Wednesday, 30 April 2014

Virtu Reality

The appearance yesterday of the U.S. Securities and Exchange Commission Chair Mary Jo White before a House of Representatives panel reminded me that a few weeks ago, the eloquent and well educated GAIN Capital trader Matthew Trenhaile challenged me to write a post highlighting the benefits of HFT (High Frequency Trading) and of its cousin in the sports investing world of CST (Court-Sider Trading). 

It’s a tough task, up there with trying to justify a belief in one specific god or in support of creationism. The evidence just doesn’t support the claims for any of these, and with trolls already only too willing to quote me out of context, the best I can do is highlight the claimed benefits of HST and then shoot them down.

Incidentally, there is definitely a god, and it’s the one true god that my parents told me about and everyone around me supports the same god and it says so in the Bible so it must be true. The other 3,000 or so gods of recorded history are just made up and silly and if you don't believe me, you are all going to hell. 

A convincing argument? No, me neither. I feel the same way about the arguments in favour of HST and CST which usually boil down to “added liquidity” and “narrower spreads”.

HST - “It's really a tool that has benefited the investment community”, says Peter Nabicht of Modern Markets Initiative. Nabicht is a senior advisor at MMI, a firm "focused on demonstrating the benefits of algorithmic or quantitative trading" and claims that HFT has lowered costs, tightened spreads and added liquidity to the markets, adding that "not all high-frequency trading is predatory."

HFT in itself is not a problem. The more people or companies trading a market, the more liquidity there is, and the tighter the spreads. However, the issue highlighted in Michael Lewis' book "Flash Boys: A Wall Street Revolt", is that of front-running, “in which the firms are able to quickly identify an investor's desire to buy stock, rush to buy it first and then sell it back at a higher price.” Lewis says:
"the market moves at two speeds: one speed for people who pay for access to the exchanges, who put their trading machines right next to the black boxes … and everybody else. And we are everybody else, everybody else being investors in the stock market."
In other words some traders are able to “have an early sneak-peak at data” which is quite different to having the ability to “react more quickly to publicly available data”.

The similarities with court-siding here are clear. Mary Jo White says that it “is not unlawful insider-trading” and that may be technically true, and court-siding is not breaking any rules either, but as the FBI, the U.S. Attorney General, New York state prosecutors and the SEC have all confirmed, they are investigating the practices of high-speed firms, and for a reason, which is to ensure that the markets are not seen as ‘rigged’.
"the staff, at Chair (Mary Jo) White's direction, is conducting a comprehensive data-driven analysis of a range of market-structure issues, including high-frequency trading practices and their impact on the fairness, efficiency and integrity of our markets."
When a company loses money on one day out of 1,238, as Virtu Financial reported in their accounts, it is clear to any reasonable person that they have an advantage, and by definition others in the zero-sum market are thus at a disadvantage. It would be irrational for someone at a disadvantage to continue trading under such circumstances, and while not everyone is rational, the result of losing money consistently will ultimately remove all but the most stupid, stubborn and deep-pocketed from the markets. New entrants will come and go of course, but are unlikely to remain for too long once they realise they are at a disadvantage, so it is reasonable to say that in an unfair trading environment, liquidity will ultimately decrease.

Active in the financial markets are there are those who are there to invest and those who are there to trade. It’s probably true to say that the impact of HFT to the small investor is minimal.

Sports markets are a little different, but if you think of the pre-in-play punter as an investor rather than a trader, the parallels continue. The punter enters the market before the shenanigans of the in-play take place, and either wins or loses.

When the market goes in-play, the court-sider has an immediate and persistent advantage. They may not be able to match Virtu Financial’s 1,237 profitable days out of 1,238, but in the long run their advantage will be profitable at the expense of others. Air fares, accommodation and expenses for a trip to Australia do not come cheap.

When Mr. Nabicht says “I think there's a confusion here. I don't think high-frequency traders, professional traders, are competing with the general public. They compete with each other but not with the general public” he has a valid point.

In sports markets though, the majority of participants are not professionals. There are a few full-timers out there, and as the Australian Open revealed, at least one enterprise willing to pay for “an early sneak-peak at data” the rewards from which clearly justify the expense involved.

As with the financial markets, once it becomes clear that they are trading at a disadvantage, and it should be clear from the price movements just how much of a disadvantage they are at, any rational person would step away, and liquidity will decline. The market may not be ‘rigged’ but it is not a fair one. 

Unfortunately, the inherent nature of betting means that most bettors do not act rationally. That markets with court-siders active in them are as liquid as they are is testament to that. Again, some new blood will always be coming in, but ultimately the liquidity will fall perhaps leaving a situation where court-siders are competing with themselves.

As for the argument that spreads are tightened, this is certainly not true in the sports markets. The presence of court-siders means that they are, or should be, the only ones willing to front-run the markets, and they don’t look to back at 1.71 and lay at 1.7. The spread is more like 1.8 / 1.6. Should the non-court-sider wish to back at 1.6, he is either getting poor value (but may of course still win on occasion) or he will not get matched because the court-sider has had a sneak-peak and adjusted his prices.

