Archive

Archive for the ‘Reviews’ Category

The Football Analyst analysed

July 27, 2011 6 comments

The Football Analyst is a system-based football betting advisory service run by a fella called Graeme Dand. Graeme contacted me through my Soccer Systems blog partway through last season and initially there was a bit of a personality clash and we perhaps got off on the wrong foot. Since then, and especially in the last few weeks, Graeme and I have had some good challenging discussions on the SBC forum and via email and I think we both now have a much better understanding of where the other is coming from. Most of the discussions we have had concern The Football Analyst so I figured it was about time I had a look at the service in more detail.

Overview
The Football Analyst uses a system-based approach, which makes sense when you understand Graeme’s fondness for data, spreadsheets and analysis techniques. Over the last couple of years Graeme has developed a number of betting systems, many of which underwent a live trial on his blog last season resulting in what have come to be known as the established systems. There are also several new systems being trialled over the coming season but as these have yet to have any live testing I will limit my analysis to the established systems only, a total of 11 systems.

It’s not my place to discuss how the systems were developed but I think a few explanatory words concerning the names are required. The various systems on offer have very simple names: System 6, System 21 etc. It’s refreshing to see a nice, simplistic approach to naming rather than trying too hard to come up with grandiose sounding titles. The naming convention also allows the user some insight into the system development methods. The 11 established systems are split into five ‘single-number’ systems and six ‘combo’ systems. The former are individual ideas worked up into systems so System 6 presumably looks to exploit a different angle to System 8 and so on. The combo systems are the common selections from the constituent systems, e.g. System 6-21 is the overlap between System 6 and System 21 so selections in System 6-21 are those that appear on both System 6 and System 21. Not everyone would think to develop new systems from the intersection of others so full credit to Graeme for thinking outside the box here.

Data
In order to analyse any system I need data. Graeme was kind enough to send me data for all of his systems covering backtesting and live trial data where available. He also sent me performance summaries for the established systems, showing the number of bets, winners, SR, profit and ROI for each system broken down by season, division and so on. But not being one to take too much on trust when it comes to gambling I wanted to verify these findings for myself.

Unfortunately, I can’t completely verify Graeme’s figures as the team names have been removed from his data. Bets are identified by date and division and include the predicted result, actual result and the odds. Normally I wouldn’t take the match result and odds from a tipster’s record but would extract them from my own database but that’s not possible here as I require the team names to uniquely identify the matches. I can recalculate whether each bet was a winner or a loser and the profit/loss of each bet so I can at least double-check some of Graeme’s workings but I have to take the rest on trust, which doesn’t sit too comfortably with me. That said, I don’t think Graeme is trying to screw me (or anyone else). He told me that he doesn’t keep the team names in his analysis sheets which is a little unfortunate as it means I can’t verify as much of the data as I would have liked so I will just have to make do really. It puts a slight caveat on my findings but as I said I don’t think Graeme is deliberate obscuring data to make verification impossible, it’s just how he chooses to work with his data.

Analysis
As the service on offer includes a number of systems it may be tempting to ask the service operator for some advice on how to get the best from what is on offer. Graeme will be the first to admit he doesn’t have all the answers when it comes to the best way to use his systems so I will perform some analysis of my own. It’s a matter of personal preference anyway when it comes to balancing risk and reward and volumes of bets so I would always prefer to do my own analysis rather than rely on someone else’s opinion.

I have split the data from Graeme into the individual systems, recalculated some aspects of it and also analysed the results in terms of seasonal returns, performance by division and so on. Where possible I have verified my figures against the summary data I was sent so I have confirmation that my figures and Graeme’s match. Let’s take a look at each system individually.

Over the past five seasons System 6 has seen in excess of 3000 bets with a strike rate of just over 45% generating 468.85pts in profit at a ROI of 15.06%. On the face of it these are attractive figures but it is worth noting that there has been a year-on-year decrease in profit and ROI with last season returning less than 25pts profit at a ROI of less than 5% (24.68pts at 4.31% to be exact). This is significantly down on the first season performance which saw a profit of 169.07pts with a ROI of 22.66%. The SR is also much lower for last season than for any of the previous seasons. Only 41.71% of bets last season were winners, a few percent lower than average and nearly 2.5% lower than the previous worst (44.10% in 07-08). There is not too much of significance in the performance breakdown by division. The system is profitable across all divisions although not so much in the Championship. One can’t expect any system to be equally effective across all leagues though; there will be some natural variance and I think that’s all we’re seeing here. The same is true of the monthly split so no need to look any deeper at that.

The breakeven SR, based on average win odds of 2.55, comes out at 39.27% which is quite a bit lower than the average SR for the past five years suggesting there is an edge to this system but last season’s results show a much reduced edge with the lower SR. The longest losing run (LLR) saw 13 consecutive losers, which is perfectly in line with the calculated expected longest losing run (ELLR) of 14. In other words, things have already gone about as badly as we could reasonably expect them to go on at least one occasion and the system has survived. The profit peaked at 471.64pts and reached a low of -6.99pts but those figures assume each bet runs consecutively rather than some running concurrently so they are a little misleading but useful nevertheless. Based on the ELLR and previous drawdowns one could probably get away with a bank of around 30pts for this system but a larger sum would obviously provide more security.

System 7 shows a number of similar characteristics to System 6 as it happens. It’s not as busy as the previous system with 1516 bets of which 729 (48.09%) were winners generating a profit of 258.20pts (17.03%) but the SR and ROI are slightly higher. The same trend across the seasons can be observed too with the SR for the 10-11 season much lower than previous seasons (43.34% against a previous low of 46.25% in 07-08) and nearly 5% down on the five-year average. Unsurprisingly the relatively low SR results in a smaller profit too, just 16.19pts last season at a rate of 5.52%. In fact the year-on-year drop in ROI is evident once more with the latest season’s ROI not much more than one-fifth of the 06-07 figure (25.26%). The divisional breakdown doesn’t show anything of great concern. The relatively poor Championship performance is evident once more but I don’t think it’s worth worrying about. Incidentally the strike rate for the SPL is under 42% but the bets in that division show an above-average return of 19.23% demonstrating that strike rate isn’t everything. Presumably there has been some value opposing the Old Firm over the years. The monthly splits show losses in February and April over the years but I don’t feel it is anything significant.

With average win odds of 2.43 resulting in a breakeven SR of 41.09% it looks like there is a decent edge here again but once more the most recent SR indicates a much smaller edge. System 7 has yet to experience the worst of it, based on the LLR and ELLR of 8 and 12 respectively. In other words there has been at least one losing run of 8 bets but given the sample size and strike rate one could reasonably expect to have seen a run of 12 consecutive losers. The total profit peaked at 272.57pts, over 14pts higher than the end of season total which suggests there was a rough little spell towards the end there. Indeed, a quick check shows that April ’11 saw a loss of nearly 15pts which pretty much explains that one. Despite that, a bank of 25 to 30pts ought to be sufficient I’d say.

Another continuing trend is that of a decreasing total number of bets – System 8 has far fewer than System 7 which was well down on System 6. Over the past five seasons there have been 830 bets on System 8 of which 417 were winners for a strike rate of 50.24%, the highest of the systems I have looked at so far. The ROI is also the highest of the three systems analysed, registering 23.24% with a total profit of 192.93pts. However, the familiar trend of a below average season in 10-11 is evident once more. The SR drops of to 44.38% for last season with 14.34pts profit coming at a return on investment of 8.96%. Admittedly that rate of return is better than for the previous systems in the same season but it is still below average. The Championship performance is weaker than the other divisions once more with the SPL not too far behind but this time there are no concerns over the monthly figures.

