grew steadily during the first half of the season with variable staking and dropped more quickly with the higher stakes being played during the second half of the season.

Obviously, if I had looked at this at Christmas, variable staking would have beaten level staking easily as the first half of the season was much better.

One thing that people need to realise though is that based on my previous 4 season’s results and backed up by this season’s results, there is a seasonality element that seems to affect the results. Before this season started, the best months for the systems have been September, October, November and December. The worst months have been January, February, March, April and May.

What has happened this season? The first 4 months created profits of 300pts. What has happened since Christmas? The systems have created a loss of 63pts.

Hence, variable staking (in the sense of increasing stakes as the season progresses or as the bank grows) could never be a way to maximise profits with my systems.

The systems have only had 6 losing months to date. May 2007, Jan 2008, March 2008, October 2008, February 2009, February 2011. April 2011 will be the 7th losing month in 5 seasons.

If you were looking for a variable staking plan and were much smarter than me, reducing stakes from Christmas may have been the way to go based on past seasons. Unfortunately, I didn’t heed this warning and neither did any others I know using the systems this season……

Overall then, I’m not a fan of variable staking as it assumes that each bet in a sequence has the same edge as the previous bet and you could argue due to the fact that you are increasing stakes, you should be looking for a greater edge with each next bet in the sequence. In reality, historical evidence as well as this season’s evidence points to the fact that increasing stakes as the season progresses and as your bank grows is not a way to increase profits. It could actually be a way to increase losses as we’ve seen this season!

In reality, I could look at DC or DNB this season and I would bet that it shows variable staking being better or as good as level staking in most cases. Again though, this is linked to the large number of draws after Christmas this season and nothing more than a short-term blip I hope!

At the end of the day, people can pick and choose whatever staking plan they want to play. I play my own staking plan and I’ll happily admit to the fact I’ve made lots of mistakes this season but this season was always going to be a trial season for me as well as any others following. I intend to do better next season……..

If anyone wants me to run some sort of staking plan on a system this season (or even for double chance or DNB), just leave a comment and I’ll take a look at it for you and post the results and compare them to level staking.