Wednesday 3 September 2014

Guest Blog Post - @pokersunited

One of the things I was keen on this Summer was to try to tap into the vast experience of some of my subscribers. A few people offered up potential ideas for posts but I decided to take a new subscriber up on his offer of a blog post as although this subscriber had no experience of TFA, he had plenty of experience of football systems, poker and bank management and I thought it would make a change to hear from a new TFA user about how he plans to use the systems.

The brains behind the post is one of the authors behind the very interesting and useful site pokersunited. You can also follow at @pokersunited on Twitter.

I hope you enjoy the read and thanks again for taking the time to write this post @pokersunited. :)

...............................................................................................................................................................

As a first time TFA user and former poker professional, Graeme asked me to write a blog post about my approach to TFA systems as it might be slightly different than the way most other people use the service. As a poker player, I’m a very risk-orientated person and have learned that the most important factor when gambling (with a positive expectation!) is money management.

Poker/sportsbetting/investing are all kind of similar in a way that you can make good decisions, but still lose money. The utmost important thing is to minimize the risk of going broke to as close to 0% as possible. As a poker player who has played probably close to 10 million hands, I know a thing or two about variance. 

So just like in poker, my main concern when sports betting, is how to minimize the risk of wiping out my bankroll in a bad run. That doesn't even mean the chance on going broke right away in a bad run, but even losing 75% of a betting/poker roll is a killing for me.

Sports betting has always been a sort of hobby for me next to poker, but over the last year it has turned around and my main focus is on sports betting as poker has changed a lot since black Friday and playing at nights isn't recommended with children.  :)

I found my way into sports betting by first betting games in my own league (Dutch Eredivisie), but last year I discovered blogabet and started following (a lot) of tippers. But as most reading this probably have discovered, there is a lot of garbage in tippers, almost all are looking for a quick buck and do not really understand what sports betting is about. It made me clear that next season should be about quality picks (tippers) and not quantity.

Somewhere last year I discovered Steve’s blog (daily25.com) and through him I found out about TFA. Despite reading about his bad run with TFA I knew that TFA would be one of my (max) 5 services I would follow. Graeme seems to know what he's talking about and I really like the idea of betting through systems and get the emotion out of it.

The first thing I did was analyse all the bets from TFA and thinking about a smart way to make a selection from all those different systems.

Right from the start I didn't like the idea of having to choose a couple of systems and blaming myself at the end of the season picking the wrong ones (like the whole TFA <> Daily25 discussion). 

So I had to find a different way of betting with TFA than the portfolio approach....

Risk management

What surprised me most when reading Steve's and Sportingvalue’s blog  (2 TFA users) was that both had weekends were they had huge bets on one (or even more) games, sometimes up to 9% of their roll (Sportingvalue). How can that ever be justified from a risk perspective....??

So my TFA betting portfolio won't ever be built up by choosing a couple of systems. It will be based on unique picks with a certain value generated by all the systems. This way I may have less picks and therefor a lower ROC, but my max drawdown will be significantly lower than a combined portfolio with a lot of correlation (read: double/triple staked picks).

From what I read from Steve he is very focused on turnover/ROC, but what he has to remember is that increasing turnover by picking a lot of double/triple staked picks is nothing more than upping stakes and therefore, should be accompanied by a higher bank size, which will neutralize the higher ROC he is aiming for. 

To explain myself further I did an analysis on Graeme's data and analysed 3 different 'portfolios':

1). Steve’s portfolio from last year with all the correlating picks staked with the same amount (I made no difference in home/away bets staking wise, I know Steve did, but for this analysis I just use flat staking for all picks).
2). Steve’s portfolio again but this time all correlating picks in different portfolios will be staked just once (resulting in 1/3 of the number of picks in portfolio 1).
3). Not using a portfolio of system but just betting all unique UK picks above a 3% TFA value.

When looking at the past performance of Steve's portfolio, it's clear to see why Steve made the choice for those systems. They've had an amazing year in 2012/2013, with a combined 17.7% ROI, but also 2010/2011 had a 19.2% ROI and 2011/2012 wasn't that shabby either with almost 10%. 

The tricky thing here is that a lot of people will see the 4000 picks/bets that were made during that period (3 seasons) as quite a big sample, certainly big enough that it's almost guaranteed making a healthy profit next season again. Well it would be if those 4000 picks were actually unique ones.....the problem is, those 4000 bets are actually just 1200 unique picks and while 1200 is a nice sample, it's certainly no guarantee that next season will be in the same range of 10-15%.