An overall strike rate of 50.24% suggests a significant edge when one considers the average winning odds are 2.45 giving a breakeven SR of 40.77%. It’s true once more that this edge was much smaller last season with the lower SR but even last season’s figure of 44.38% allows plenty of wiggle room. The recorded LLR is 11 consecutive losing bets which is actually slightly longer than the ELLR of 10 so followers of this system have already had a taste of how bad things can get and should be prepared. There was a peak in the total profit at 202.84pts which occurred back in mid-March so the system had a bit of a rocky end to the season. As with System 7 it was a tricky April that did the damage. Given the greater SR and the impact that has on the ELLR a bank of perhaps just 20pts would suffice but personally I would look to go with at least 25pts if not 30pts.

Attention now shifts to System 21, a much busier system than the previous one with a total of 2455 bets, 1148 winners (46.76%) and a profit of 477.68pts (19.46%). This system fits right into the same pattern we have seen previously though and it should come as no surprise that the figures from the early seasons are much better than those for last season. In the last campaign the SR was down over 6% on the average at 40.70% with the ROI right down at 4.69% with a total profit of only 22.92pts. The SR, profit and ROI have decreased with every passing season and one must wonder if this trend will continue. Interestingly though none of the individual divisions stand out on the figures particularly. There is some variation in performance level across the leagues, obviously, but none of this seems significant. The same can be said for the monthly figures with no time of the year outshining any other.

The average odds of the winners on System 8 are 2.55 so with a 39.15% breakeven SR there is the usual edge we have already seen with the other systems although this time that edge is pretty slim for the 10-11 season. And the figures suggest there are worse times ahead. To date there have been losing runs no longer than 10 consecutive bets but the expected longest losing run comes out at 13. This, combined with the poor performance last season (January, February and April were all losing months), is a bit of a concern. Indeed, the profit peaked in early January (at 491.02pts) and has slipped back around 14pts since. That’s perhaps not that worrying but who knows if the previous trends will continue and next season will be even worse? Based on the ELLR a bank of 30pts could be enough but given what we have seen here I would be tempted to revise that figure upwards.

The last of the single-number systems to come under scrutiny is System 22 which has seen 1219 bets over the past five years of which 586 have been winners (48.07%) resulting in a total profit of 323.57pts which came at a rate of 26.54%. Once again though last season was poor in comparison to what had gone before. The first couple of seasons saw ROIs up around 40% but this was probably unsustainable anyway. Last season’s ROI was 8.89%, which in isolation would be respectable but is poor compared to the five-year average and previous seasons, and the strike rate was way down at just 40.82%. There are no divisional concerns at all this time and the profits are pretty evenly spread throughout the season but that tendency for the latest season to be the worst of the bunch is a little off-putting.

I have no problems with the edge this system seems to have with a breakeven SR of 37.99% due to average win odds of 2.63 but obviously the edge is reduced in the latest season with the SR so low. Followers of this one have experienced the worst of it too as the LLR and ELLR both come out as 11 losers on the spin which would suggest a bank in the region of 25 to 30pts. The total profit reaches a high of 339.70pts in mid-January of this year but it’s pretty much all downhill from there. In fact last season was very much a season of two halves, more so than I have observed for any of the other systems. It was profits all the way until mid-Jan, up 37.91pts at one stage but the rest of the season saw a loss of 16.13pts with losses recorded in January, February, March and April.

Let’s now have a look and see if this idea of using the crossover between individual systems works, starting with the first of the combo systems, System 6-21. Given that both constituent systems struggled in the most recent campaign compared to previous years it is only reasonable to expect the same from this system and it doesn’t disappoint. The five-year figures show 1546 bets of which 737 were winners – a strike rate of 47.67% – generating a total of 365.51pts of profit at a rate of 23.64% but last season that SR was down as low as 41.61% with a ROI of 9.08%. Once again these figures could be acceptable had previous years not set the bar so high. There are double-figure ROIs in each of the divisions covered so this combo effort has addressed the potential Championship defect seen System 6. With the exception of below-average figures for October – where the ROI drops off to under 5% – the profits are spread right throughout the season.

The decent overall SR allows a good margin over the breakeven SR which is 38.56% thanks to average odds on the winning bets of 2.59. Results to date have been relatively blessed though as the longest losing run has only reached nine in a row whereas the sums reckon on an expected longest losing run of 12 bets. The advised bank figure, based on that ELLR, is right in line with the other systems coming out at around 25 to 30pts. The peak in the total profit came at 370.81pts back in January of this year and things have been a bit flat since then. There were three losing months in the second half of the season but fortunately they were all small losses (-1.21pts in January, -4.35pts in February and -2.10pts in April).

Next under the microscope is System 6-22, the overlap between System 6 and System 22, fairly obviously. The intersection between these two systems is smaller than we saw previously with 944 bets in the past five seasons. In that lot there were 457 winners for a strike rate of 48.41% and a profit of 283.93pts at a healthy rate of 30.08%. The decreasing SR and ROI is evident once more but I am inclined to more or less forgive that given last season’s profit of 25.26pts at a return on investment of 13.66%. Even a further dip next season would probably still result in a nice profit. That’s not to say it’s not a worrying trend, of course, just that it is slightly less concerning than for previous systems. Once again the Championship difficulty seems to have been addressed as the profits are fairly evenly spread across the divisions and throughout the year.

The edge is slightly bigger for this system than for the previous system with a breakeven SR of just 37.22%, the lowest of any of the systems analysed so far. In terms of losing runs things could get a little worse as the ELLR comes out at 11, one greater than the longest losing run experienced to date but the figures are pretty similar so the data should give a reasonable idea of what lies ahead. As with previous systems the peak profits came midway through January, reaching 296.57pts, and things dipped in the second-half of last season.

The spotlight shines on System 7-21 now, a system that picked out 958 bets of which 483 were winners for a strike rate of 50.42% and a total profit of 244.45pts at a rate of 25.52%. And finally the spell is broken – the 2010-11 figures are right up there with previous years. OK, so the first season was a stormer with a ROI of 38.11% but since then things have calmed down a little and last year’s 19.39% ROI is almost identical to the figures from 07-08 and 09-10. The strike rate from last year is down a little on previous years at 46.28% but it’s not a world away from 07-08’s 47.03%. When the figures are viewed by division rather than by season a slight question mark appears over the SPL. There are far fewer bets in that division than in the other leagues, the strike rate is significantly lower and the bets seem less profitable with a ROI of just 6.44%. It may be nothing to worry about, the effects of a relatively small sample, but I felt it worth noting all the same. In terms of monthly profits, across the years February and April exhibit a dip in performance but again I don’t think it need overly concern us at this stage.

Once again the system seems to enjoy a sizeable edge, the overall SR being well in excess of the 40.17% breakeven SR dictated by the average winning odds of 2.49. There are longer losing runs on the horizon though as the longest to date has been seven but the figures suggest an expected longest losing run of 10 in a row. Given those figures one could reasonably employ a bank of around 25pts I’d say. For this system the profit peak, 250.77pts, is only a few points above the current figure and occurred in mid-March, a couple of months after most previous systems peaked. That suggests the second half of the season wasn’t so bad for Systems 7-21 and a look at the profits by month backs up that view. Losses were recorded in February and April but the magnitude of each was no worse than a typical winning month so in effect they both just cancelled out a winning month.

System 7-22 is one of the quieter systems on offer providing 667 bets over the past five seasons. Of those bets the system picked the winner on 336 occasions for a strike rate of 50.37% and made a profit of 224.14pts for a 33.60% return on investment. Unfortunately the all-too-familiar trend of the seasonal ROI dropping on a year-on-year basis is back again but despite that last season’s profit of 32.65pts was generated at a rate 24.92%. This may be nearly 20% down on the 2005-06 figure but it’s still bloody good. The SR is low again for the 10-11 season, the 45.80% recorded being the worst seasonal performance of the seasons I have data for, and it’s around 5% down on the overall average but that seems to be the way of things. When viewed by division it becomes apparent that profits are spread very equally throughout the leagues and the story is similar for the year-round split although February raises a very slight cause for concern.

