Another shocking week for the systems and I’m starting to
worry just a little bit now. Given we’re only about 60% of the way through the
season, my worry now is that the betting banks for various systems are starting
to come under a little bit of pressure. I’ll do a piece of analysis at the end
of the month to look at the drawdowns on each system as quite a few are now at
historical drawdowns I suspect but I’m praying that things turn soon.

After this weekend, my proofed results for this season
are now showing a loss and to be loss making after 728 bets this season isn’t a
great situation to be in. I think the thing I’m personally struggling with is
that the systems only suffered 3 losing months in their first 20 live months
and yet, January is going to be the 3

^{rd}losing month in the last 3 months! My overall ROI since inception has dropped to only 11% which is very disappointing.
It’s quite difficult to understand I think and my own
betting P&L has taken a massive plunge over the last month or so and
ultimately, I’m sitting here writing this tonight and questioning whether this
is all worth it or not. I’m not a great loser as I’ve said before and I do take
losing runs quite badly and at the moment, this is easily the lowest I’ve felt
since I switched my attention to football betting 3 years ago.

As I said in the email last night, I just need to
continue on and hope things improve. Thousands
of live results and tens of thousands of backtested results lead me to believe
that things will turn around at some point but boy, could I do with it
happening soon!

So, the last week then.
Can be summed up in one phrase I think….too many draws.

__Est Systems__

__New Systems__

A massive loss here. A loss of 33.86pts from 70 bets.
Horrible. Clearly, the draw was to blame in the main with a profit of 3.33pts
is using AH0.5 on all bets.

__Misc Systems__

Another big loss here. A loss of 20.61pts from 42 bets. Disgusting.

__Under/Over Systems__

A profit again on these systems. 2.31pts profit from 8 games.
Shocking (in the surprised sense!)

__Euro Systems__

I do feel a little sorry for these Euro systems this week. It’s
a massive loss but with a bit of luck, they should have had a good weekend,
bordering on a great weekend. A loss of 32.9pts from 80 games and yet, a profit
of 6.3pts if using AH0.5. Just about sums it up.

Overall, another terrible update. A loss of 104.71pts
from 242 bets.

Already looking forward to the next update……NOT!

Hi Graeme,

ReplyDeleteHope things work out for you. Just have to hope this is a blip and results turn your way. Try to keep positive. Always here if I can be of help.

Best wishes,

Jason

Chin up fella. Is there really anything in the recent results that is out of whack with expected variance? I don't know, but I suspect not.

ReplyDeleteI actually love these difficult drawdown periods for tipsters I follow, as it weeds out the dilettante gamblers who haven't got the stomach for it. And it's this weeding out which means the edge you have is sustainable.

Am sure things will turn around at some point. And remember, as I'm sure I've heard you say before, if this was easy, everyone would be doing it and the edge would very quickly disappear.

Cheers

Paul

Now with more than 2.5 years of live results it’s possible to draw some statistical conclusions about the TFA betting system.

ReplyDeleteStatistical inference can be obtained via tests. A test calculates a p-value, which represent a measure of ‘luck’.

F.ex. system 6 has now 1709 tips with a ROI of 4.1% and average odds of 3.3. So if you had followed system 6 for all 1709 tips you would have earned a nice profit of 4.1%. System 6 has a p-value of 13.9%.

You can think about this number in the following way.

Imagine it was possible to replay these 1709 games 100 times under the exact same conditions. Then by simple chance there would be 14 of these 100 replicates that gave the same or better profit as TFA obtained.

Most statisticians would require a p-value of 1% or less to consider the actual result significant.

The p-value for system 21 is 1.19%.

So if I had to make a choice between system 6 and system 21 the answer would obviously be system 21.

Now one of the basic assumptions about the TFA system is that by filtering f.ex. system 21 you get system 22 and this system is better than system 21.

I have not seen any evidence of this yet, so I have tested the assumption.

The p-value for system 22 is 12.7%, so system 22 is definitely not better than system 21.

And unfortunately this is not isolated to system 21-22.

The exact same conclusion can be drawn by comparing system 6 to system 8, system 31 to system 33 and system 41 to system 42.

Now this result can be ignored if you follow a single system like system 21.

But if you follow f.ex. system 21 AND system 22 you get in trouble.

Then you play 2 units on system 22 teams and 1 unit on teams which only are system 21 picks.

So you risk more on the teams which statistically perform worse.

I can not imagine anyone is playing ALL systems.

However the performance of the TFA system is measured by the grand total, which represents such a system.

