Monday 25 January 2021

New Year, New Challenge

Apologies for the lack of posts in recent weeks. As some of you will know from Twitter, I had an accident at the start of the year which means I shall be mostly laid up in bed or on the couch until mid-March, neither location being conducive to creating new posts. The phone works just fine for brief entries such as this one, but it’ll be a while until normal service is resumed. Surgery is this week, then recovery can begin, all being well. Stay safe. 

Wednesday 6 January 2021

When 2 + 2 = 5

Artwork by Joe, age 41 3/4
The last post was a little rushed, but as at least one person is looking forward to an update I thought I'd clarify some of my thinking.

@CatsofOxford pointed out that because of the partial correlation between the two individual Senate races, no bookie would offer the double, and that's true, but we can still use the implied probabilities in each market to calculate the outcome in terms of overall seats.

It was very strange to me that the odds for the Special race Democrat were so much shorter than for the Democrat in the Regular race. There were differences of course, mostly that while both Republican candidates were incumbents, usually a big advantage, one (Loeffler in the Special) was appointed, and the incumbency advantage is usually lower. Perdue in the Regular race, had name recognition, but even so, for the two Democrats to be separated by so much seemed odd. The majority of voters would be expected to vote by party, so why did one Democratic candidate have an implied probability of 0.62 and another just 0.465? 

Anyway, the point I wanted to make was that these races only had four possible results and with the Democrats starting from 46 seats and the Republicans on 50, the Senate makeup could be easily projected. Two coin tosses have three net outcomes - Two Heads, Two Tails, One of Each, with the latter most likely at evens. 

To stay on 46 seats the Democrats would need to lose both races, a probability we can calculate from the odds available as 0.2033 (0.38 * 0.535).

To move to 48 seats, the Democrats would need to win both races, a probability of 0.28 (0.62 * 0.465), and to end up with 47 seats, and the Republicans with 51 seats, the probability would be 0.5167 (or 1.93 in Betfair terms), yet the price for this last outcome was 4.9. 

What gives? Even allowing for some wiggle in the prices, that's a big difference but if 4.9 intuitively seemed to high, what price was fair? And if 4.9 was too high, then the other outcomes must have been too low. Or the root prices of 1.61 and 2.15 were wrong. Warnock's (1.61) was probably reasonable, but Ossoff should have been closer to 1.94. Something to watch in the future perhaps, but as these events are so rare there's not much to be gained from analysing them to death. I made a little on all three markets by betting with my heart rather than finding an edge, but I can't help thinking I was too late looking at these markets and missed a bigger opportunity. It happens. The bigger picture is that the Democrats look to have secured control of Congress and the Presidency, a nice turnaround from four years ago when the Republicans ruled the roost. So much for all this winning by Donald Trump.

Georgia Arbs

A rather unusual opportunity in the markets surrounding the Georgia Senate elections, of which there were two - one Special and one Regular. 

The odds are fluid as the election is taking place currently, but as polls were closing the Democrats were around 1.61 in the Special and 2.15 in the Regular.

Based on these odds, the probability of the Democrats winning both seats and increasing their total to 48 was therefore around 0.28, the probability of winning one seat was around 0.52, and the probability of losing both races was around 0.2. 

The "Number of Democrat Senate Seats" market for these three outcomes respectively was 3.0 (too short), 4.9 (great value), and 2.7 (too short). There was also the mirror market for the "Number of Republican Senate Seats which offered similar odds, but strangely not identical. Backing 46 Democrat seats is the exact same as 52 Republican seats and so on.

Great value doesn't guarantee you a winner, but the odds seem way off. Be careful, because since I started this post, the Special odds have moved out, but there is opportunity here.

It maybe because some low-information bettors don't read the Rules for these markets, but good luck. 

Friday 1 January 2021

Au Revoir 2020

2020 has certainly been an interesting year. Aside from my Mum's passing in April, which appeared to be non-COVID related, and the continued slow and steady decline in my 93 year old Dad's health, I have to say that this year has been pretty good to me.

Unusually for me, I took a vacation early in the year, and had a great two week trip to South Africa with my son in January / February to watch cricket, parasail, swim with sharks, jump off 700' bridges with a rubber band fastened to my legs, taste wine, see wildlife including a lion which was rather neat and basically just enjoy life. Retirement was a real possibility at this stage, with a trip to Queensland in July next on the agenda, but then two things happened. 

One was I got promoted at work, and two, the pandemic hit which meant the Queensland tickets were cancelled and there was little point in retiring if I couldn't do any traveling anyway.

My job isn't really work these days. It's all meetings and management and if I can do it full-time from home, and by full-time I often mean for two to three hours a day, then I might as well hang in there until the next bonus and stock options are doled out at least, which is next month. Were my job hacking away at a coal face, my attitude to retirement might be a little different.  

So since March, like many readers I suspect, I have mostly worked from home, with the exception of three business trips later in the year, meaning a large saving in travel costs although slightly offset by now having to pay for my own coffee! 

More significant than the cost saving perhaps is that of the extra time I now have to myself, rolling out of bed five minutes before my first meeting and sliding my chair across the floor to my non-work desk within seconds of the last meeting ending. Of course, I have too much integrity to ever think of looking at personal projects during work hours.

Financially this year has been second only to 2013 with the US Indexes yet again leaving those of the UK in the dust, as I've mentioned before on this blog. This year the difference between the two was the largest ever, well at least since 1992, at more than double the difference in 2013.
Worth pointing out that £100 invested at the start of 1993 in the FTSE100 Index would now be worth £93.23, while $100 invested at that time in the S&P500 Index would now be worth $255.65.