The winners have come at average odds of 2.65 for a breakeven SR of 37.70%, a long way short of the overall system SR indicating a solid advantage here. Once again the expected longest losing run is a couple of notches higher of the worst run experienced to date. The ELLR comes out at 10 but we are yet to see a run longer than eight losers in a row on this one. Once again though a bank of somewhere around 25pts would seem sensible based on these figures. The total profit reaches a maximum of 234.43pts in mid-March, again suggesting a calmer second half of the season compared to the earlier systems. In fact System 7-22 underwent a similar season to System 7-21 in that respect.

We’re coming towards the end of what’s on offer now and second-to-last we have System 8-21. The past five seasons have seen a total of 325 winners from 636 bets for a strike rate of 51.10% and 168.42pts coming at a return on investment of 26.48%. The seasonal strike rate for 2010-11 is again well down on the average at 46.09% but isn’t a million miles away from the previous worst figure of 47.92% in 2007-08. It’s still the worst of the bunch though so it’s not all good news. The ROI for last season is also significantly lower than in previous years at 15.91% but the previous trend of an annual drop-off is not apparent here with a more random variation in the rates of return. As we have seen with some of the previous systems a question mark hangs over the SPL figures which actually show a slight loss but I have no real concerns other than that as the profits seem pretty well spread throughout the campaign.

The difference between the actual strike rate of 51.10% and the breakeven SR is around 10% again as it has been for previous systems I have analysed. The average winning odds are 2.48 resulting in a breakeven SR of 40.40%. Unlike the last few systems I have looked at though this one has seen the worst run out already. There has been a run of 10 losing bets in a row and that’s exactly what the ELLR predicts should happen. The peak profit is around 10pts higher than the current figure at 178.67pts, a value that was recorded in January but was almost equalled again in March so it seems again that we have another relatively stable second-half to the season for one of the systems. A closer look at the figures shows February was not as bad for this system as for some of the others but April was worse with a loss of more than 12pts from just 20 bets. Swings and roundabouts though I guess.

Finally we come to System 8-22 which has seen the fewest bets over the years of all the systems, just 457 of which 237 (51.86%) were winners creating a profit of 160.47pts at 35.11% return on investment. Like the previous system the seasonal strike rate was lowest of all in the most recent season, as was the ROI, but the variation in those figures shows no clear pattern and is more random than we have seen before. Also in line with the previous system the divisional performance figures are excellent for all divisions other than the SPL which does little better than breakeven really. A similar question mark can be placed over the figures for April as the bets in that month show a loss across the years, albeit only a slight one (-1.92pts, -3.92%).

The average winning odds for System 8-22 are a touch higher than for System 8-21 at 2.61 which gives a lower breakeven strike rate of 38.38%. This compares very favourably to the overall SR and one can expect to see a solid advantage from this system. The longest losing run to date has been eight which is one notch lower than the ELLR figure of nine. Similarly to previous analyses the peak profit is around 10pts above the current figure, standing at 171.22pts which was achieved in January and all but equalled in March. In fact this system seems to have operated almost in parallel with System 8-21. As with that one February was a steady losing month while April was a proper stinker, a loss of all but one of the 14 bets struck for a loss of 12.30pts.

Conclusions
With 11 established systems to analyse there is a lot to take in. Fortunately there is a general template that nearly all those 11 systems seem to follow and that’s one that includes a relatively poor season (in comparison with what has gone before) when the system is opened up to live trialling. We’ve seen that year on year decrease in the ROI and SR in the majority of the systems analysed and one can’t help but wonder what the cause of that is and whether it is set to continue. Why should there be such a drop-off in strike rate? Was last season an odd season results-wise (other analysis I have done on football data suggests otherwise)? Perhaps the systems aren’t able to live up to the billing afforded them by the backtested results. Maybe there is some other reason but in all likelihood we’ll never really know.

It’s an important point to raise because at this point we don’t know if this trend is an ongoing one. Are last season’s results typical of a live season for these systems and thus next season we should expect more of the same? Or next season could we expect a further drop away in performance? If so, it raises doubts about whether one should be following some of these systems. On the other hand though, despite the annual drop-off in performance, several of these systems have shown excellent profits in their live trials and that has to be respected. And of course if performance were to return to previous levels the profits would be incredible. Future performance can never be known in advance and ultimately there isn’t enough live trial data to go on at this stage to determine how these systems are likely to perform in coming years.

Graeme had mentioned previously that last season was one of two distinct halves that my analysis agrees with that finding. Lumping all the systems together we see a very profitable run covering September through to the end of the year, the ROI of which was running close to 25%. January is a transition month with only a small profit and then the rains came. A loss of over 60pts in February was cancelled out by March but there wasn’t enough action in May to rub out the 90pts-plus loss registered in April. One wonders what caused these poor months and the same questions around results etc arise again. If we look at SRs over this period the winners were coming along at an average rate (48.01%) until the turn of the year and it was the second half of the season that suppressed the seasonal strike rate.

On that basis, rather than worrying about last season’s figures as a whole I probably need only concern myself with wondering why the second half of the season was poor. As mentioned above, I don’t expect to find an explanation easily but what these figures tell me is that problems with these systems are either something that has only kicked in recently or are temporary. Until I have more live trial data to go on I won’t know for sure but unless the system rules have been changed mid-season (and Graeme assures me they weren’t) then I have to assume the downturn is a temporary one. Overall though the performance figures all include bad runs such as the one observed last season, in which case if I am happy with the numbers – which include 20%+ ROIs don’t forget – then the bad runs shouldn’t unduly worry me.

Given last season was the live test I have to give the figures from that year more weight than those from previous years. On that basis I am happy to rule out System 6, System 7 and System 21 due to the low ROI figures recorded in the 2010-11 season. The returns are better for System 8 and System 22 but are they good enough? I am happy with a reasonably heavy workload when it comes to football betting as I want the volume, or churn as it seems to be termed these days, but I can still afford to pick and choose the systems I follow. The six combo systems give better returns than their constituent parts and would give me around 1000 bets a season, which is plenty, so System 8 and System 22 also slip by the wayside.

Do I want to follow all six combo systems though? Looking at it again the discrepancy between the 2010-11 figures for System 6-21 and previous seasons is pretty large which puts me on the back foot. But the returns from that relatively poor season are still pretty good so I am tempted once more. The drop-off for System 6-22 and System 8-21 is less pronounced so I am happy enough with those I think. System 7-21, System 7-22 and System 8-22 don’t exhibit this same drop-off tendency and offer excellent returns so all make the grade. It then seems a little daft not to take on System 6-21 but to follow all the other combo systems. After all, System 6-21 still made a nice profit last year and more of that this coming season would be perfectly acceptable.

So there we have it, more or less. From the 11 established systems available I plan to follow the six combo systems with each tracked separately in my spreadsheet and each utilising a separate bank of around 30pts (exact figures to be confirmed). I will be on at low stakes as I look to build up more live test data before perhaps eliminating the weakest of the systems for 2012-13 and perhaps replacing them with other systems that are due to be trialled this season. I am going to keep it simple and stick to level stakes throughout with equal sums going on the home and away selections with no draw cover. That’s the staking I used when analysing the data and that’s what these findings are all based around so it seems only right that I stick with that staking. I saw recently that Graeme is using a different staking strategy but I won’t be following suit even if he is the expert on these systems. It’s my bankroll and I’ll play by my risk management rules. I also can’t verify the odds for any of these bets to date so the profits have to be taken on trust somewhat and that’s another reason for me to keep the stakes down.

