Wednesday 24 December 2014

The Reality Of Naïveté

Betslayer took the plunge, and dived in to comment on yesterday’s Garbage Goal Times’ post, essentially suggesting that it’s ok to have bad data in your numbers, just so long as you don’t have too much of it.
I think that is a very naive view, so some goal times aren't accurate?? so sample size isn't key to smooth out the data then? I don't think the odd delay would have that big a difference when you are working on large margins. 
The problem is that the number of goals varies significantly from one club to another, from one league to the next, and for any individual league from season to season, so it is useless comparing goal times in such a broad manner.

Data from the 2005-06 Ligue 1 season is quite distinct from the Bundesliga last season for example, as was data from Serie A in 2009-10 when compared with the following season.

At the risk of laboring the point, any system applying the simplistic logic that a goal scored by Liverpool last season in a certain minute should be considered the same as a goal scored by Aston Villa in the same minute this season, is headed for the poor-house.

Our sample size is always going to be small. If we look at the top 38 match leagues, there are on average less than 1,000 goals in a season, but distributed (albeit unevenly) over ‘90’ recorded minutes, and with each of the 20 clubs having a unique scoring footprint, it should be obvious that meaningful data is at a premium. 

There is certainly not enough meaningful data to be able to say, with a straight face, “that sample size is the key to smoothing out the data”.

Sample size could be, but the sample size simply for this strategy isn’t there unfortunately. I’ll let the readers decide who is showing the naïveté here.

Betslayer continues:
From analysing inplay using minute by minute data I can tell you it largely fits very very accurately to bookmakers inplay pricing. 
I think that’s exactly what we would expect to see, and precisely why the humble home trader is highly unlikely to have a long term edge versus the serious players in these markets. Certainly a broad-brush strategy based on something as simplistic as goal times isn’t going to cut it. I believe I have mentioned before that I have a nephew employed by one of the big players in London, and the data obtained and analysed is far beyond the reach of individuals trading at home.
Funny you think elo ratings are accurate enough for your betting and systems but a theory with 15% ROI over hundreds of bets is deemed to be flawed.
If I have ever claimed that Elo ratings alone are the answer, it was certainly many years ago. Again, such a simplistic approach won’t work, but an Elo ratings based approach when factoring in other ‘context adjusted’ data has proven to be profitable, at least for now.

As for the suggestion that "a theory" can have an ROI% of 15%, while in theory anything is possible, in practice it is not. Although an ROI of 15% or more is quite feasible over a smaller number of bets, it is somewhat improbable over several hundred. Such a large edge would soon be identified by the thousands of other traders swarming over these markets and eroded away. It is simply not credible that such a huge edge could possibly exist for any length of time. 

The latest FTL table has just three entrants with an ROI north of 15%. One is from just 51 selections and is destined to come down. A second is from 123 entries, and only the leader Fairfranco has over 200 selections. I would be very surprised if anyone had an ROI% above 15 come the end of the season.
Betslayer later added a second comment, although it was mostly him talking to himself, which is never a good sign! One line worthy of mention was that:
The market makers have the same limitations in timings of goals and indeed data.
No, actually they don’t. As I’d already mentioned before seeing this second comment, there are companies (both trading and bookmakers) with far more resources available than the home trader has, which means that they do NOT have the same limitations as most of us. They definitely do not put a 9th minute goal scored by Real Madrid at home to Elche in the same bucket as a 9th minute goal (after a 10 minute injury stoppage) scored by Hamburg away to Eintracht Frankfurt.
I know of no database for football without errors and trust me its an area I have a large amount of experience in.
The same experience that leads to the belief that a long-term ROI of 15% based on goal times is possible? 

Anyway, continue:
I also wonder where you get this mythical data to make your elo ratings, so you clean up goals so that no incorrect ones are included? what about incorrect offsides? Goals off peoples arses? No thought not, all data and systems are flawed in nature, doesn't stop people having an edge!
A few erroneous and bizarre assumptions made there, and again he’s missing the point that an Elo rating can only ever be part of the process. That there is an element of luck in football also seems to come as a surprise to Betslayer. 

Football is played and officiated by human beings. Most human beings have emotions, and many decisions in a match are by definition, subjective – all of which means that there is a certain unpredictability about any match. It’s why football is such a popular sport of course. The best team doesn’t always win and while no method or system can account for luck in any one match, over time luck balances out. 

Value doesn’t mean winners every time, or even over a period of time, but if you are identifying value correctly, in the long run you will come out ahead. 

Unfortunately, for it means more work, football betting is very much about context, and a one-size-fits-all approach such as using past goal times to predict outcomes for an in-play game isn't going to gain you a long-term edge, especially if the data you are using is wrong.

I’d also like to touch on Betslayer’s implication that betting is a game of us versus the market makers. Who are the market makers these days? The days of the markets being ‘made’ by a few bookmakers are long gone.

Money moves the markets. Pinnacle Sports’ prices are determined by punters opinions and the money they are prepared to put where their mouths are. They open with a soft price, with lower limits, and allow smaller sized bets to come in, and constantly adjust their prices to reflect the money being bet. As the prices firm up, the limits are increased. Exchanges work in a similar way, although rather than profit from volume and tight margins, they of course take commission.

Traditional bookmakers do not change their prices so often, but you will soon have your account limited or closed if you only take their prices when they are either market leading or close to it.

The advantage an individual has is that he can be very selective. Most of the time, any edge will be very small, but a disciplined approach and patiently waiting for the right opportunities can be rewarding.

1 comment:

Betslayer said...

Excellent work and as usual you jump to many conclusions without actually covering the key points. Do you actually know how I use goal time data? No, so how can you jump to conclusions. Plus you skirt around the fact any data you use is not clean, I know for a fact football data has a large number (not in reality, across all leagues) but still into 00's of incorrect scores and HT/FT's. As for me talking to myself, going off that rant you make a living at it. At least I get my 5 mins of fame hey :)