mathare's gamblog

[youngtag]

Excellent Thread Mathare. It's certainly eye opening to see the rigorous testing and scientific and mathematical approach you take to your gambling.

Unfortunately some of the more advanced portion of your statistical analysis are above my tiny mind's comprehension but I think I get the overall thrust of your analysis.

My question is regarding your testing of the system. Are you saying that you have to go back over games from 10 years ago onwards and find out the best available odds and average odds for each game? Would this not take a small army of monkeys tapping away at keyboards in a room for several years?

Secondly, can I ask if where you get your stats from? Is it from the internet or can you go out and buy the stats somewhere?

Keep up the good work, I'll keep my eye on this blog as I think it's good to see how it should be done...

[mathare]

My question is regarding your testing of the system. Are you saying that you have to go back over games from 10 years ago onwards and find out the best available odds and average odds for each game? Would this not take a small army of monkeys tapping away at keyboards in a room for several years?

It would have taken a small army of monkeys hidden away in some room somewhere had some clever chaps not set up a website back then that captured all that data. I get my data from said website (football-data.co.uk) and use that in my analysis. They have odds for the English leagues going back to the start of the noughties. They carry historical odds and data for the European leagues but I don't know how far back that goes - I haven't got rounded to using it yet so haven't checked.

Hope that helps

[mathare]

23rd November 2009
I'm wondering if I have jumped the gun again with this team rating system and started the paper trade without fully working through the analysis to make sure the system stands a decent chance of being profitable. It occurred to me over the weekend that while the system had shown a decent profit of nearly 120pts in four seasons that this didn't mean much unless the profits were spread fairly sensibly across those seasons. Was one season showing exceptional profits and the others generating much smaller profits, or even losses? Only one way to find out....

The 05-06 season (the first test season) produced over half the profit with 61.315pts. The next three seasons produced 14.605pts, 15.86pts and 27.755pts respectively. So far the 09-10 season (excluding the games played over the weekend) is running at a loss of -14.59pts. This season's loss is not insignificant but the easiest way to see if this system could work is to paper trade it so that is what I am doing (in the relevant section of this forum). After all, a paper trade costs nothing and one can learn from it so I may as well give it a go for this, a system that may well produce reasonable profits by the end of the season but only time will tell.

If this season ends with a loss will I ditch the system? It depends - mainly on how large the loss is. One losing season may not mean the system doesn't work as we have seen there have been four profitable seasons prior to this one so I may give it a second chance next season but let's wait and see what happens between now and May first.

I had a few more ideas for tweaks to this system and other related ideas to check out too over the weekend so today I plan to start working my way through those to see what, if anything, comes of them.

[MattR]

Mat, did the previous seasons also go through a losing spell early on and start off in the red? I'm wondering because I would expect a rating system to start a little erratic until the form settles down a bit and the league takes shape.

[mathare]

Mat, did the previous seasons also go through a losing spell early on and start off in the red? I'm wondering because I would expect a rating system to start a little erratic until the form settles down a bit and the league takes shape.

Dunno yet Matt, that's next on my "to do" list. This morning (and early afternoon) I was working on some code to make my under/overs systems easier to manage. It's a bit of a grind getting the best odds for each match as you have to select the match in Oddschecker, select the market and then scan the prices for the best odds and they aren't always consistent about whether they list over first or under first so it's always open to me getting the prices the wrong way round. Fortunately I have now written a bit of VBA to extract the prices from Oddschecker automatically saving me the trouble. Makes it a lot faster to run as well as more reliable.

I'll start looking at those team ratings profits and how they vary throughout previous seasons now. You're right, it should be interesting and it's easy to see how they could take a while to settle down each season

[mathare]

Mat, did the previous seasons also go through a losing spell early on and start off in the red?

The 05-06 season took off like a rocket, levelled out around November, slumped around 20pts in the new year and then shot back up. 06-07 started with a loss of around 30pts in the first 6-7 weeks, recovered about half of that, lost another 25pts or so and then staged a remarkable recovery pulling in over 60pts between mid-December and the end of the season. The following season saw a drop of about 20pts early on, bobbled about a bit, had a strong new year then dropped back a bit towards the end of the season. 08-09 had a nothing start to the season with a few wins and a few losses all but balancing one another out for a month or two before a steady climb ensued. There was a slump around March followed by a late season rally. And this season so far has seen a steady start to the season followed by a recent slump.

All of which suggests the losses observed so far this season are probably normal and can easily be recovered from.

[MattR]

That looks encouraging then Mat

Would it take long for you to run that on some other major leagues like the Italian, Spanish, German etc? Or is is just a case of pasting in the relevant data and then letting your software do it's work? I would think in theory that if it will work on the Premier then it should work on other 'top' divisions of a country. Most of those leagues like the Premier are stable with the established 'better' teams performing similarly most seasons.

[mathare]

Would it take long for you to run that on some other major leagues like the Italian, Spanish, German etc? Or is is just a case of pasting in the relevant data and then letting your software do it's work?