In short (ha!): there’s a lot to like about what’s on offer and I have cherry-picked what I feel are the best systems but as independent verification of the odds and profits is not possible I will only be on to low stakes.

Categories: Football, Reviews

Another look at the Simplex systems

July 21, 2011 1 comment

Introduction
This weekend sees the start of the Scottish Premier League and it’s still only July. Madness! Everyone knows the football season starts in August and ends in May. OK, there may be exceptional circumstances that necessitate a slightly early finish, towards the end of April in a World Cup year maybe, but a July start? I presume it’s to try and get as much play as possible in before the bad weather hits in December and January but it still seems odd to me, not least because I don’t really get back into the football vibe until August most years. I’ve known for weeks that the SPL starts this coming weekend but it’s only now, with a couple of days to go, that I have got my arse into gear to ensure I am ready to go come Saturday afternoon.

Admittedly I don’t have a lot to prepare for a new SPL season as the only betting interest I have is my SPL Simplex system but if I am prepping that for the coming season then I might as well go the whole hog and sort out the rest of the Simplex systems in my portfolio at the same time. That means making sure I have the right spreadsheets in place to make it easy to identify the bets as they occur and also to keep track of stakes throughout the season. And talking of stakes, I need to confirm I have the appropriate staking levels set as I can’t really change the value of each point during the season due to the way the staking plan works. I need to set that before the season starts and stick with it throughout the campaign.

I have a staking progression plan in mind with each of the Simplex systems taking another step up the staking ladder each season, the value of each point increasing each time. This allows me to build confidence in the systems while the stakes are low so by the time the stakes get serious I have experienced the highs and lows the system has to offer and am more prepared. This method also takes advantage of the organic growth that I am so fond of. Right now I want to check that each system has a suitable bank for the next staking level without the need for further leveraging (there is more on leveraging to come, I wrote up more of my thoughts on it recently but I am aware that I have owed you this post for some time now and worry that it may be overhyped) and also what sort of maximum stake I could be opening myself up for.

Last season things got scary on the Prem Simplex system. Wolves went on an extended run that saw the stakes on them reach nearly 5000pts. I don’t care what value you assign to a point, when the stakes get up that high things get worrying. Fortunately stakes that high are not the norm but they still occur every few years it seems. With that in mind I am cautious about stepping the value of one point up too high. While most stakes will be nowhere near those levels, in the tens or perhaps hundreds of points at most, I could do without the stress that would inevitably come with stakes like I had on Wolves.

When Wolves were on that scary run I looked at some of the maths involved. I have since thought more about the maths behind such runs and have some new figures to bring to the party. What I really needed to know then was how long I could expect a losing run to go on for as that will tell me how high the stakes are likely to get. Fortunately it’s a relatively easy calculation so I will be looking at that for all three systems.

While I have the systems up on the ramps, as it were, I will also be giving them a quick once-over to ensure that the variables they use are all correctly optimised. This is all ‘under the hood’ stuff really but it’s worthwhile giving it a quick check to make sure it’s not likely to blow up a few months down the line.

Losing Runs
I’m not going to discuss the calculation used here in any detail. There’s no need, all that matters is the result really. I will give the equation used but I won’t be going into how it is derived or why it is as it is. The equation is:

expected longest losing run = ln(num matches) / -ln(1 – strike rate)

That ln() business is the natural logarithm of the quantity in brackets. You don’t need to understand it to use it as most scientific calculators can work it out for you as can programs like Excel but in case you do want to know more you should probably read this. When using that formula above you want your strike rate as a decimal value rather than a percentage, so 0.5 instead of 50% and so on. Note also that the result won’t necessarily be an integer value so you may wish to round it. After all, a losing run of 13.46 bets is pretty nonsensical isn’t it? How can 0.46 of a bet be a loser? Personally I always round up, so 13.46 would become 14, rather than using the traditional rounding methods but you can do whatever suits you.

There are two important parameters used in the losing runs calculation: the number of matches played and the strike rate, i.e. the number of matches that ended with the desired result. There are several ways in which one can obtain the number of matches and strike rate depending on what data sample one uses. For example, should I use the data from my own betting records? That will give me a larger sample for the Premiership than for the SPL and Serie A as I have been betting on that division longer. I only have one season’s worth of Serie A data so calculations based on that sample may not be terribly accurate. Should I therefore use the full set of matches from the last few seasons? If so, how many seasons? It would be interesting to see how the expected longest losing run varies in length depending on the number of seasons included in the calculation. With all that in mind I have decided to examine my own records along with full records from the last 10 seasons, 5 seasons, 3 seasons and last season plus limited data sets including only those matches that meet my Simplex criteria for the same timeframes.

Premiership
I had wanted to display a table of values here showing the sample sizes and strike rates used in the calculations but I tend to have a few problems with tables in my blogs so after playing around a bit with the HTML for tables I have given up and gone for plain text instead. It’s not as clear but it serves a purpose.

My record contains just under 500 bets with a strike rate of 27.91% (0.2791) resulting in an expected longest losing run (ELLR) of 19. The figures for the  full 10 season sample produce an ELLR of 28 and from there the value decreases as the size of the sample also decreases. The calculation gives an ELLR of 25 for the past five seasons, 23 for the last three seasons and 18 for last season only. The values are identical when the data set is limited to the Simplex matches only, with the exception of last season’s value which comes out at 16.

These values fit pretty well with my previous analysis which indicated a run of 19 games without a draw is possible but that it is unlikely to go much higher. Sure, the ELLRs from the multi-season samples are higher, being in the twenties, but note also the trend for the ELLR to decrease as only more recent seasons are included. The number of draws seems to be increasing with each season which in turn brings down the size of the expected longest losing run (run without a draw). Including only those matches that meet the Simplex criteria doesn’t change things much. Based on last season I could expect a losing run of 16 and no longer but this is somewhat optimistic as I have already seen longer losing runs than that. I think working on a figure of 19 to 20 seems like the most sensible approach.

SPL
Using my own record of nearly 200 bets and a strike rate of 25.39% I get an expected longest losing run of 18, one less than the Premiership equivalent. Substituting in the figures for the past 10 seasons produces an expected longest losing run of 30 while the figure is 26 for the past five years and 23 for both the last three years and last season on its own. Limiting the data sample to the Simplex matches only leaves the 10-year value unchanged by increases the remainder by two.

I’ll be honest, these figures are bigger than I expected. There was a run of 25 matches without the desired result for one team back in 2003-04 and a couple of years before that one side went on a run of 22 games without one ending all-square but looking at the data these seem like aberrations rather than the norm. However, the sums say I should be prepared for a  possible repeat performance so something in the 20-22 range is probably a sensible figure to take forward.

Serie A
I only have one season’s worth of Serie A bets in my spreadsheets but despite that the sample size is almost the same as the SPL sample despite the latter covering two seasons. Funny how these things happen at times isn’t it? Anyway, that sample gives an ELLR value of 18 which, as with previous divisions, is lower than the values calculated from larger samples. For the samples including all matches the 10-year figure comes out at 25, the five- and three-year figures are both 24 while last season produces an ELLR of 21. Unfortunately my Simplex spreadsheets don’t go back far enough to allow a 10-year Simplex-only calculation but the five-year ELLR is 23, it’s 25 for the last three seasons and 19 for last season. Note that in this last case the general trend of decreasing ELLR figures as fewer seasons are used is bucked. The three-year Simplex-only ELLR figure is greater than the five-year equivalent due to a dip in the number of draws in recent seasons. It remains to be seen whether this trend continues long-term but last season was right back on track so it may have been a temporary blip.

