Tuesday, 18 February 2014

Correlation and Variance - A bad combination!

I think without a doubt, the biggest disappointment of the season so far has been the performance of the Established Systems. These are the systems that I developed first and before this season, they didn’t have a blemish on their record. 3 seasons, 11 systems each seasons, a profit on every system, every season. Not only that, before this season, no system even looked like having a losing season, so it’s not like you could say that a losing season was expected.

I discussed it on a recent post regarding the portfolio that Steve built for this season but Steve went in really heavy on 5 of the Established combined systems. Although I discussed the correlation point (never a great idea to follow 5 systems so closely correlated in a portfolio IMO), the fact that Steve built his betting bank based on historical drawdowns meant that in a way, he had taken the correlation into account. Hence, I don’t think Steve is to blame for his losses this season (I didn’t say that in the post but someone else accused me of saying that by the way!) and the only thing we can blame is the poor performance of the systems this season.  

Here are the results for all the Est systems, split by season (up to the weekend past):

It’s amazing that a set of systems which have created over 200pts profit each season for the last 3 seasons can be looking at a loss of 87pts so far this season. Last season was far from a vintage season for these systems but even so, an 8.1% profit across nearly 3,000 bets hardly points to the returns falling off a cliff this season.  

The strike rate this season is the issue and we are missing a full 6% or 7% of bets being winners compared to the last two seasons. May not sound like much but across this many bets, that’s 89-104 winners we’re missing compared to the last couple of seasons. I appreciate the average odds aren’t the same and therefore, any calculations here are not exact but you get the gist of what I’m doing hopefully. Add another 40 winners to these systems this season and we’ll be break-even. Add another 40 winners on top of that and we’d be having a great season.

Obviously, you can’t just go around pretending you’ve hit 80 winners more than you have hit and therefore, where have the winners went this season?

Well, it’s not that difficult to find where they should have been.  

Every 8-22 bet must appear 11 times since it must have appeared on the other 10 systems. If you look at the performance of this system this season, you see the issue. 5 winners from 27 bets. A 14.6pt loss. To break-even, this system is missing 5-6 winners. That would be 55 or 66 winners on all these systems. Looking at the historical results, the system is probably missing 8 or 9 winners. That would be 88-99 winners on all the systems.

There we go. Those 8 or 9 winners are the difference between success and failure for these systems this season.

Of course, pinpointing where the 8 or 9 winners have went is impossible but it shows the thin line at this game. The issue that anyone following a set of correlated systems suffers from is that if these 8 or 9 games don’t go your way (and they haven’t done this season), you suddenly appear to be missing 8 or 9 times the number of systems you follow that would have these bets. If we take the example of Steve (sorry mate!), that would be 40 to 45 winners he’s missing. Steve’s P&L this season for TFA would look much different if he replaced 40-45 losers with winners!

None of this analysis helps us understand why the systems are making a loss and why these 8-9 games haven’t gone our way but that’s the point I’m trying to make. It’s 8-9 games.

Many of the people following TFA run a portfolio of tipsters. They follow TFA to try to diversify the risk within their betting portfolio and I think this makes perfect sense. They try to spread the risk across different tipsters and different sports and in the case of football, they may follow other footie services to diversify the risk away from TFA. Great idea (especially this season!)

However, what is the point of doing this to achieve diversification if for TFA, they lump on a very small number of teams every season and basically, run the risk of variance killing them in any season?

System 8-22 has had 27 bets this season. 27 bets.  Given these 27 bets account for 297 system bets out of 1,487 total system bets this season on the Est Systems, that’s 20% of the overall system returns for this set of systems is dependent on 27 bets. Yes, it’s been a horrible 27 bets so far but do we really want our betting returns to be based on such a small number of bets?

Take the case of someone following system 6-21 this season. 281 unique bets this season. Yes, this system is missing the same 8-9 winners as system 8-22 but with a large sample of bets, the chances of variance killing the returns is much smaller.

If people take anything from this season (I count myself in this too by the way), it has to be that we need to get away from being so dependent on such a small number of teams. Yes, things went our way over the first 3 seasons and system 8-22 produced profits each season but looking at the bigger picture, it doesn’t take a big swing in luck for a system with less than 100 bets to go from a winning system to a losing system.

That’s not to say that I think system 8-22 hasn’t had much luck this season. I think it has had its fair share of good and bad luck as most other systems have but when you rely so much on a small sample of bets, it doesn’t take much for the system to make a loss.

I think the amazing thing is, if you look at the returns of these 11 systems since they went live (including this season), they still don’t look too bad.  Here’s the results:

I’m sure many a tipster would cut their arm off to achieve these returns over these sample sizes but maybe this season has seen a correction of some sort? I’m starting to think the higher combined systems probably overachieved during the first 3 season and if you look at the overall returns now, maybe this is more like the normal returns we can expect. Of course, it has taken a terrible 6 months this season to correct the returns and maybe I’m kidding myself but looking at these returns, would I say that the edge has gone from these systems? Not sure I would.

Anyway, I’ve been wanting to do this post for a while but I was waiting until an upturn started! Given we’re still on a drawdown, I’m probably tempting fate as a few more bad weeks and the returns will take an even bigger hit and we’ll be missing more winners than 8 or 9 I suspect but at the end of the day, I don’t control the results.

Good luck for tonight’s bets. 

No comments:

Post a Comment