It's mainly just a case of pasting in the relevant data (in the appropriate format), clicking a few buttons and waiting while the PC churns through the data. That said there are a few options that can be tweaked that can sometimes make or break the system so while it may work in theory on the other leagues it may need some tinkering, and that's what would take the time were I to apply this to other divisions. However, I feel this is one of my weaker potential ideas for systems and I have some much better stuff to come so I am unlikely to try the team ratings idea out on the other European leagues just yet. It could be one for the future though as I would like to test a lot of things on the European leagues at some point.

I would think in theory that if it will work on the Premier then it should work on other 'top' divisions of a country. Most of those leagues like the Premier are stable with the established 'better' teams performing similarly most seasons.

True, and that's something I plan to exploit via other systems but this one doesn't perform significantly better for one division within the English league than the others so whether it would still work for the European leagues when you factor in the Segunda Liga, Serie B etc I don't know. Few countries have anything like the depth of professional football that we have in this country.

[mathare]

23rd November 2009 (part two)
I wasn't thinking too clearly on Friday, I can't have been, as I rushed to get everything ready to post the qualifiers for the Team Ratings system for the weekend matches and completely missed the fact that where one system exists there may be several. And indeed I think I have found a couple more along the same lines.

A lot of the team ratings analysis I have done has revolved around the concept of value. This value has been established using three different best fit types on charts of the data. The system I posted up at the back end of last week was for just one of those fit types, and is actually the weakest of the three fits. It occurred to me this afternoon that I ought to play with the numbers for the other two fit types (logarithmic and polynomial) and lo and behold I have been able to find a reasonable looking system for each fit type.

Like the linear system I posted up last week these new systems use margins of value within a certain range. Even though the fits are better the profits are not quite to the same levels as the linear fit system though, unfortunately. That said they are not too bad with profits of 75.665pts (9.28% ROI) for the logarithmic fit and 97.09pts (19.07% ROI) for the polynomial fit system. Better ROIs than the linear fit if not quite the same profits! I have attached the number of bets, profit and ROI per season for these systems. As you can see 06-07 wasn't a great season for either system and for both systems the profits could be spread a bit more equally (ideally) but at least they are producing a fairly consistent number of bets each season and the overall profits are there at least. I'd say it's certainly worth paper-trading these two just to see how they pan out between now and the end of the season.

[mathare]

23rd November 2009 (part three)
There are still doubts as to whether I am thinking sufficiently clearly as my head is absolutely buzzing with ideas at the minute and bits of it are only slowly dropping into place. For example, the above thinking on the logarithmic and polynomial fit systems got me thinking about the linear system again. I made a very slight tweak to one of the value margin limits and improved the ROI by around half a percent, adding another 15pts of profit. However, the profits for this version aren't so evenly spread as for the original, as the attached table shows. In the original version the worst full season still netted nearly 15pts while in the new version the worst season was only a little better than breakeven. But one season is improved by 25pts so we seem to stand a better chance of a big winning season but at the risk of making very little at all. Is it worth it? At this stage I don't think it is worth tinkering too much. I will test the system live on the forum and see how it goes till the end of the season at which point I can really work on a full optimisation to squeeze every last point of it if I still feel it is worth it. Yeah, that makes sense. I think!

[mathare]

26th November 2009
Progress on researching football systems and processing all this data has slowed of late as I haven't been 100%. My various health issues (none of which are major but are enough to keep me under the weather) have been getting in the way and some of the new tablets I am on have whacked me for six the past few days so I am not getting through as much of this work as I would have liked. That said, some progress has been made.

My recent blog entries saw me unveiling the discovery of some decent candidate systems that are now being trialled live on the forum. They haven't started well though I have to admit. Still, it's early days yet.

Those systems were the end of one line of investigation on the team ratings idea so I was able to start work on some other related ideas. The two I have finished involved the same basic idea of value bets with best fits used to determine where that value existed but this time I split the data into home and away teams. That is I created a ratings model whereby the ratings took into account home and away form but the fits used only home or away form and would consider only bets on the home or away team as appropriate. So the home ratings model used fits to all the home ratings and looked at value bets on home teams, and similar for the away model but using the away teams. The standard value test provided nothing of interest (as expected) but there were some profitable systems when one limits the margins of value being considered, as we did with the original system to produce the three systems above. However, none of these new systems produced significant profits on a season-by-season basis - either the profit was too small to really concern ourselves with or one or two good seasons were producing all the profit which means there are obvious long-term doubts about the usefulness of these systems. So nothing has really come of these alternative models which means I can shelve those and move on to something else. I started on another idea this morning but I haven't got that far yet. I will update here when I have more to report.

[mathare]

27th November 2009
My terrier-like attitude continues as I have not yet dropped this idea of team ratings and moved on to areas I think will be more productive and meaningful. There is still the possibility of wringing a few more semi-decent systems out of this idea I feel.

Having broken down the data into separate home and away models and reapplied the fits the next stage is to split the data by division instead. I have just completed the analysis of the Premiership model and might have a couple of interesting systems to test. The reason I think this divisional split will pay dividends is due to how the ratings are constructed. When teams move divisions their rating is reset to the initial value. In the Premiership there are only three spots that involve a change of division - the relegation places. There is no way out the top of this division which means the best teams will continue to increase their rating at the expense of the teams which finish towards the foot of the table. Therefore the range of ratings within this division is much greater than for all other divisions - a team in the lower leagues won't have a rating above 2000 without getting promoted really whereas Man Utd and Chelsea have recorded ratings up around 5000. These high ratings will be affecting the fit that is then applied to the 72 non-Premiership teams whose ratings is more in the 300-2000 range so it makes sense to split out the top flight and study it separately.