My spreadsheets for this system don’t go back as far as the Premiership and SPL versions but they are still very useful. They show that in 2005-06 there was a run of 20 games without the desired result so, bearing that in mind along with the calculated values above it seems an ELLR of 20 to 22 is appropriate here too.

Staking
What does any of this mean? Why do I need to know the length of the likely longest losing run? The reason is the staking behind these Simplex-style systems. There is an element of loss-chasing although I should emphasise that it is controlled and not the terrible idea that most seem to think it is. Well, it works well under these conditions but that’s not to say I would consider loss-chasing staking plans for other systems. But because of the way the staking works I need some idea of how long the losing runs can get as that determines how high the stakes can get and with that information I can sensibly decide whether or not each system can take the next step up the staking ladder. Obviously the more I stake the greater the potential returns (all else being equal) so while I am keen to maximise my returns I need to balance this against the risk of overstaking.

The figures I calculated above are very useful when it comes to sorting out my staking but they are all estimates really. The calculations assume that all teams are equally likely to draw and while this is not a strictly valid assumption one has to make a few simplifications when modelling real world situations. They also don’t tell me how often I could expect a losing run of this magnitude but do allow me to ensure my staking doesn’t get out of hand by indicating what sort of max stakes I could reasonably expect to have to put into action. And that’s how I plan to use these figures.

While I haven’t shown any of my workings here I have also given each system an MOT. I have taken each of them apart, checked all the main variables, polished them and put it all back together again. In doing so I was able to use the key variables to calculate the largest stake I should expect for each system based on the ELLRs above. It transpires that I should allow for stakes to reach around 8000pts for the Prem Simplex system, something close to 1200pts for the SPL and when betting on Serie A matches I should be prepared for my stakes to reach figures of around 3500pts.

They are all non-trivial sums whatever value I set for each point but at least I know now what to expect if (when) the worst-case scenarios come into play. The question now then is knowing those figures am I prepared to increase the stakes on each system?

The Prem Simplex system is the oldest of the three I currently run and each year I have increased the stakes. Previously I have done so without first calculating how bad things could get but I wasn’t going to make that mistake again after Wolves’ efforts last season. And it’s only that run that’s really got me thinking here. Without that I’d be upping stakes like a shot as the system is showing a ROI in excess of 50% according to my records and is simple to follow. I think making the increase is the right move but I am going to think it over a little more before I decide for certain. After all I have a few weeks until the Premiership starts so for this system there is no real rush.

The same is not true of the SPL system of course. I need to get the stakes for that one nailed in the next day or so. Looking over the various spreadsheets it seems like a safe move to take the next step up the ladder, especially given the expected longest losing run would only require a max stake of 1200pts or thereabouts. I can handle that, yeah. I could be tempted to jump a rung on that staking ladder given the way this system has gone in recent seasons but that may be pushing it. The longer losing runs came a few years back now which could mean that draws are getting more common so losing runs are naturally shorter or it could mean the stage is set for another big losing run. I just don’t know which it is so I’ll play it safe and just take one step up the ladder.

I only have one season of Serie A Simplex under my belt but it was a promising start and I feel that more success is on the cards. Given that this is the youngest of the systems it has the smallest stakes having started last season at on the bottom rung of my staking ladder but this season it looks well set to step up. In the worst case scenario the stakes will be great but still not as scary as last season’s Premiership so I am sure I can cope.

Categories: Football, Reviews

Review – Football Elite Shortlist

June 11, 2011 2 comments

Overview
The Football Elite service is one I have been with for a couple of years or so now and I am very happy with it. Their tips are split into two different types: the main recommended bets (often called account bets) and shortlist bets. The main bets get a full write-up illustrating why Matt thinks the bet offers value whereas the shortlist bets are just given as Team A to beat Team B along with the prices for the main markets (win, draw no bet, double chance, Asian handicap). Matt also stated throughout this season that there is not enough data yet to prove whether the shortlist bets are profitable in the long run. There are now two seasons worth of data for the shortlist bets so Matt has taken a view on the value of these bets now, something I will come to after first looking over my own figures from this season.

Detailed Analysis
As I mentioned earlier, the Football Elite Shortlist bets come with advised prices across several different markets. Prior to the start of the 2010/11 season I decided to follow the Football Elite Shortlist bets on the draw no bet (DNB) market. This seemed to be the best option for me based on the previous season’s data. In 2009/10 the shortlist bets made a profit of 15.80pts betting draw no bet at a ROI of 12.20%. This was not the most profitable option as the double chance bets finished the season with 18.61pts profit (14.40% ROI) but on the basis that more firms offered DNB markets, including Betfair, I went down that route. I could have constructed my own double chance bets but I took the easy option in exchange for a small amount of profit.

I didn’t get on all the Shortlist bets this season due to my unavailability for betting at a couple of points during the season so my figures may differ a little from the official figures. And of course there is no SP for football bets so my figures are based on the odds at which I struck my bets and include my commission rate at the time of settlement where the bets were placed on Betfair. I placed 134 bets of which 48 were winners and a further 38 were draws which returned stakes. Overall a loss of 5.92pts was recorded at a return on investment of -4.43%. The longest losing run spanned nine bets while the longest winning run was five consecutive winners.

Matt’s official figures for this season’s Football Elite Shortlist bets are 152 bets of which 57 were winners and a further 39 were drawn matches which resulted in a loss 4.98pts (-3.2% ROI) on the DNB markets. As you can see from that I missed a few bets and as a result cost myself a few points profit but my figures were more or less in line with the official standings. That’s good to know.

Why are these bets shortlist bets and not main account bets? There could be several reasons but they probably include factors such as having less confidence in the bet, any stats/trends used to identify likely bets may be less reliable than for the main bets or the league as a whole may be less reliable as a betting medium. Whatever the reasons Matt has rightly identified the shortlist bets are not as strong as his main recommendations. That doesn’t mean the bets don’t serve a purpose though. I’m sure considerable amounts of effort are put into selecting each and every tip that forms part of the Football Elite service whether it’s an account bet or a shortlist bet. None of these bets would be sent out to members unless it was felt they had merit. So what is Matt’s take on the Football Elite Shortlist bets?

In his end of season review Matt identified the fact that the shortlist bets are best followed as win only bets. That’s the only market that has made a profit in both of the seasons for which there is data. It’s not a very big profit though and Matt is quick to accept that. In fact he goes on to say that the shortlist bets are probably best used to smooth out the ride somewhat. For example, during a horror spell in January for the main bets the Shortlist made a nice profit that would have softened the blow. To my mind it seems the Football Elite Shortlist bets are more or less akin to the ProBandit NQ bets – included in the service as something of an optional extra to protect against some of the more severe losing runs the service may experience but not necessarily significantly profitable in their own right.

It seems that next season the Football Elite service will continue with the Shortlist bets separate from the recommended bets and that they will be advised as win only. I think this is a sensible move allowing members to follow or ignore the Shortlist bets as they see fit. Personally I shall be following the recommended bets only. My focus at this time is on systems and services that can generate a solid return. I get that from the main Football Elite bets but not from the Shortlist selections. Matt rightly identifies a number of situations where one may want to use the Shortlist service to increase turnover with a small risk of making a loss but none of those currently apply to me. I am looking to increase my overall ROI rather than increasing turnover so the Football Elite Shortlist bets don’t fit into my plan and will be dropped. I must emphasise though that this decision doesn’t reflect badly on the quality of the Football Elite service as I will still be following the main recommendations, it is only the weaker Shortlist bets that are being dropped.

Categories: Football, Reviews

Review – Football Betting Data

May 28, 2011 1 comment

I have a good idea what the conclusions coming out of this review will be before I even get started but I want to be fair and give the service a fair chance so I will review the figures as I would for any other system/service.