The investigation into the financial markets continues, while Betfair and the Gambling Commission continue to drag their feet, allowing a few individuals, including former employees, to clean up as they slowly suck the in-play markets dry. After all: who would want to play in a casino so explicitly rigged?

Tuesday, 29 April 2014

Defending The Draw

It was another 'below expectation' weekend for draws across the top five leagues in Europe with just 11 from the 49 matches, four of which were in the Bundesliga. This follows a round of 11 from 48 matches, and the 2013-14 strike rate for draws continues to be below long-term averages in all leagues:

Danny Murphy questioned (very politely) my assertion that the reduced number of draws in the leagues I track, relative to their long-term averages, is a valid excuse for a negative ROI on my draw selections.

He wrote:
Still trying to get my head round your comments about the lack of draws in the five leagues.
It's hard to explain but how can you be sure about the link between the lack of draws in the matches you have selected and the lack of draws in the league in general? I don't see why a lack of draws in the league would affect your selections. Once your selection has resolved itself the other results won't matter. If I back Man U to win at home and they draw what does it matter if home wins have been less in the PL this season? I don't see any relationship between one and the other. Yet I have to admit your excuse sounds highly plausible until you really start thinking about it.
In a way you are saying you aren't doing well because draws have a random element to them. But I say don't back the draw for exactly that reason, they are almost impossible to predict and it's going to be very difficult to find much of an edge.
My reply was:
I understand where you’re coming from, but the issue for me is that the predictive model (it really does sound pretentious when all it is a spreadsheet) is built on Elo ratings and data from previous seasons in the leagues I follow. The idea is that over time, 10 years plus, the numbers of Home wins, Away wins and Draws, as well as goals scores tend to be remarkably consistent within each league.

Each national league has its own ‘personality’ and this is factored into the model. So the model makes a broad assumption that in say the EPL, the Home wins will (over time) continue to be around 47% of matches, with the Aways around the 27% mark and the Draws 26%.

(There’s a reason why there are usually around 14 draws each week in the XX Draws selections).

Over a season, this assumption would usually be reasonable, but in an extended period such as this season, the template for the model simply isn’t a good fit, not just in one league but in four of the five. The model would slowly adapt, but more likely is that the reduced percentage of draws is not a long-term trend, just one of those outliers that come along every once in a while. It’s just very unusual to see the same trend across all the leagues in the same season.
I went on a bit more about specific numbers, (it made me feel better but probably bored Danny to tears - he's not written to me since), but if we take an extreme example, if this season had been one where the draw percentage was say 33%, i.e. considerably higher than the model 'expects', it is almost certain the XX Draws would have looked great, and almost as certain that I wouldn't have paid so much attention as to why.

As for Danny's last comment, I agree to a point. Draws are often random, but if the selections are games where the 'true' goal expectancies are lower than the market's expectation, the draw will be value. The 0:0 is the perfect draw, the 1:1 isn't too shabby, but I don't get too upset about the 1:0 or 0:1 results. The randomness that makes football such an exciting game means that while you can't predict exactly which matches will result in draws, you can predict the matches where it is more likely that the prices suggest.

Certainly 3:3 (or higher scoring draws) are fluky and the 2:2 result doesn't exactly fill my heart with joy, but the 0:0s and 1:1s are pleasing enough.

Given the efficiency of the match odds markets, any edge is always likely to be small, but the pool is a reasonable size for a part-time investor - 1,826 in the big five, but many more if you care to include the lower leagues or some of the other leading leagues.

The big question for me is whether the lack of draws this season is an anomaly or the start of a trend. More goals means fewer draws as I have written before, and it is a fact that there are more goals being scored this season in all five top leagues than their 10 year averages. What is not clear to me is why this is so.

No Place To Hyde

With the finishing line in sight, here are the numbers from the latest round of matches. The top six 'in-profit' entries remain the same, although Webbo dropped a couple places with the XX Bundeslayga and Fedslam selections moving up one place each.

The one Cassini Value Selection won for a profit of 1.47 points as Sampdoria (priced by my spreadsheet at 1.96) came from behind to beat Chievo Verona with a 90th minute goal. Skeeve went for an unusual final day selection of Overs in the Hyde v Luton Town match. With Hyde already guaranteed last place on 10 points before the game, and Luton guaranteed first place on 98 points, this game must be a record for a league game points differential, or close to it (excluding point deductions). Anyway, a day for Overs it was not as Luton ended their Conference career with a 1:0 win. The five Bundeslayga qualifiers produced three winners for a 2.04 point profit, Fedslam was inactive and Webbo had just one winner from ten selections dropping 4.72 points. 