Anyway, I have done what I usually do for these things now and have plucked out a couple of systems that show decent but not mind-blowing profits. The first has produced 56.73pts of profit from 429 bets (13.22% ROI) from the start of the 05/06 season but is having a rough old time of it this season. The system hasn't previously experienced a run quite so poor as the one it is on at present but we shall see if it can recover when paper traded. The second system has generated 26.23pts of profit which isn't that impressive over nearly 5 years but has done so from 254 bets at a ROI of 10.33% so I am interested to see how it goes for the remainder of this season.

I will try and have the bets from these new systems online later today/tomorrow all being well. I will post and track them in my Team Ratings thread with the previous three systems this line of enquiry has thrown up.

[mathare]

30th November 2009
Guess what my next line of enquiry was having investigated the Premiership team ratings? If you said looked into the Championship team ratings then give yourself a pat on the back as that's exactly what I did, and with pretty good results too I think. My team ratings system portfolio is expanding rapidly now as I have uncovered another three systems (one for each of the best fit types I considered) each applying to away teams only again. The idea that away teams offer the best value is really starting to hit home from this exercise.

These systems are odd though. Remember a while ago I talked about margins of value and found a few systems when I looked for odds that were better than my fair odds but not so great that I suspected my model had missed something? These systems use the same concept - twice! There are two profitable ranges in these value margins for each fit, one at the lower end and one higher up. This suggests there is value to be had when the bookies have got the odds either slightly wrong or significantly wrong but not inbetween. I know, it's odd isn't it but it's a clear trend than appears in the data so who am I to argue with it.

The headline stats for these systems are:
Linear: 511 bets, 58.005pts profit, 11.35% ROI
Logarithmic: 478 bets, 70.685pts profit, 14.79% ROI
Polynomial: 363 bets, 87.595pts profit, 24.05% ROI

Not a bad set of stats for just one division I think you'll agree. When I get myself sorted I will add these into the relevant spreadsheet and track the selections in my Team Ratings thread.

[MattR]

Looking good Mat. I think there is something in aways as well. I have been testing something recently where I am generating half time odds and paper testing the qualifiers (those that were deemed value in relation to the actual odds on offer). I'm testing them in three systems, home/draw/away. To highlight what you are saying about the aways being undervalued by bookmakers here's the results so far.

These are to £2 level stakes

Home Wins
55 bets
17 wins
Loss -£13.63

Draws
46 bets
19 wins
Profit £3.32

Away Wins
63 bets
16 wins
Profit £31.28

Interestingly as well with the home wins is when the home team is the underdog, those with odds 3.00 and above, the loss is running at just -1.72.

Another interesting stat is that 16 of the 63 that have been ahead at half time, only 10 have gone on to win, and of the 47 that weren't ahead at half time only 5 have gone on to win. So there have been 16/63 half time wins but only 15/63 full time wins. Not sure which would have produced the better profit as I haven't record the match odds price, just the half time odds. Looking at some games today the half time odds seem to generally be a tad higher until you get to seriously odds on favourites when the match odds are higher on the underdog than for the half time win.

[mathare]

To highlight what you are saying about the aways being undervalued by bookmakers here's the results so far.

These are to £2 level stakes

Home Wins
55 bets
17 wins
Loss -£13.63

Draws
46 bets
19 wins
Profit £3.32

Away Wins
63 bets
16 wins
Profit £31.28

Take a look at those SRs as I think they are telling in this case. Less than 31% for the homes, over 40% for the draws and around 25% for the homes. Long term, across all divisions, there are around 46% home wins, 27% draws and 27% away wins (roughly). Your SRs are way under that average for home wins, well above it for draws and about right for the aways yet it is this latter category that is showing the greatest profit. It's a small sample but it looks like further evidence that the away teams are overpriced and that you don't need an SR up around 27% to make a decent profit.

This is interesting stuff Matt, thanks for that. Certainly food for thought.

[MattR]

That's a good point about he strike rates Mat. The odds range on those away bets has been from 2.32 to 15.00. While the odds range on those that have won has been from 2.56-10.00 so they have won right across the board. I think I'll start posting up the draw and away qualifiers from today actually - there's one starting right now in fact.

[mathare]

4th December 2009
It doesn't take a genius to work out that after devising football systems specific to the Premiership and Championship that I would start to look at League One and League Two to see if I could pull the same tricks there too. That's exactly what I did, and it worked. Three more systems for each division - one each for the three types of best fit I have used for the data being studied.

Things looked promising for League One when the simple value test (backing all teams available at value odds for 1pt each) had a ROI that was around 4% better than for all the divisions combined. This suggests the fits for this division suited the data much better so I was hopeful that when it come to examining the margins of value that had proved so fruitful before that I would uncover some gems. Yep, right again. The linear system returned a profit of 82.69pts from the start of the 05/06 season taking in 503 bets for a ROI of 16.44%. The logarithmic system was the weakest with a ROI of 7.61% (56.805pts profit from 746 bets) with the polynomial system producing 46.88pts profit from 524 at a ROI of 8.95%. Those last two systems may not be world-beaters but are decent enough and an average profit of around 10pts or more a season means they may make worthy additions to a portfolio so all three of these League One systems will be paper traded to the end of the season.