Last week there were 13 UK teams advised, but 154 units staked on these teams. Ranging from 1 unit on Cambridge and Ross County to 22 units on Oldham and Bury.

NO football system should use variable staking this extreme, especially not with the results shown in this post.

SecondWalz

SW - could you post the p-values for all of the systems please?

ReplyDeleteHi Jason.

ReplyDeleteCheers for the comment. Now I’ve been doing this for a few years, I’m never too surprised at the ups and downs that come with betting on football. I think because the odds are fairly short (I come from a horseracing betting background where hitting 30-40 losers in a row wasn’t a disaster for me!), I automatically expect a much smoother ride for my betting.

This season has been a farce really for me and putting aside the systems’ performance (outside of my own control), my P&L for the 2012/13 season looks like it would be the best rollercoaster in the world. I was on a massive drawdown again after the weekend past and it is mentally draining at times.

Although this isn’t known on the blog (try my best to not moan too much on here!), this season has actually been really difficult for me in terms of getting the bets out on time, placing my bets, updating results, keeping track of my own P&L and transferring funds between my bank and bookmaker accounts etc. I don’t want to try to count the number of hours a week I spend on the footie these days but it’s an unhealthy amount considering I work in a pretty demanding full time job which has involved working far too many hours in the last 6 months or so!

My posts on the blog do sometimes come across as a little too honest but I don’t want anyone to think that making money from following my systems or even from gambling generally is easy. I don’t believe it is!

Cheers for the comment.

Graeme

Hi Paul.

ReplyDeleteI think for a lot of the systems, it may be the case that none of the things we’re seeing is too unexpected although it’s all relative to what you expect to happen I think. I had a plan this season for what I was aiming for and what I expected to happen and ultimately, in my own portfolio of systems I follow, I’ve had two losing runs that were as long as the portfolio had witnessed over all the previous seasons historical results! Having one bad spell a season isn’t a disaster but my own P&L has had two!

I know there are a number of people following the systems who aren’t having a disaster of a season but then again, I have heard from a few who are having a very poor season. Over the first two seasons, I had followers who all had variable P&Ls but it was about who won the most money. At the moment, it feels if you’re not losing money, you’re doing well!

Similar to my comment above, my posts on here probably match my mood at times from seeing my own P&L results rather than the results I’m posting up here but let’s just say that my portfolio has seen better months than Jan-13! :(

I’m looking forward to Feb-13 being a better month.

Graeme

Hi Secondwaltz.

ReplyDeleteNice to hear from you again. As you know, I’m not really into statistics as much as you (I doubt few people are) but I don’t think you need any statistical significance tests to draw conclusions about which systems are statistically significant. You could take a pretty good guess from looking at the system results to date and draw a similar conclusion.

The issue doesn’t lie with knowing system 21 is more significant than 6 and that 6 and 21 are more significant than 7,8 and 22. For most people following, it is about making profits from following the least number of bets possible.

I don’t think people follow a system like 7-22 because they think it’s more significant than system 21. Clearly, one system has over 1,200 proofed bets and one has about 200 proofed bets, so system 21 is clearly more proven to work but people don’t want to follow hundreds of bets a season. They want to follow the least number of bets possible in my experience.

As for your point about following all systems, that’s not really an optimal solution and never will be. Staking 28pts as a max bet and 1pt as a min bet can’t be an optimal strategy! However, this blog is detailing the results of all systems. I know a 100pt loss doesn’t mean everyone lost 100pts this week but then again, it paints a good picture of performance of the systems at any moment in time.

Cheers for the comment.

Graeme

Hi Graeme

DeleteWe don’t live in a deterministic world.

But some of us believe that it’s possible to tell something about the future by looking at the past.

Maybe even you think so too. After all, you took 10 years of football results and made an attempt of predicting future games based on these many datapoints.

Mathematical statistics is a way of quantifying the predictions, putting numbers, probabilities on future events.

You should be happy when I tell you that system 21 has a p-value of 1.19. This is not a proof that this system will make money going forward, but a very strong indication. Very, very few tipsters are selling systems with such a fine p-value.

On the other hand, you should be concerned about a system with a p-value of say 40%.

Regarding the results that system 22 is not statistically better than system 21 could inspire you to investigate and examine if this part of your model (the filtering) is good enough. The tests do indicate that something is not working like you thought it would do.

And lastly my remarks about using the grand total as a measure for the combined TFA system.

If, and I say if, there are good and not so good systems among the many subsystems, an average can not tell the important things.