Overview
Football Betting Data (FBD) is Oxfordshire Press’ football betting service concentrating solely on the Premiership. It’s not really a tipster service as such, in fact it’s a little tricky to describe exactly what this service offers subscribers. There is a web-based database containing match stats for all Premiership teams and members are encouraged to use this database to study head-to-head match-ups in order to identify potential bets. Matt Nesbitt (the face of many of Oxon Press services) emails out on a weekly basis with his views on coming matches including a number of recommended bets with advised odds. But these are not really tips as no staking plan is advised and as far as I am aware the performance of these recommendations is not officially monitored. The cynics amongst us may have already spotted that this allows the service to crow about significant profits when they happen but sweep losses under the ‘not a tipping service’ carpet when it suits.

Many of the recommended bets are in relatively low liquidity markets. Corner handicaps, team to receive first card etc. Yes, there are some straight win bets and some under/over 2.5 goals bets where the liquidity will be much greater and prices will generally hold up on such bets but for the minor markets the price has often gone even if you’re pretty quick to get your bets on. The advice doesn’t come with a minimum price either (a significant flaw in my book) so it’s not always clear when you should chase a price down and when to just let it go. If the recommended bets are based on statistical trends then surely a minimum value price drops out naturally. I asked Oxon Press about this and got told they were not a tipping service so didn’t advise minimum odds. Convenient.

In addition to the recommended bets, if you read between the lines there are sometimes other bets that have been identified but don’t carry a full recommendation. Again, the cynics may say this is an ideal way of boasting about profits but ignoring losses and I am inclined to agree. This review focuses on the recommended bets rather than the ‘read between the lines’ advice.

Detailed Analysis
FBD has been part of my portfolio since the start of the 2008/09 season so three seasons in all. A loss has been recorded in each of those seasons so we’re not off to a very promising start. In 2008/09 I lost 8.26pts, 2009/10 was nearly breakeven with a slight loss of 0.13pts while I was down 15.42pts in the season just gone – a total loss of 23.81pts.

Those figures include a number of ante-post bets, which I will come to later, but for now I want to focus on the recommended bets on individual matches. My records show 430 such bets over that period with 215 winners for a strike rate of a nice round 50%. Those bets have contributed 17.51pts of the total loss which is equivalent to a ROI of just over -4%. Not great by any means.

I also have the figures broken down by year. In this case the 2008 figures only include the first half of one season whereas the 2011 figures include only the second half of a season with 2009 and 2010 containing the equivalent of a full season, albeit made up of two different halves. If we look first at the two complete years we see a fairly consistent level of return. 2009 saw 143 bets generate a total loss of 4.35pts for a ROI of -3.31% while the following year saw an overall loss of 6.88pts from 190 bets at a ROI of -3.88%. The losses in the part-years were greater (-6.04% ROI in 2008 and -9.43% in 2011) but the sample sizes are smaller so these differences may not be significant. There has been very little deviation in the strike rate throughout this period so on that basis one can assume the service has been consistent but at the same time a losing proposition.

The SR only tells part of the story. To get an idea of whether this service enjoys any sort of an edge one must also look at the average odds of bets struck. It transpires this is a tad less than evens at 1.98. If half of all bets placed at such odds were to win then one would suffer a slow but steady loss. The strike rate is too low for the average odds. Even more so when one considers the average winning odds which are 1.89 or slightly under 10/11. That means the SR is a few percent too low to turn any sort of profit. There is no edge, quite the reverse in fact.

With football services it is often worth analysing a monthly breakdown of profits to see how performance changes throughout the season. It is not unreasonable to expect performance to improve once there is more form in the book, so to speak. But there are also end of season issues with teams in mid-table not giving their all and so on. Therefore one could perhaps expect October to March to provide the bulk of profits. That’s not the case with the Football Betting Data figures. In fact there are no clear trends to the monthly breakdown. There is no obvious pattern to the SR or profits. August (unexpectedly), December and April are the only profitable months and all have better than average SRs but there is no obvious explanation why those months should be profitable and other months not.

As I mentioned earlier there have also been a few suggested ante-post bets. My records only really show these for the 2010/11 season so I’m not sure if they are a relatively new part of the service or whether I have missed them in previous years. The bets are all top goalscorer bets although they cover a number of different sub-categories such as club top scorer (e.g. Fernando Torres to be top Liverpool goal scorer) and top midfield scorer. I think (but can’t be sure) that these bets all came out of a single pre-season report looking specifically at Golden Boot bets. Anyway, the bets were rubbish! OK, it didn’t help that two of the club top scorer bets were players who moved clubs in the January transfer window (Torres and Darren Bent). A lot of faith was placed in Jermain Defoe (taken to be top scorer at Tottenham and also on the Premiership golden boot handicap) who simply didn’t deliver and Steven Gerrard who was injured for a lot of the season. Perhaps then a heavy slice of bad luck scuppered the bets. Perhaps, but the results tell the tale and heavy losses were incurred. My records show an outlay of 8.5pts and not a single winner. A small number of those bets were placed win only rather than recommended each-way as the advised price was long gone and as stated before the bets come with no minimum price stipulation so there were questions as to whether I should have even placed these bets at the lower price. Had I staked as advised I would actually have lost an additional half point.

Conclusions
I don’t think there is much point analysing the figures any further. The outcome of this review is fairly clear now. The recommended bets supplied as part of the weekly emails have no edge. They lose on a consistent basis albeit as a fairly slow drip and no matter whether you break the figures down by season, year or month there are no signs that a decent profit can be made. As such I don’t really see a reason to keep following the recommended bets – no point continuing to throw money away just to have an interest on a few Premiership matches.

However, I won’t be dropping the service completely. As I have membership to FBD as part of the X Circle deal from a few years back it is not costing me anything to stay signed up with FBD so I could still make use of their database and analyse head-to-head match-ups as and when I see fit (and find time) to do so. You never know, I might be able to spot a few angles that way and make back some of the losses incurred from the recommended bets over the years.

In portfolio terms I will be removing FBD from my portfolio breakdown and discontinuing it on my spreadsheet with immediate effect. If in the future I identify any suitable betting opportunities through the FBD database they will be recorded as being my own tips. That may seem slightly unfair but then I am just using FBD as I would any other data source, and not as a tipster/advisory service.

Categories: Football, Reviews

Review – Mark Johnston

May 26, 2011 Leave a comment

It seems that whenever I return from holiday I approach my gambling with a renewed enthusiasm and a real sense of purpose. The other day I said I was happy with my portfolio and while that is true at a high level there are potentially a few systems and services that should be dropped. I have on my Watch List page a list of those aspects of my portfolio I wish to review so it’s about time I worked through at least some of those reviews to give my portfolio a bit more of a polish. And I’m going to start with the Mark Johnston system.

Overview
The Mark Johnston system is one I picked up from a Win2Win eBook several years ago. It concerns Mark Johnston runners (duh!) during the bulk of the flat season and focuses on the tracks where the trainer has a particularly good record. The system has a fairly low strike rate, around 15% or so, but has picked out some of my biggest winners over the years. Over the last couple of years it feels as though the system has lost some of its edge. Also my mentality may have changed slightly as I am no longer as comfortable with the long losing runs I associate with this system. Perhaps that comes with going full-time and feeling some of the pressure to deliver consistent profits where possible. Whatever the reason, I no longer feel I am getting the best from this system so it’s time to have a look at the figures in more depth.

Detailed Analysis
Ultimately the system has proved profitable for me, earning 87.25pts from 859 bets. My staking has been all over the place with this one over the years as for a long period it threw the balance of my portfolio completely out the window, returning significant profits and swelling the system bank when other systems were struggling to the extent that this system dominated my results. I reduced the stakes I used for this system a few times to build more security into my bank but also to reduce the degree to which my results were dependent on Mark Johnston. As a result my cash ROI is lower than the 10% I could have earned by placing 1pt level stakes on all bets.