If I thought things were looking up for League One then League Two was in a different class altogether. The simple value test of backing all value teams regardless of how much value the odds offered (as long as there was some real value in the odds) was profitable! That means all one needs to do to make money is churn out the value odds for each League Two match and back any team whose odds exceeded these value odds - simple! Of course, I wasn't going to settle for that as the ROI was only 1 or 2% so I knew from experience that the application of a couple of filters would greatly improve these figures. How does a profit of 79.685pts from 464 bets (a ROI of 17.17%) sound? What about 359 bets returning 99.03pts profit for a ROI of 27.58%? Or another 99.54pts of profit from 480 bets at a ROI of 20.74%? They are the figures for the linear, logarithmic and polynomial League Two systems - not bad eh? I need to see how these fare in the 'real' world so will be paper trading them alongside the other systems taking the number of systems uncovered from this line of research to 14 so far, with a couple of ideas still to investigate.

I will be posting qualifiers from all systems later on this afternoon or tomorrow - I just need to get the spreadsheet up to date to handle these new systems and then we're away.

[mathare]

7th December 2009
I said a week or two back that the Premiership was probably distorting the fits for the other divisions because the teams with high rating points tend to keep most of those points rather than have their score reset when they get division (via promotion) so I set about splitting the dat up by division and modelling that. I produced a number of systems that way but there are other ways to skin a cat. If the Premiership is the major factor in the fit not working for the lower divisions when then separate the Championship, League One and League Two? Why not build a model based on all three divisions? That's exactly what I have just done, and again it has turned out a number of interesting-looking systems.

The headline stats look pretty good:
Linear: 1469 bets, 112.95pts profit, 7.69% ROI
Logarithmic: 1302 bets, 139.54pts profit, 10.72% ROI
Polynomial: 1538 bets, 123.01pts profit, 8.00% ROI

But break these figures down by season and you'd see that the seasons 05/06 to 07/08 are steady, 08/09 is a stellar season making plenty of profit and then this season is a nightmare so far with each system seeing a drawdown in the region 25-30pts. Is that a correction for last season, putting things back to normal? Or is this a sign that the system has died and will record a big loss this year? There is no way of telling at this stage so I will, of course, be paper trading these three new systems alongside the others to see how they pan out for the rest of this campaign.

And after nearly a month of working on it (on and off) that pretty much brings to an end the study of these team ratings. I will take a bit of a break from this when I have tied up a few loose ends and come back with some new ideas in a little while. There is plenty of work in the pipeline...

[mathare]

7th December 2009 (part two)
At the back of my mind for the past few weeks have been conflicting thoughts. On one hand there is an optimist thinking these football systems can be combined to form lovely little portfolios that can be relied on to bring in a nice sum every football season. On the other is the realist asking why these potentially excellent systems don't seem to be living up to their reputations in the paper trades. Sure, it's early days for the paper trades but there are more red numbers than black in the columns that matter.

It's fair to say that the work I have done to put together these systems includes a fair amount of what could potentially be construed as data mining. The system rules are justifiable and I can explain why each of them makes sense but all of these systems were constructed using an artificial set of odds really. They use the average odds from a variety of bookmakers but no-one intentionally bets to average odds do they? This is where the lack of an SP in football is really irritating and frustrating. Because I am using rules that apply both a minimum and maximum to the odds at which any given team should be backed (through the margins of value) I may find that the averages odds lie squarely in that window but that the best odds are outside it. Then what? Back at odds that are not the best available? That way madness lies but should I spurn the bet simply because one bookie has gone out on a limb and set odds outside my value range? It's a real conundrum.

What I have done to try and work through this issue is bring in the odds from a couple of bookies and use them in place of the average odds and then recalculate the system figures. I have done this for the Premiership Linear system initially. The average odds give a profit of 68.92pts from 408 bets. The Bet365 odds result in a profit of 18.62pts from 390 bets - a much lower profit although roughly the same number of bets. Hmmm. Use odds from Ladbrokes and there are 325 bets resulting in a loss of 23.65pts. It's 5.06pts from 280 bets with Stan James and 28.47pts from 364 bets using Victor Chandler's prices. Even using the best odds from BetBrain it's only 27.35pts from 414 bets.

None of those are even close to the profits I claimed before are they?

I have checked back on how the odds I used previously were derived and that's all fine. I have double checked the formulae used to calculate the profits and they are all spot on too.

I still don't quite understand it and will have to let it churn around in my brain for a while and see what comes out later - I'll probably have some idea when I settle down to go to sleep tonight! I think all I can do at this stage is paper-trade the systems and see how that goes. Those trials are using the best odds from the Oddschecker bookmakers, which is what I would use when backing the selections myself and seems as good a set of odds to use as any. Anyway, I have slapped the optimism back down a bit and mentally started halving any of the ROI figures I gave previously to make them potentially more realistic and achievable. But the time for really reviewing this is in May, not now, but I felt it important to share this insight and add the potential health warning that the claimed profits may not be possible but not because I have been fiddling the figures - I can't quite explain why. Yet.