Now let’s look at some stats about portfolio 1:


While the ROI is the best over a 4-year period (10.5%), the maximum drawdown during this period is a massive 157! points (in the 3 very good seasons before choosing this portfolio it already had a 95! points drawdown, while those were obviously very good seasons) 

On average there are 169 bets per month with an average profit of 17.76pts per month with this portfolio. By calculating the standard deviation (46.2) we can make confidence intervals for monthly results. 

For portfolio 1 we can say that with 99% confidence the results in a month will be between -100 and +136 profit....It's clear that one needs a bank of at least 160 points here, but 200 would be my advice at least (2 times the max loss in a month 1 in 100 times).
If we now calculate ROC by advised bank sizes we can see that ROC has been 319.66% in 4 years. Not bad at all.

Now let’s look at some stats about portfolio 2:


While the ROI is slightly lower here with 9.4%, the maximum drawdown during this period is just 43.6 points (about 25% of portfolio 1). On average there are 54 bets per month with an average profit of 5.07. By calculating the standard deviation (11.4) we can make confidence intervals for monthly results.

For portfolio 2 we can say that with 99% confidence the results in a month will be between -24 and +34.4 profit....This leads to a minumum bank of at least 50.
If we then calculate ROC by advised bank sizes we can see that ROC has been 364.95% in 4 years. So with much less picks (30% of portfolio 1), a lower ROI, we still have a better ROC!
Basically we save time/capital and stress by choosing a portfolio like this one above portfolio 1.

Now let’s look at some stats about portfolio 3:


This one has the lowest ROI of the three with 7.4% , the maximum drawdown during this period is just 32.8 points. On average there are 78 bets per month with an average profit of 6.8. By calculating the standard deviation (12.8) we can make confidence intervals for monthly results.

For portfolio 3 we can say that with 99% confidence the results in a month will be between -26 and +40 profit....This leads to a minimum bank of at least 50 again.
If we then calculate ROC by advised bank sizes we can see that ROC has been 409.71% in 4 years. 

So what we see is that the portfolio with the lowest ROI, has the best ROC. This is because it has 40% more picks than system 2 and a lot better risk profile (resulting in a lower bank needed) then portfolio 1.

With the conclusions I make here I don't want to ruin Graeme's portfolio approach, because I still think his idea is nice, I would just use his systems in another way. My idea would be to make a couple of absolutely NON-correlated portfolios namely: UK home, UK draw, UK away, EU home, EU draw, EU away and maybe even something like away favourites/away outsiders or you name it. 
Graeme could then decide when to add a pick to UK aways, if it shows up in just 1 algorithm or maybe only when it shows in 3 of them etc...?

In this way there would be 10 portfolios at most, a lot less picks and swings and probably more profitable and happy customers in the end :-)

6 comments:

  1. It is impossible to put in to words how much I agree with this post. I even like the suggested alternative portfolios (If only Graeme could generate more away favorite picks). Thank you Pokers United. Now if TFA can only create an Asian Handicap system that generates even more bets we can really start to turn over a bank.

    Regards,

    Matthew Trenhaile.

    ReplyDelete
  2. Thanks for the kind words Matthew, appreciate it.

    Regards,

    Martin/@pokersunited

    ReplyDelete
  3. This is an absolutely excellent post. I have been very busy for a while and was planning to do some work on how to use the picks, but you have saved me a lot of time and work. I can't thank you enough for some exceptionally detailed and clear work. I'm a fellow ex-professional online player and I can't help but wish I had applied the same rigour to my bankroll management while I played (my mentor was/is an utterly godly player, but has always been staked and has always has massive disdain for bankroll management; I'm just learning how important it is).

    ReplyDelete
  4. Great post Martin!

    For the alternative portfolio 3, which rating algorithm did you use to determine if the pick was above a 3% value? For example, some selections will show as above 3% value on one system, but as negative value on another.

    Regards

    James

    ReplyDelete
  5. Hey James,

    I took all unique picks that had at least one rating above 3%, the system doesn't matter there. Could be a nice research if results are getting better, when taking only picks that have on average more then 3%....but haven't done that yet...maybe when I have some spare time left ;-)

    Regards,

    Martin

    ReplyDelete
  6. This is great. Thanks a lot.

    I had already made the decision to only follow a +4% ratings system but this really nailed the analysis for me. Gives a lot more comfort for my decision going forward.

    ReplyDelete