If we look at how the profits have fallen over the four years I have been running this system a very interesting pattern emerges. 2007 and 2009 were hugely profitable (+50.47pts and +47.66pts) respectively with small losses shown in the even years (-7.44pts in 2008, -3.45pts in 2010). Going by that very small sample one may expect a bumper 2011.

Last year saw a massive increase in the number of bets struck, the figure almost double the 2007 number. 176 bets in 2007, 134 in 2008, 208 in 2009 but a massive 341 in 2010. Why? Has Mark Johnston expanded his operation significantly or changed his modus operandi and started throwing more runners at races? I don’t follow racing enough to know the answers to these questions. Sure, my holidays and whether or not I continued to bet while I was away could be a factor but surely not the major factor here as we’re only talking about a couple of weeks a year really. I am concerned that if this trend continues I could be set for a relatively large number of bets on this system during 2011 and I’m not sure I want to place that many bets on a system I don’t believe in 100% (which I don’t really if I am conducting this review).

The monthly split of the profits is also very interesting. The system only runs from May to November so there are only seven months of the year to consider and the first and last of those are relatively quiet. Even so, one wouldn’t really expect July and August to dominate as much as they do. Profits of 109.58pts in July and a further 42.94pts in August. October is a big losing month (-43.66pts) with the other months relatively breakeven but mostly slight losers.

The obvious conclusion to draw here is that the May to November time span is too broad and it should be limited to just July and August but is that fair? A profit of 152.52pts from just 332 bets certainly seems attractive but there is a considerable element of back-fitting there. It kinda makes sense as one could reasonably expect a big stable such as Mark Johnston’s to peak in the middle of the season and take advantage of the big prize money on offer around that time. But without a detailed database of horse racing results with which to check whether backing his runners in just those months would be profitable. I have had a quick look at the Racing Post site and their database and it is inconclusive, although the functionality is rather limited. I could get on HRB or similar and have a proper look but my instincts are telling me it’s not really worth it. It’s unfortunate but I am just not quite getting the right feeling from this system and already I am starting to doubt it’s continued inclusion in my portfolio. But let’s finish the review to be fair.

I said earlier I felt the system has lost its edge in recent years so let’s take a look at that angle. There were four winning months (May, July and August) and three losers last year. The system had a good start to the year, a dip followed by the traditionally strong summer peak and then tailed off. That may well go some way to explaining why I feel the system had lost it’s edge – it had a tough end to the year and the memory of losing has stuck in my head. In the last few months of last year the system hit several losing runs in excess of 20, including one of 28. Such runs are to be expected based on the SR but when they come towards the end of the season (the system ended the year with 20 consecutive losers) and you know there will be no more selections for six months it can be hard to take. There’s nothing to say these runs can’t happen at any time but the limited timeframe in which this system operates doesn’t help matters. It can feel like a long winter coming off the back of a long string of losers.

On the face of it last year was nothing unusual. OK, the number of bets was larger than expected but the SR was in the right ballpark, the longest losing runs were in keeping with previous years as were the average odds and various other key indicators of performance. I can’t actually find clear evidence that the system is losing/has lost its edge after all.

Conclusions
If the system doesn’t appear to have lost its edge and is performing ‘normally’ then where does that leave me? I came into this review thinking I could look for evidence of an eroded advantage and drop the system on that basis but that’s not how it has transpired. I suppose it then boils down to how this system as a whole fits into my current betting style.

Generally speaking I prefer consistent, steady profits. That’s why I am attracted to lay betting although I recognise the greater ROI associated with backing horses. Steady profits come from a steady stream of winners so we’re not talking about longshot low SR systems, and that’s what this system is really. On that basis then this system isn’t one I would naturally choose were I constructing a gambling portfolio from scratch. Then again I have spoken at length about the need for diversity, both in the sports covered as well as the types of services and systems that form one’s portfolio. I am genuinely torn over this one.

OK, here goes then. If I had access to HRB or similar I could deconstruct this system and tinker with the rules – an advantage of systems that I have spoken about before. I could look at the effect of narrowing the timeframe to just July and August. In doing so I would end up with a very limited sample each year but with a view to plucking out just the most profitable bets. But if I do that then I should also look at the course-by-course breakdown of results too to see if the results at any of the individual courses are dragging things down. In short I’d be tearing the system completely apart and almost starting from the beginning, and that’s not a route I currently want to go down. My future direction is football not horse racing, at least in terms of system development. I have my doubts over the Mark Johnston system in its current guise and because of that and the fact I don’t have the time and motivation to give it the attention it needs I am going to drop it from my portfolio with immediate effect.

A tough decision that one but the effects of the long losing runs are burned into my brain and I would rather steer my portfolio towards consistent returns at the expense of the occasional big win.

Categories: Horse Racing, Reviews

Review – ProBandit A List

March 11, 2011 Leave a comment

This is a review I hadn’t really planned until very recently. I had been perfectly happy with how the ProBandit A List service was going until someone on the Secret Betting Club forum questioned the performance of the Near Qualifier (NQ) bets. For those of you who aren’t aware, A List bets fall into two categories: System bets and NQs. No-one was doubting the success of the System bets but were the NQs actually profitable?

Curiosity was starting to get the better of me. Were all the profits due to the System bets? When I reviewed this service for SBC, getting on for a year ago now, I concluded that one should treat the two types of bet the same and back them equally. But now I had more data to work with I figured I should check whether that was that still the case.

Overview
The data set I will be using for this review comes from my own betting records. There are 870 bets (counting each way bets as two separate win and place bets) of which 287 were winners (32.99% SR). These bets generated a total profit of 39.23pts at a ROI of 11.43% to advised stakes.

This data set divides down into two subsets as follows:
System Bets: 680 bets, 248 winners, 36.47% SR, 40.40pts profit, 14.20% ROI
NQ Bets: 190 bets, 39 winners, 20.53% SR, -1.17pts profit, -1.99% ROI

You can see at a glance that the NQ bets are not profitable and have a much lower strike rate than the system bets but this alone isn’t sufficient justification for me to drop the NQs from my portfolio. I want to look deeper into each subset of data and get a better feel for them.

TSM Analysis
In situations like this, where I am splitting one of my data sets, I can’t readily use my normal betting records spreadsheets for the analysis so I turn to The Staking Machine instead. In this case TSM tells me the System bets have average win odds of 3.25 which combined with the SR give an edge of 18.47%. With such a strike rate we can expect a losing run of 14.4 bets and indeed a run of 14 consecutive losers has been recorded in this data set so it seems reasonable to assume we’re dealing with a representative enough set of selections.

Add the NQ bets back into the mix and while the average win odds increase to 3.45 the edge drops to 13.89% due to the lower SR. The full data set contains a streak of 21 consecutive losers, exceeding the statistical estimate of the longest losing run which comes out as 16.9. It just goes to show that sometimes you can experience a downswing worse than anything you thought possible which in turn highlights the importance of good bankroll management, something I will be touching on again later.

The Official Line
With the SBC forum thread gaining more and more interest it was pleasing to see Paul Bell of ProBandit give his view of things. The official word is that the NQs are there to help smooth the profit flow. Longer losing runs would be experienced without the NQs and they themselves are not intended to provide a serious long-term profit. The system bets are the main drive but the NQs just smooth out some of the bumps along the way and it is up to the individual as to whether they follow them or not.

A Smoother Path to Profit?
Paul’s argument sounds convincing and at the time I was happy to go with it but I have started to think more deeply about what he said and to question it slightly. Have a look at the graphs below, for example. Is the blue profit line (representing the profits accumulated by backing all selections) particularly smoother than the profit line for just the System (red) selections? Similarly the graph showing the variation in strike rate suggests that the NQs are doing little to smooth the ride.