[MattR]

Interesting Mat. What if you just used those bookmakers and took the best odds on offer from those for each game. Where does the ROI and profit sit then?

[mathare]

Interesting Mat. What if you just used those bookmakers and took the best odds on offer from those for each game. Where does the ROI and profit sit then?

I don't know and I have killed off the spreadsheet I was using for it now. Oh well. As I said, the paper trade for the rest of this season is the acid test as far as I am concerned. I will look again at the end of the season how the odds I used during that trial compare with the odds available elsewhere (using data from football-data) and compare ROIs again at that stage

[mathare]

9th December 2009
I am very angry and frustrated! I didn't realise till around 9.30 last night that in amongst all the Champions League games and FA Cup replays were a number of Championship fixtures. Because I hadn't spotted these were being played I haven't included them in any of the paper trades I am doing on this division, and while they are only paper trades I was trying to make sure they were based on real odds that I recorded so I knew they were accurate. I can't get the under/over odds now either.

With all this in mind I set about trying to work out how best to update my various football spreadsheets to account for these missed fixtures and in doing so have uncovered some glaring errors in some of my code that invalidates a lot of the work I have done so far in this division, and perhaps in others too. This is BAD news!

I will update this thread and the appropriate system threads when I have finally worked through the mess and assessed the true impact of these errors

[mathare]

9th December 2009 (part two)
Oh, what a mess!

I have sorted out my Unders-Overs spreadsheet and associated systems now, including double-checking the bets for this season are now spot on. They may not exactly match what I have posted as part of the paper trade before but the figures are right at least. I'll sort all that out on the appropriate thread shortly though.

I think the team ratings systems will be less of a hassle to sort out. I hope so anyway!

[mathare]

9th December 2009 (part three)
My Team Ratings spreadsheet was much easier to sort out, thankfully!

It's a good job I have an honest face and am only paper trading these systems though as last night would have netted a profit of 23.40pts largely thanks to away wins from Blackpool and Cardiff, both of which came up on all three Championship-specific systems. My figures will include these bets (which I have posted up after the event) for completeness as I am tracking the systems performance over the season as well as during the paper trade so need all the games in there. It's typical that the evening I miss is a cracker though isn't it?

I'll have to be more careful not to miss games in the future. I know how it happened too, which is more annoying. I usually check a certain webpage for a service I use at lunchtime and that lists all the fixtures for the next few days along with their ratings. Somehow, yesterday I forgot to check this webpage and missed those matches. Oh well, mistakes happen I guess, and I have learned to be more attentive in the future :)

[mathare]

11th December 2009
Why? Why do I always seem to do this? I have several projects on the go and then one day I wake up and find something else, something that sounds more 'fun', that I want to do too and throw that into the mix as well. Why can't I prioritise and actually finish one project before starting another? I am juggling so many things at the minute it's ridiculous, and at the back of my mind is the thought that the missus probably wants me to go back to work one day not spend the rest of my life starting umpteen little (!) projects and not actually completing any of them.

The latest project to be thrown into the mix is the back testing of my False Favs. I started to go through it manually this morning (as that is how I pick all the selections on a daily basis anyway) before realising that to pick the horses, then fill in all the results with SP and BSP is going to take ages. I know Excel and VBA well enough to know this can be speeded up a little, as long as I have all the data in the right format. Which I don't but I can work on that, right? So that's what I set about doing, organising and reformatting racecard data so that eventually I will be in a position to hopefully automate the back testing of the False Favs, at least to some degree. And it has taken all morning so far and I am still not that close to completing this portion of the task. Meanwhile I have various notes and files littering my desk and HDD relating to the various other projects I have on the go. Would it reallly be that hard to finish some of them off? For me, unfortunately the answer seems to be YES!

[mathare]

11th December 2009 (part two)
It may have taken all morning and most of the afternoon but I hope it was worth it. I have hopefully proved to everyone (including myself) that my False Favs lay system is profitable in the long-run despite the recent poor performance. I have put up over 16 months worth of qualifiers from the last two years and the graphs show a steady upward trend - the system definitely works but like any laying system it has swings, both upward and downward. But it is the long-term trend that is most important and that is certainly in the right direction.

As a result of this back-test I have also improved on some of the tools I had previously built to help me with projects such as this one and I now have a fully working toolset for back-testing various systems, which will surely come in useful in the future. I feel rather content about all of this now, and I am certain that this is the end of that project too - I won't back-test the lays any further back, I see no need to, and there is no way I can fill in the gap between November last year and June this year - so I can draw a line under it and check it off as done on my ever-increasing 'to do' list.

Now what's next?

[tophatter]

Now what's next?

A cup of tea!

[mathare]

14th December 2009
It was a staking review morning for me today as I was getting sick of seeing the stats on my spreadsheet that tell me several of the lay systems I follow are in the black when viewed in terms of points profit but in the red for cash terms. It means I am losing when the points are worth more and winning when the points are worth less - exactly the opposite of what I should be doing. I'm maximising my losses and minimising my wins, to some degree at least. So I have taken some time out to review the staking for all the W2W lays systems I follow.

