Monday, 4 June 2012

Systems 31-33 Reviews

Below are the first 3 reviews of the new systems from last season. As I said on a recent post, these will be shorter in nature since it's the first season for these and for most systems, I don't have enough results to draw any meaningful conclusions. 

System 31

A slightly disappointing first season for the new 3rd rating algorithm this season although quite early on, I thought it was going to be a disaster.  An ROI of 3.8% across 671 bets is about the minimum I’d accept over a season I think over such a large number of bets though, so it wasn’t a disaster.

The P&L graph this season is a bit of a rollercoaster. Before the season had barely started, the system was -20pts after 60 bets.  It then recovered to break-even by around the 150 bet mark. A period of stability then appeared with nothing happening over the next 300 bets! It then won 20pts quickly and lost them just as quickly!

So, break-even across the first 500 bets and I was wondering why the hell I bothered with building a 3rd algorithm.  Out of nowhere basically, a profit of 50pts across the next 50 bets and at that stage, it was running with an ROI of about 8% and I thought this was highly acceptable.  Unfortunately, a poor run struck again and it lost 25pts before stabilising and finishing the season on a profit of just over 25pts.

The AH returns struggled all season. After a poor start, it was basically 500pts before AH0.25 and AH0.5 got level on the season.  Interestingly, both ended the season with returns not too far behind betting outright, so it’s a difficult one to fathom!

A quick look at the results by League and you can see a very nasty 25pt loss in League Two this season. Drill down deeper and nearly 23pts was lost on League Two aways. That alone accounts for much of the underperformance this season.

Only 3 winning months from 9 just about sums up how much the system struggled this season. However, when it gets it right, it can produce some excellent returns in a short space of time I think. It just doesn’t get it right too often!

Homes and Aways produced very similar returns over the season, but way down on what was hoped for before the season started.

With no live data last Summer, I suggested a betting bank of 70pts.  A profit this season of 25.7pts means a ROC of 36.7%.  Not what I was hoping for but when you think about it rationally, it’s not a disaster over the course of the season.

The max drawdown this season was actually lower than historically, so I think 70pts is probably about right. Although backtested results aren’t useful in the sense of setting target returns, for historical drawdowns, I’ve always liked the fact they help you set a betting bank which isn’t just finger in the air!

Ok, going forward, 70pts bank, probably looking at 650 bets a season and a 5% ROI as a target.  That will give a target 46% ROC. Not bad I think considering I’ve called it a rubbish system all season!

System 32

The test for this system is whether it improves the performance of system 31. We can definitely say it has!  An ROI of 11.1% on this system across 222 bets is actually a very good performance and it’s the most pleasing performance of all the new systems I think.

The system started the season poorly but soon recovered and went on a quite amazing run. It then lost form for a long time, made small losses over a large number of bets and then bounced back a little at the end of the season!

Interestingly, not sure how it managed it but this is maybe one of the first systems I think I’ve looked at who made a profit on League One and League Two aways this season. Not just that, it actually made a good profit on both types of bets.

System 31 had 28 winners from 127 away bets in League Two. System 32 has had 16 winners from 52 bets. The other 75 bets that system 32 didn’t like only produced 12 winners. Quite amazing really! 

5 losing months in 9 months though shows the turbulent nature of the system on a month by month basis and interestingly, this was apparent in the backtested results too. Aways produced a return of 12.1% with Homes producing 8.5%. Homes tend to not do great during backtesting anyway, so 8.5% isn’t bad on this system.

The AH returns on this system weren’t bad too with a 6.6% return coming from backing all bets AH0.5.

Last Summer, I suggested a betting bank of 50pts based on the max drawdown of 17pts. The max drawdown is only 18pts now, so I think a 50pt bank is sufficient.

A profit this season of 24.6pts meant a ROC of 49%. 

Going forward, a 50pt bank, 225 bets and a ROI target of 10% I think. Gives a ROC target of 45%.  Much in line with system 31 interestingly.

System 33

This carries the distinction of being the first ever losing system I have had over the first two seasons.  11 systems during the first season, 20 systems during the second season and yet, this is the only system that has ever lost over the season!

To keep things in perspective, the system lost 3.9pts across 126 bets.  Yes, it’s annoying to have a losing system over a season but with one or two winners, it would have turned around and scraped a small profit.  In addition, if you look at the table of returns, covering the draw in any way (AH0, AH0.25 or AH0.5) produced  a profit over all bets, so it’s simply the case the system has had a few too many draws this season which ultimately, has ensured it has made a loss.

The fact that system 32 had a profit of 11% across 222 bets and yet, system 33 has taken these 222 bets, picked out 126 of them and made a loss shows the impact that variance can have on a system I think. 

In a way, it’s good to see that a system can lose over a season as it shows what I’m doing isn’t infallible. They call this gambling for a reason. Thankfully, I don’t know anyone who made a loss over the season as no one would surely choose to follow system 33 as their main system given the proofed results of all other systems, so it’s not a disaster that this has happened.

I’m not going to overanalyse 126 bets as I’ll be wasting my time but Homes made a loss of 9.3% and yet, if dutching the draw and the selection (AH0.5), a profit of 15.5% was achieved. A massive swing and clearly, too many of these Home bets finished in a draw.

Small losses were achieved in most leagues and 5 of the 9 months were losing months!

Last Summer, I suggested a betting bank of 40pts. The largest drawdown was 16pts, so even though it was a losing season, the betting bank was never under pressure and the drawdown this season is only just worse that the max historical drawdown, so I don’t think there is anything to change on this system apart from hope there is a bit better luck next season!

So, a loss of 3.9pts and a bank of 40pts means a ROC of -9.8% this season.

Going forward, I’d say a bank of 40pts is adequate. I’d expect around 100 bets and maybe a ROI of 10% is what we should be aiming for (it can’t be any lower than system 32 as it’s meant to be filtering the best bets from system 32!).  That would give a ROC of 25%. 

Given the performance this season, my recommendation would be to NOT follow this system next season. 

No comments:

Post a Comment