Note that my sports investments these days are a very small part of my portfolio, with most of my money tied up in US-Index funds, property, and with play money invested in companies I've mentioned in this blog such as Tesla, Lloyds Bank, Boeing, and Pfizer, plus a few I haven't. David Sumpter suggested that:
Well, it's for old men too, or at least this one. Hard to argue against a stock that increased by 652% in a year. Bitcoin's mere 260% gain on the year seems rather tame by comparison, plus it seems that not only is it still incredibly difficult to use, but is also open to unregulated abuse as told in this sorry tale here

Overall, it was a volatile year for me as I expect it was for others. Q1 was the worst in spreadsheet history, with a 10.2% decline including my all-time record losing day of 16th March when I 'lost' six figures for the first time. 'Lost' is in quotes, because it was a paper loss only. I didn't sell anything, which is when you would actually take a loss, but it was a bold red entry on the spreadsheet and a record I don't want to see beaten any time soon. I've been around long enough to know that such days historically present buying opportunities.

In complete contrast, Q2 was the best in spreadsheet history, +14.3%, while Q3 was relatively quiet with a 7.7% gain. Q4 was the best of all, +13.2%. Six figure gains in both November and December made this the best quarter in real terms, if not quite by percentage.

As for the sports betting systems, the very selective EPL "Toss-Up" Draw System had a great 2020 winning us 12.05 points, but with just 25 selections, the 48% ROI is rather meaningless. With the over-round adjusted to a standard 3%, this system since 2000 has an ROI of 7.8%is overall, but 23% when the favourite is the Away side (or both Home and Away). Results are also improved by ignoring mid-week matches, but that's for another post.

Unfortunately the 'Close' selections ended the year with a small loss of 3.5 points from 64 selections, falling victim to the empty stadiums and a 15 game losing sequence in which the average goals per game climbed above 2.9. 

The Small Road Dogs in the NFL went 34-23-1, while the NBA version of this method was 56-47. Adding in the Time Zone filter I mentioned a few days ago, and the record improved to 21-13. 

Overs on high totals was again profitable, with around 54% or 55% depending on your entry point, but the one sport that was terrible last year was baseball.

The season was drastically different to usual, a mere 60 games in the regular season which didn't start until July. The long profitable T-Bone System lost between 8.8% and 10.5% depending on whether you played Straight Up or the Run Line, while backing "Hot Favourites" lost 4.8%.
The lines may yet move, but currently it looks like the final two matches in the AFC East on Sunday will be the final selections for this season's Small Road Dogs. 

Finally, a big thank you to @beigemartin who let me know that my latest post had been reproduced without permission or credit on another blog. Perhaps not the biggest issue in the world, but annoying that someone thinks this kind of behaviour is acceptable. At least they were quick to remove it when I requested them to do so. 

And with that, Happy New Year. 2021 will be the fourteenth year for this blog.

Holiday Draws

Did you lose money backing Home teams in the EPL over Xmas? You probably weren't alone.

Blindly backing Home teams over the holiday period (defined here as matches played from the 23rd December to the 3rd January) is historically not a good idea, with an ROI since 2000 of -7.5% from 621 matches.

As you may have guessed, this observation came to light while I was looking at whether the Draw was value at this time of year, the rationale being that with people taking holidays, and matches coming thick and fast and with less time to evaluate recent data, there may be opportunities.

I'm pretty sure I have covered here before how certain times of the season are better / worse for specific outcomes, looking at the results by month, but I mentioned a couple of posts ago some work done by Dave Holdsworth, who has taken this idea a step further, and analysed the data by Week or Round and come up with a simple strategy. "Simple" here is defined as easy to follow or use rather than a reference to the work behind it which is extensive and anything but simple.

For example, Dave's research suggests that backing Home teams between weeks 27 and 32, which he defines as "Crunch Time", is a profitable strategy, and accompanies this suggestion with a rationale for why this might be so.

Here are the Draw numbers through the end of 2020 by month, on the left all Draws, on the right, the Draw in "Close" matches only:
Blindly backing the Draw is clearly not a great idea in general, although it does appear to have some merit in January. Of course, we know by now that in most matches, the Draw should be ignored, and the matches where the Draw price is too long are those where the win probabilities of the two teams are relatively close. 

Last season was obviously an outlier with its lengthy suspension and ultimate return with no fans in attendance, and this year the latter continues to be an issue as well as the late start meaning that rather than being halfway through the season, we are a few games behind schedule. We may fall a few more games behind schedule if the uptick in COVID related postponements picks up, but that's another topic. 

Back to the Draw, and clearly in the first half of the season, August through December, the 'sensible' Draw is where this method should be applied, with an ROI of 13.9% from 873 matches.  

So I was wondering what happens in December / January to cause this second-half of the season decline. Obviously there's the Festive Fixtures, the onset of Winter, and fixtures are now between teams that have faced each other already that season. 

Over the last 20 seasons, the halfway point of the season is, with three exceptions in early January, reached in late December (see left). The actual mid-point is between the 28th and 29th of December.

Promoted teams have lost their element of surprise, and perhaps there's a feeling of rounding the turn and the second half of the season is where things get really serious. There's also the distraction of the FA Cup and for some the League Cup is still of interest.

I looked at the Festive Period and broke it up into a Xmas period of the 23rd to 28th, and a New Year period from the 29th to the 3rd - two six day periods. 

Here are the results as the season ends one half and starts the second half.
The sample size, even over twenty years, isn't huge, but there is a clear difference in returns.

Comparing the two halves of a season in their entirety shows these differences even more starkly:
In case you were wondering, for all Draws the first half of the season has an ROI of -2.95%, while the second half is again worse at -5.31%. Like I've said, don't back silly Draws.

Not all of you are interested in Draws, so what else do we see when we break a season down? 

For Homes, the second half of the season is where fortunes can be made.