Before discussing the new staking plans I will be applying to these systems I ought to explain the staking I currently use as my default plan. I use a 1% fixed stake rounded down to the nearest 50p. So for a £500 bank my stake is £5, which is how it remains till the bank reaches £550 (or £450) at which point the stake is recalculated. It made sense to me when I first started to use this plan as it's effectively staking slightly under 1% on average so has that little bit more safety than standard 1% staking. It also makes life simpler, staking £5 rather than £5.23 etc.

Now I am beginning to realise this 'one size fits all' approach doesn't work that well so I am looking at plans that suit each individual system better. I am looking only at simple plans - namely fixed stakes, fixed liability, percentage stakes and percentage liability - as I don't feel it is worth complicating issues more than necessary. I don't want to even think about loss-chasing plans and so on - keep it simple, stupid. So I have split out the data from each system under review and applied each of these plans to the selections using a £500 starting bank with settings of £5 fixed stake, £50 fixed liability, 1% stake and 10% liability. I have used these settings as they mirror my default staking plan fairly closely in as much as I use a 1% basic stake (which is £5 from a £500 bank) and cutoff at 11.0 which means my maximum liability is ten times my stake so £50 or 10% of the bank. All figures take a 5% commission into account and are based on my data rather than official bets and odds.

CD Lays
Using a £50 fixed liability is best here, nearly doubling the bank by the end of the review period (924 lays from 29th April 2007 to yesterday) at a ROI of 1.08%, which is 0.40% better than the next best plan which was £5 fixed stakes (£171.90) profit. Fixed liability is only really worse than fixed stakes in one brief period near the start of the lays and is usually way ahead in profit terms, especially in the second half or so of the bet sequence. The bank reaches a lowest point of £397.06 from a starting value of £500 so the plan is very safe too it seems.

The actual staking I had been using comes out worst of all the plans.

Ladies Lays
The most profitable staking plan is fixed liability with a profit of £184.65 from the 342 lays on this system, equivalent to a ROI of 1.08%. Fixed stakes comes next with £165.75 at a ROI of 1.72%. Percentage stakes has a ROI of 1.26% generating £141.06 of profit while percentage liability and my own staking plan are dreadful and best forgotten about. The graph for these staking plans shows the fixed liability plan to have deeper troughs and higher peaks making it a much swingier plan than the other two, which look pretty similar but the vote goes to fixed stakes for simplicity and the better ROI (and better profit, ultimately). The £500 reaches a low of just under £350 for fixed stake, £226.70 for fixed liability and £360.97 for the percentage staking plan so fixed stake again seems safe enough with 70% of the starting bank never brought into action.

Lay'em
Here there are a massive 3004 lays so the staking analysis has plenty of data to work with. Percentage liability results in a loss of £189.66 so I'm not using that plan. Percentage staking has a ROI of 0.43% for a profit of £488.26, fixed stake results in £792.95 profit at a ROI of 0.82% while fixed liability generates £1024.83 at a ROI of 0.68%. So fixed liability results in the most profit but fixed stake has the best ROI. When the lays hit a bad run the fixed liability plan drops hardest and fastest but it is fixed stake that gives the smoothest ride so looks the best candidate here. The record lows on the banks are £346.60 for fixed stake, £85.93 for fixed liability and £288.68 for percentage staking. It has to be fixed stake on that basis then.

Max Lays
Of the gour new plans only the fixed stake and fixed liability show a profit and both are in three figures, which is a decent start. The actual figures are £105.55 (0.37% ROI) for fixed stake and £167.55 (0.36%) for fixed liability. Unfortunately the fixed liability bank busts at one point so we'd best not use that one, eh. Admittedly the performance of all plans has improved greatly since that point, which came a good while back (around 400 lays into a 941 lay sequence covering 19th March 2007 to the present day) but even so the liability plan is very twitchy with high highs and low lows while the fixed stake plan just gets on with it and runs relatively smoothly and steadily. The lows on these plans are £189.55 and -£144.54 for fixed stake and fixed liability respectively but as I said this was some time ago and things have improved a lot since then. Anyway, fixed stake is better than the staking I am currently using so I guess I should switch to that.

NH Profit Machine
My staking isn't so bad here with my plan coming in 3rd but still well behind the two fixed plans. Fixed stake makes a profit of £858.90 (2.76% ROI) whereas fixed liability comes out at £1533.65 (2.82% ROI). The percentage plans make more profit, significantly so for the percentage liability plan which makes nearly £3500 profit, but the ROIs are much lower. Also the percentage liability graph is mental! It has crazy spikes that come crashing down as fast as the profits build - I don't need that sort of scary ride thanks very much. Fixed stake gives the smoothest ride but fixed liability isn't that bad for this system. The ups and downs are a little more extreme but not massively so and the fixed stake and fixed liability lines are diverging throughout indicating the latter will continue to generate greater profits and it has the better ROI, just. The banks are safe with lows of £388.77 on fixed liability and £469.00 on fixed stake. There is no significant extra risk associated with using fixed liability so that is the plan I will use here.

I'm not suggesting anyone else following these systems should immediately change their staking here, by the way. These are reviews done on my data and are changes I will make to my staking - how you stake these systems is up to you. I have found that the aforementioned plans seem to suit the data best according to my records but the same may not be true for you so you should conduct your own staking reviews before making any changes to your existing staking methods.