If you have 2 people with an IQ of 120 and 4 people with IQ 90, they have a sum of 600, exactly average for 6 people. But if you should hire one of these people to your company I feel confident, that you want to know who has IQ 120, not just that their average is 100.

So I don’t agree with you, that the grand total paints a good picture.

SecondWalz

Hi Anon.

ReplyDeleteAs I said to Secondwaltz, I don’t think you need to be a rocket scientist to look at the results of all systems and rank them in order of significance. Just because a system has a good p-value doesn’t mean it is any more likely to produce a profit going forward!

Graeme

OK Graeme, you non-rocket-scientist.

ReplyDeleteI have three betting records.

Please rank them in order of significance.

System “Home favorite”: PL= 70, 1000 bets, average odds=2.00

System “Away dogs”: PL= 100, 1000 bets, average odds=5.00

System “Horses”: PL=150, 1000 bets, average odds = 10.00

SecondWalz

Hi SW.

DeleteI’ve got enough on my plate at the moment with betting and doing ratings, never mind playing games out on the blog!

Anyway, to answer the question posed……

I think it’s fairly easy in this example. If given the choice of these 3, I would say the order of significance is as you have it. The lowest odds being the most significant, then the 4/1 and then the 9/1. How did a non-mathematician such as myself arrive at this?

Well, I always relate things back to experience and prior beliefs rather than statistical testing. I know that following bets at odds of 2.00 means that the losing runs are going to be short, the betting bank needed won’t be huge (35pts?) and therefore, to make 70pts is really good. Making a 7% ROI on this odds range is actually very difficult I think and therefore, to do so over 1,000 bets is exceptional. I’d say this is a proven system based on these figures.

For the average odds of 5.00, losing runs will be a bit longer (potentially 2.5 times the expected drawdowns on the previous system?) and therefore, to be as good as the previous system, I’d be looking for a profit 2.5 times higher than the first system. 30pts more isn’t really that much and if given the choice of following the first two systems, surely anyone with a brain would choose system 1?

As for system 3, playing at average odds of 10.00 is similar to horseracing and having spent a good few years betting seriously on the racing before my football journey started, I know only too well the drawdowns and losing streaks involved here. So, 150pts isn’t actually that much profit at these odds and a couple of losing streaks or winning streaks could see huge swings, so I’d be willing to bet a fair bit this is the least proven of the systems statistically.

Likewise, if given the choice between 1 and 3, it really is no decision. It’s like chosen between a football tipster and a racing tipster. Too many chase the high ROI!

So, that’s how someone such as little old me judges systems.

Now, I may not have an average IQ (achieved 156 in a mensa test at university) but I don’t think you need to be a rocket scientist to choose between betting systems. Yes, if using significance tests, you’re taking away the human aspect and forming accurate judgements but my point to Anonymous in the post was that if you analyse the results yourself, you can come up with some sound conclusions.

I think 99% of subscribers know that the lower systems are the most significant statistically and are also the systems that can create the highest ROC. However, you have to work hard to do that and most subscribers want shortcuts where they can follow much fewer bets for higher stakes and achieve good profits. Anyone doing this knows the risk they run with variance. Similarly, anyone using highly correlated systems like the 6 combined systems and following all of them knows they are using a 1-6pt staking plan and therefore, if the 6pts bets start losing, they’ll be in trouble.

Anyway, let me know if I’m right or not above in my non statistical analysis!

Graeme

Congratulations Graeme!

ReplyDeleteThe solution is correct.

You draw on your great experience in sports betting to arrive at this.

This year it’s exactly 50 years since I started studying mathematics at university.

So I use this experience when solving betting problems.

But how can we expect Anon, who asked the question in the first place, to solve the puzzle?

Also notice that your solid reasoning builds very heavily on your knowledge of the odds.

You do never report average odds for the systems in your summaries.

This means that neither you or a rocket scientist can evaluate the systems correctly without further information. It’s simply logically impossible.

Even though I love to play all sorts of games, these posts are not games for me.

I hope your readers will benefit from them.

Just for the record:

System “Home favorite” has a p-value of 1.3%

System “Away dogs” has a p-value of 6.4%

System “Horses” has a p-value of 6.9%

With Excel and a simple formula it takes me about 40 seconds to get these numbers.

SecondWalz

Phew, feel relieved now SW. I can get on with updating the blog for the midweek results. :)

DeleteAs you probably know, I do use average odds a fair bit when analysing my data at times but you are also correct in that I don’t show it in my analysis posts on the blog. I’ll start to include it in tables etc. now for you and any others to use.

Cheers,

Graeme