Wednesday, 18 June 2014

Preview of Euro Draw Systems

One of my big pieces of work since the close season has been to develop a couple of new rating algorithms to allow me to build a set of Draw systems for the European leagues. I suspect if it wasn’t for the fact my UK draw systems were my best performing systems last season alongside the fact my outright European systems ended the season strongly, I’m not sure I would be bothering looking at another set of systems. Given I already have 50 systems live for next season, adding another 11 systems might sound a bit crazy but as always, I’m always looking for another edge at this game to allow me to diversify within my betting portfolio.

Last season in my own portfolio, I followed UK outright bets, UK draw bets and Euro outright bets. The missing piece of the jigsaw for me was Euro draws and as far as I am aware, there are no standout Euro draw systems available in the market. Being honest, even if there was a standout tipster or system in the market, it would have to have a long-term proven track record to get me to pay for it and follow it myself.

I’ve been here before though and my solution is always the same….build something to follow myself in future once it is proven to work. Therefore, it seems a natural progression to move onto looking at European Draw systems.

I actually planned to do this later in the Summer but I realised when I was writing the reviews of the systems from last season that I needed a break from writing and therefore, I started playing about with the data for the European leagues. As always happens, once I get started with data and ratings, I struggle to stop and I managed to build these ratings fairly quickly since I had all the European data to hand as I was refreshing the Euro ratings for my other systems this Summer too.

I’ll do as I usually do and share the backfitted and backtested results but as always, these are not meant to provide a guide of what we can expect going forward. Depending on how good my ratings are, the true profitability tends to be around 25%-35% of the backtested results although it varies by rating algorithm. All I can guarantee is that the results next season live won’t look anywhere near as impressive as the results I’m about to show!

I tend to share the historical results so people can see how the systems look in terms of bet numbers and average odds. Knowing that the results tend to be lower when live, we can use the backtested/backfitted results to help us set some targets for the season forward although for things like betting banks, we need to be really careful as the drawdowns historically won’t be near as bad as the drawdowns we can expect in future.

Anyway, caveats aside, let’s see what I’ve come up with for these systems. I’ve spent a little bit of time thinking through which seasons to backfit and which seasons to backtest and I’ve gone down the route this time of the backfitting the last 3 seasons worth of data and backtesting the previous two seasons. I could have went back further and included more data but I think I’m of the opinion now that the best data to backfit on is the most recent data and as long as the results look OK for the previous seasons, everything should be OK. I used to think that it was better to build the ratings with historical data and backtest over the recent seasons but I think the issue with this is that the ratings then get out of date fairly quickly as I’m using data that is so old. Hence, when I’m refreshing all the rating algorithms this season on the other systems, I’ll be flushing through more recent data and tweaking the ratings.

The best place to start is probably the overall results to get a flavour of what the returns look like for all systems and to check the filtering works as I intended to. Here’s the results by system for the 5 seasons combined:

I always love this picture when I’ve developed a new set of ratings as it is my first check that I’ve built systems that could be useful. The filtering on the systems works like a dream (as we would expect based on my other systems I’ve built) and I think the overall ROI of 24.4% is in line with most of the sets of systems I’ve built. I think if this was below 15%, I would probably be worried and likewise, if it was above 35%, I would be worried that I’ve fitted the data too much during the backfitting.

Of course, the table above doesn’t help us too much since it has 3 seasons of backfitting and 2 seasons of backtesting included. Here’s the results by season:

I quite like this picture too. I think we can see quite clearly which seasons are backfitted (ROI is circa. 10% points higher) but the two seasons which are backtested still show a very strong return of 18%. Basing it on historical performance, I think the backtested return can convert to something like a 5% or 6% return as a minimum live return and this would be acceptable I think over a large number of bets.

Again, we can see that for any season, the filtering may not work as perfectly as hoped but over a large sample of bets as we’ve already seen, the filtering can work well.

One place I have struggled a little with the other Euro ratings has been managing to find something that works across all 4 Euro leagues. I do tend to find France and Italy quite difficult leagues when looking at Euro ratings as what works in these leagues tends to not work so well in the other two leagues. Interestingly though, for these new Draw systems, I didn’t seem to have too many issues with France as the draw strike rate tends to be fairly high in France. Here’s the results split by league:

I find this quite pleasing that there appears to be an even split of returns across the leagues historically. The Italian league has a lower ROI but it does have the highest turnover by a good bit (lower goal expectations in Italy relative to the draw odds I found) and therefore, the returns and turnover look acceptable.

It’s interesting to look at the returns by league and season as over smaller samples of bets, we can always see some crazy returns for some leagues. Here’s the results:

I do find some of these returns quite amusing! It’s quite interesting how 2 of the last 3 seasons in France has saw low returns even though this data is part of the backfitting process but the first two seasons have great returns! Likewise, Spain has really flat returns over the last 5 seasons which probably points to Spain being the most consistent league out of the 4 which fits in with my findings for the other Euro systems. Spain is the only league I seem to have a good handle on with the other Euro ratings.

I think that’s enough of an introduction to the European Draw systems for the moment.

As always, the best way to end these posts is with an overview of the targets for the upcoming season. With it being the first live season, it’s difficult to set accurate targets but I always try to err on the side of caution. None of us know how good (or bad!) these systems may perform next season but it’s always worth setting out targets so we have a benchmark to judge the season against.

Here’s the targets for this season:

I think the system that stands out is DE1-DE6 as being the one with the highest potential ROC but this is usually the case for the combined systems. My only concern with following this system is the high turnover and the fact that if anything goes wrong with these systems next season, this system is likely to lose more money than some of the higher combined systems. However, as we saw with the UK draws last season, if following the highest combined systems, you run the risk of filtering out too many winners and making a loss even though the ratings have an OK season! It’s a tough balance following a set of new systems as you don’t want to push all in which leads you to filtering the bets someway but if you do this, you run some other risks,

If I was giving my own honest advice of how to play these systems, it would be to give them a miss this season completely or to follow enough bets to ensure any edge shines through but to play trial stakes. Trial stakes is something I’ve been doing for the last few seasons now on my unproven systems and it sort of works for me. Yes, the money won or lost doesn’t actually make that big a difference to my overall P&L but I get a feel for what the systems are like and how the losing runs are etc. This means that if the systems perform as I hope, then when I migrate these systems to my main betting portfolio, I already have a feel for how these systems perform and what they are like to follow.

Overall then, I think this concludes the preview of the new European Draw systems. As always with my systems and ratings, things look very promising but we’ll need to see what happens when the bets go live. I think for next season, my hope would be to see the systems make a profit in the first instance. I’m still not as confident with the Euro leagues as I am with the UK leagues and therefore, I’m not wanting to set the bar too high in the first instance. If all systems could make a profit and hopefully have a fairly stable season, then this would be a good first season I think. I guess I want something to build on in future and it’s always easier tweaking ratings to improve them rather than build something new from scratch!

We’ll see what happens…..

No comments:

Post a Comment