Incidentally I did the same exercise for my own False Favs and will be changing my staking there to fixed stake.

[mathare]

14th December 2009 (part two)
Having got back into playing poker again in the last few decision, I decided this evening that I really need to get out of the habit again. Due to health concerns I have been out of work for nearly four months now and have been using the time to learn more about what it takes to make it as a professional gambler and while I am a way off making a living from the game I have streamlined my portfolio, made sure I am staking properly and generally learned a lot about the mindset required to make an income solely from gambling. It's a rocky ride but I think I could ride it out - if the missus will let me!

My thinking has changed quite a lot from when this was just a hobby, a small second income alongside my salary. Everything is about the bottom line now more than it used to be. How am I spending my time and what is that doing to contribute towards making a profit? I allow myself the evenings off but during the day I want almost everything I do to somehow be working towards making professional punting pay off for me. Staking reviews, systems research, number crunching etc. is all good ground work that will benefit me long term. Taking the time to place my bets, checking the markets so I can get good odds etc are moves that will see a more immediate pay off.

Poker, on the other hand, is lucky to bring in 20c an hour! What use is that? It's profit, right, and I am gaining experience true enough but I am struggling to see the pay off for the effort I am putting in. I have been around poker - playing, reading, learning - for several years now yet I am still piddling about on the micro stakes cash tables and have yet to make anything like a decent score from the cards. That's not to say I am a really bad player, I don't think I'm too bad when I am truly focused and playing my 'A' game, just that for whatever reason (mainly not having the time to really focus and dedicate my efforts into playing to the best of my ability I suspect) I haven't been able to get far off the ground with poker. Not that I am giving up, it's just that I want to move poker to one of my many back burners and focus more on projects that I have had in the pipeline for ages now and I really need to do something about them. They could have a very bright future and pay off in several ways, not only making me money but also giving me an opportunity, so I ought to take them seriously and give them the time and attention they deserve and require.

The thing is with me is that while something is new, fun and interesting I can't let it drop and have to keep plugging away with it. Until it gets a bit stale, or is usurped by something more fun or more interesting to me at that moment in time. Then the original project is left to stagnate and I find it hard to pick it up again. In order to breathe life into some of my old work I have had to start from scratch, both to keep it interesting by going back to basics and to make it possible to actually start it again as I didn't keep very good notes on what I was up to. I have tried to improve on the latter part of that by making detailed notes on everything I do now. For the football team ratings stuff I did I ended up making 25 pages of notes. In fact they are more than notes, I write something akin to a full explanation, a manual of sorts detailing the approach, the results and my thoughts along the way. I want to be able to pick these things up at any stage by reading these notes and hopefully get right back into the correct sort of mindset. That's the idea anyway.

One of the projects I have on the go is a software development project that I had the idea for several years ago. I had dabbled with bits of it in the past but never really got very far. I gave it up as I didn't have the time to dedicate to it (sound familiar?) but now I could make that time and do something useful with my time. If I need to go back into the job market, is it better to refresh and develop my software development skills or my poker skills? Much as I would love to say poker is more important, it's not is it? I took this project up again recently and got going but then hit a major stumbling block. I didn't (and still don't) know how to get to the vision of the software I have in my mind. I know what I want it to do but I don't quite know how to get there. I started to learn Visual Basic as that is what I will program in but the training materials I had didn't take me far enough. So I have got another course, a more thorough one, and I know I should see it all through from the very start but it is boring as I know most of the early stuff so it's not gripping me and making me think "I need to do more of this, it's fun." Sure, I could skip some of the early parts but as I am learning a whole fresh language I want to make sure I have all the basics down pat and know all the little tricks these courses tend to contain as they will surely come in useful in the future. So I have the carrot of the bits I want to learn dangling in front of me but it's just not tempting me, it feels more like a chore than something I want to do at present. I need to change that mindset and slog through these basics so it gets a bit more involved and fun then hopefully I will be re-enthused and can start to make some progress.

I am more likely to convince the missus to let stay at home and be a (semi-)professional gambler if I can show I am doing something useful with my time - and that means making good progress with this software project or proving I can make a decent income at the card tables. At the minute I'd say the former is a lot more likely than the latter so I have to give that route a fair crack of the whip. I will still play poker but I need to think of it as a treat almost, something to be played on and off for profitable enjoyment but not one of the main uses of my 'working' day.

[mathare]

16th December 2009
The recent poor run of results on the horses continues and in all honesty it is starting to grind me down. I have started to really fight for good prices but it's not making an ounce of difference when the horses don't play their part. With the profit from the Chelsea-Everton match last weekend I am pretty much assured a profit this month unless there is a total meltdown on the nags in the next couple of weeks but that profit should be greater if this were an average month (whatever one of those may be).

I am seeking solace in numbers, as I so often do. I want to make sure my portfolio is tight and that my staking is appropriate. Fortunately, since I added in the calculation of profit in points a few months back this is much easier now than it used to be. All I really need to do is check that each of my systems is profitable when viewed as points and that my staking is sensible given things like the longest losing run to date. Here is that information for the W2W Systems I follow:

45 Males
I hold my hands up now and admit I haven't been keeping a very careful eye on this one recently. My bank isn't much above the barrier I set so I have sort of forgotten about it somewhat. Naughty, naughty. Anyway, I'm down 5.38pts at the minute with a current losing run (CLR) of 1. The LLR is 9 so my 3% staking is fine for handling the losing runs but I must remember to keep an eye out for selections.