We saw earlier from the TSM figures that because of the lower strike rate the expected losing sequences for the full data set are longer than those for just the System bets. And this is demonstrated by the actual runs experienced: 14 consecutive losers for the System bets but a massive 21 if we include the NQs too.

Staking
ProBandit advise a stake along with each tip they provide. Presumably the larger the advised stake the greater the confidence in the bet. Is this advised staking helping though or would one be better off to level stakes? We saw earlier that to advised stakes the total profit was 39.23pts at a return on investment of 11.43%. If my calculations are correct (and I am pretty sure they are), this drops to 7.81% ROI for 1pt win/0.5pt each-way. It seems that the advised staking plan is helping to generate greater profits.

Bankroll Management
Mike & Dan at SBC suggest a bank of 50pts for this service which is probably about right if you’re following all the bets but I think that is overly cautious if you plan to just follow the System qualifiers. In that case a bank of 30pts, which is roughly double the longest losing run if you’re excluding the NQs, should be more than adequate.

In fact you could tighten things up even more if you were feeling brave. While the longest losing run may be 14 System bets this only equates to a drawdown of around 6pts (based on an average stake of 0.42pts) so you could potentially get away with a bank of around 15pts. By the same rationale a bank of around 25pts would suffice for the full set of A List bets. That said, my preference is for a bank somewhere between these two extremes, around 20pts for the System bets and perhaps double that figure if the NQs are included also.

Win or Each-Way?
Finally I want to have a very quick look at win v each-way betting. Would one be better off backing all the selections win only or are the each-way bets contributing as expected? Backing win only would result in a profit of 26.73pts from a total stake of 251.19pts for a ROI of 10.64%. Limit the data to System bets only and you’d have a profit of 27.64pts from 199.81pts staked or a ROI of 13.83%. Both of these figures are slightly down on the ROIs achieved by backing each-way when advised so it seems that once more Paul knows what he is doing when it comes to advising the right type of bet.

Conclusions
I set out to investigate the performance of the ProBandit A List NQ bets and to establish whether or not I should continue to back them as I do the System bets. I quickly established that the NQs were not profitable and ran at a lower strike rate than the main System bets. The smoothing effect that was claimed was not evident either so it is difficult to see what the NQ bets really bring to the party.

I mentioned recently a need to increase my turnover but I also need to increase my profit and the bottom line is more important so while ditching the NQs while reduce my turnover it should increase my long-term profit and also my overall ROI. The peaks and troughs of the System bets are not so extreme that I can’t do without the alleged smoothing effect of the NQs. I will no longer be following the NQ bets.

While I had the data up I also briefly investigated alternative staking plans and betting methods. I found that advised staking significantly outperforms level stakes and that each-way betting, where advised, is profitable. In short one should follow the advice given with each bet.

Disclaimer – This review is based around my own data, including odds, gathered over the past year or so. As such this is very much an unofficial review of the service and all opinions stated are mine and mine alone unless otherwise stated. This review is no way endorsed by ProBandit or Secret Betting Club. I think that’s my arse sufficiently covered.

Categories: Horse Racing, Reviews

Review – First Timer

February 14, 2011 Leave a comment

First Timer is a system that has been part of my portfolio since March 2009. However, recent performances have forced me to look at the system in more depth and question whether or not it should remain part of my portfolio.

An aside
I’ve just spent a couple of hours writing some analysis code within my betting records spreadsheet. The spreadsheet is Excel format with large amounts of VBA behind it, all of which I have written myself. I feel I must be in control of my own records, including their format and presentation, which is why I develop my own spreadsheets rather than rely on the work of others. I know there are a few spreadsheets around that are suitable for recording one’s betting but I need to be in control and do things my way.

The code I have written this morning allows me to take any individual strategy from my portfolio and within 20 seconds I can have a full statistical breakdown encompassing summary stats (bets, winners, SR, profit (£), profit (pts), ROI etc.) plus yearly, monthly and daily breakdowns. I need to do a bit more work to add in the graphing functionality also but what I have now is more than good enough for this review.

I would encourage all gamblers to learn how to make the most of Excel, especially VBA programming. In the right hands Excel can be such a powerful ally and I really couldn’t manage without it.

Overview
The First Timer system selects horses that are enjoying their NH debut and are fancied to do well by the betting markets. On the face of it it sounds like this should produce a reasonable number of qualifiers but it doesn’t. The system has been in my portfolio for nearly two years and in that time there have been just 113 qualifiers. The SR is pretty poor too with just 27 of those runners going on to win their race, a strike rate of 23.89%. My records show a loss of 23.00pts (a ROI of -24.38%) for this system so it is time to take a deeper look at what is going wrong here.

A system with a strike rate of around 25% can expect to have some long losing runs. Something like 24 consecutive losers wouldn’t be out of the ordinary if you crunch the numbers (I’ll try to remember to explain the maths behind this at some point) so the longest losing run of 22 I have recorded is perfectly acceptable, given the SR. Of course, a lot of losers is going to push down the SR and increase the chances of a longer losing run. The greatest number of consecutive winners (longest winning run) is 4.

The loss of 23.00pts comes despite an overlay of 10.48%. That means that the prices I got my bets matched at were, on average, 10.48% better than starting price. This has been achieved through the judicious use of BOG bookies and Betfair. Were I betting to SP my losses on this system would be even greater.

With a strike rate of 23.89% this system needs to select winners at an average price of 4.19 in decimal odds, or a shade over 3/1 in old money. Unfortunately the average winning odds are just 2.41 or a tad under 6/4 if you prefer. If fewer than one in four selections are winning and when they do win the average odds are under 6/4 then disaster will surely follow. The average odds of all selections is 3.60, still significantly below the breakeven odds for this strike rate. There is no edge to this system, in fact the opposite is true.

Detailed Analysis
I don’t have the time or inclination (or access to the right data really) to take this system apart piece by piece and see if it can be rebuilt as a profitable betting strategy but what I can do is take a look at whether there are any time-based trends we can exploit.

I added this system to my portfolio in March 2009 and it enjoyed a quiet but profitable year with five winners from the 13 selections for a profit of 4.51pts. In 2010 there were more selections (64) but the SR dropped as only 18 of those were winners. A loss of 4.96pts was recorded taking away all the profit built up the previous year. And now the system has gone into meltdown with just four winners from 36 selections so far in 2011 for a loss of 22.55pts. OK, we’re dealing with small sample sizes (as one must do so often in racing) but the SR dropping year on year is a bad sign and with so few selections in an average year there seems little opportunity to claw back the losses recorded so far this year. Things are not looking good.

I think it’s clear where this review is heading but I want to quickly take in the monthly and daily figures, just in case. Selections in the summer months may be few and far between but based on that small sample (21 selections in May to July) they are quite profitable. This could just be coincidence and without a lot more data it’s not something I will be concerning myself with. Performance over the winter period (December to February) stinks though. Seven winners from 50 bets for a loss of 28.62pts is a sure indicator of the biggest problem this system has. When the NH season is in its prime this system is recording huge losses. The system seems broken to me.

I tend not to put much stock in breakdowns by day of the week but with many of the major races held on a Saturday it can be interesting to see how things look. The breakdown of runners is fairly even throughout the week but the profit split isn’t with Monday and Tuesday not looking too bad, 7.21pts combined profit from 41 runners. Saturdays and Wednesdays are terrible but we are dealing with statistically insignificant sample sizes.

Verdict: Any edge this system may have had at one time has been well and truly eroded and continuing to follow this service would be madness. There hasn’t been a profitable month since July 2010 so this system will be dropped from my portfolio with immediate effect.

Categories: Horse Racing, Reviews