Claiming Jockey
I only have 40 bets recorded and am down less than a point so I can't draw any meaningful conclusions here. An LLR of 6 (so far) plus a CLR of 3 would suggest that may standard 3% staking is fine for the job and it's a case of waiting to build up more data on this one to see how it turns out long-term.

Double Top
Up 144.65pts from 1404 bets so well in profit. The LLR is 16 so the conventional wisdom of doubling that to get the suggested bank means my 3% staking is pretty much spot on. The CLR is 7 with a short-priced winner coming just before that sequence started. It's been struggling the last month or two (down nearly 10pts since the start of November) but remains a solid system. It's just suffering the same blip that everything else seems to be having around now and as this is my biggest backing bank it is costing me more in cash terms than poor runs on other systems. But this does mean when it comes good again, as it will, then I will make back those losses in decent time. The graphs show this is doing well and is steady in the long-run at least, so I am happy with that.

First Timer
I only have very few bets on this system since March 2009, 14 to be exact, which makes me wonder about the rules and how Auto-Sys is handling them. They look fine, but the rules themsekves are quite restrictive so it kinda makes sense that I haven't had many bets. I'm up 5.51pts but this isn't that meaningful on such a small sample size.

Info Rated
Only 54 bets in this one too so again no significance can be placed on the findings. November saw the longest losing run to date at 6 but there have been a couple of runs of 5 losers in the past. I'm down 2.87pts thus far but the 3% staking should easily handle any losing runs and again it's a case of letting this one get into the long run rather than focusing on short term results from a small sample.

Johnston Fav Tracks
A profit of 149.75pts from 611 bets on this low SR system which has seen a losing run of 29 before now, making the current run of 6 look pretty insignificant. I reduced my stakes a while back on this system to protect the bank and it seems like a good job I did with such a high LLR figure. But if you can stomach the losing runs associated with a low SR this is a worthy addition to any portfolio with Mark Johnston bagging winners at good prices in the summer. The graphs show this year has followed the same basic trend as last year, a steady start followed by some big wins before slipping back a bit late in the season. This one is solid enough.

LTO3
A profit of 15.58pts (from just under 400 bets) but a cash loss of just over a tenner suggests I am doing something a bit wrong here. The profit graph for this one is spiky rather than a relatively smooth upward trend so I have probably been staking high as the profit comes down and then low as it climbs again, as can happen with percentage staking. This is a candidate for a staking review but as a system is profitable, on the whole. The LLR is 11 but it has been in good form recently with only 3 losing bets in the last 10. October and November saw a run of 9 losers though and only 2 winners in 21 bets so perhaps this one experienced the bad run earlier than most other systems.

Michael Stoute
+38.59pts from this one to date with an LLR of 17 (so 3% staking is safe enough really) and a current losing run of 5. This one has been pretty static this year really but long-term it's alright. Actually, last year wasn't so hot for this one either (a loss of 8.25pts) whereas 2007 was a bumper year (+43.45pts) so it is worth monitoring this one to see if the bubble has burst. I have sensible staking and a bank barrier in place though so I have the tools to be able to keep my eye on this one and am protected if it really has seen better days. Hmm, I hadn't realise how this had stagnated. Interesting.

NH Media
This system has seen a steady increase over the past couple of years, or at least as steady an increase as gambling banks can have given their ups and downs. I only have 19.44pts profit from 910 bets which is pretty small I have to admit. The current run is 3 winners in a row and just 3 losers in the last 10 bets whereas the LLR is 13. My staking is safe; this just doesn't make much profit each year but it can tick along in the background, and it's interesting to note that this is one of few systems currently experiencing a purple patch.

SixPlus Loser
Up and down, up and down. But mostly down this one, since the summer at least. A loss of 4.67pts from the fairly small number of bets so far (143 bets) with an LLR of 11 but a recent run of 2 winners from the last 5. Most losing runs seem to be around 3, 4 or 5 punctuated by a winner or two. Another to just let run and see where it goes I think. My analysis suggests this is best in the summer but can be very swingy but I am using a small sample so I can't be confident about those conclusions.

Trainer 3
I restarted this around March of this year and am up 10.07pts so far from 175 bets. Things peaked around October and have steadily fallen since. The LLR of 11 came in early-mid November but since that point it has been pretty much on message with 14 winners in 34 bets resulting in a net loss of 1.57pts, so the SR is about right but the prices on the winners are a bit on the low side. I don't have enough data to be totally confident of the long-term stats for this one but I think it's yet another one that needs to be left to bubble along and see where it takes me.

It seems that some of the systems are actually performing well at present, some are perfectly average and others are indeed running below the expected levels but this is natural for any portfolio I'd say. It's nice to discover that not everything is broken though - just the systems that are carrying more of my cash perhaps. I haven't found anything other than the LTO3 system that has really made me question my staking too so I am happy to let things go on as they are for the most part. I'm doing my bit but I need the horses to play along too and then I will surely reap the benefits of all these systems.