There are now three entries in the 'close-to-profit' part of the table, i.e. down by single figures, and they are:
The XX Draws (Unders) had a good weekend with 10 winners from 14 selections, and a profit of 5.31 points, while The Football Analyst continued his strong finish with another 3.30 points. The liability remains at £125.

And finally, the red section of the table: 
A 2.89 point profit (5 winners from 14) for the XX Draws moved these up two places and for the first time (I think) this season, all four Cassini categories were in profit on the same weekend. Peter Nordsted's Drawmaster found two draws from four selections and made 3.02 points. Football Elite (2.00 points) and Punter's Friend Neil (1.15 points) were also profitable, while Jamie A, Rubicon, Peter Nordsted's Premier Betting and Fairfranco all made losses of varying size).

Thursday, 24 April 2014

FTL Easter Update

The Easter weekend was a great one for The Football Analyst, who moved seven places up the FTL table after making a profit of 14.06 points. There were no Pinnacle prices for Tamworth's win at Dartford on Easter Monday, so I compromised and took an average of the seven books that did have prices for this game. Those seven places mean that Graeme reduces his liability to £125 and another weekend like last and he'll be in the green.

Not much in the way of changes in position other than this one, and the top six (all in profit) all held their positions. The Cassini Value Selections dropped 3.6 points after identifying value in five games where the selection had nothing to play for, and the only winner game in game six, where the winner (Everton) DID have everything to play for. Skeeve made small profits, and looks well placed for the first cash prize, while Webbo lost ground although still has the advantage in the race for second cash spot just ahead of the XX Bundeslayga selections and Fedslam.

Here are the profitable entries:

 And here are the not-so-profitable entries:

Sunday, 20 April 2014

Motivation And Perception

I am off to San Francisco tomorrow, so expect a delay in this week's FTL update and posts in general. A quick look at today's results show that the Cassini Value Selections had a bad weekend with all four selections losing. Unfortunately at this time of year, motivation is a factor and three dead-rubber matches in Serie A all lost and Levante, with nothing to play for, where held by Getafe, with everything to play for. Whether the motivation factor is real or an example of perceived wisdom being wrong is a good topic for debate. Football Elite's Matt produced some numbers in his email this week that it is the latter, but the minuscule sample size today didn't add anything to that finding unfortunately.

Not too many draws still, (8 from 28) but the XX Draws and others had a few hits in England, Wales and Spain today. On last week's Draw update, Betslayer commented:
You have missed off sotdoc / sotdraw, how did he do?
Although these selections are not in the FTL, I did take a look and it was another losing weekend for them - down 0.7 points for a running total of -25.19 and a negative ROI of -33.14%.

Thursday, 17 April 2014

FTL Update 16.April.2014

A little later than usual, but I thought I'd include the midweek matches in this update too. Without further ado, here is the profitable section of the FTL:

Only one change in position, although Webbo was saved by the draw at Manchester City tonight. The Cassini Value Selections had another good weekend making 4.83 points from eight selections.
It's a dangerous time of year with motivation questionable, and there was an interesting exchange of views before the games about the merits, or otherwise, of some of the selections. A 28% ROI from 100+ selections is far above my expectations, and with 5 of the 6 wins being by one goal, perhaps luck was on my side last weekend. An early dismissal at Swansea City certainly helped.

Skeeve came up empty, but the loss was limited to 2 points by the FTL format. Small losses for XX Bundeslayga and Fedslam.

The red section looks like this:
Hofs Hackers were the big movers, moving up two places, but also a good weekend for Peter Nordsted whose Drawmaster selections were up 1.42 points (and up one place) and his Premier Betting Account Bets moved off the bottom losing just 0.25 points. The Football Analyst made a small profit and needs a strong finish to trim the bounty liability which remains unchanged. 

Not FTL related, but some of you might be interested in this article I wrote last December explaining why Crystal Palace would survive. 

Sunday, 13 April 2014

Draw Focus

A couple of Monday night games still to come, but even if both end as draws it is still yet another dry weekend for Draws. In England, there was one from eight matches, Germany saw one from nine, and there were two draws in each of Italy (from nine games) and also France (from ten), and a massive three draws in Spain (from nine games). Nine from 45, one perfect draw, and the averages continue to decline.

As a group, the draw-hunters did quite well considering the unfavourable market conditions. Peter Nordsted's Drawmaster had just two selections, one of which was the game at Celta de Vigo. The draw at West Bromwich Albion was included in Peter's Premier Account Bet selections for a rare winner. The XX Draws found winners at West Bromwich Albion, Ajaccio, Reims and Bologna but the king of the draws this weekend is Jamie A who found all three in Spain plus in the game at Sassuolo from nine completed selections. He also has the draw (and Unders) in the match at Athletic Bilbao tomorrow night.

More of an update later when the prices are available.

Friday, 11 April 2014

Intrade Tradesports

US Presidents 44 and 45
I am indebted to Scott for this link to an interesting article at Business Week about Intrade, and their plans to return to life as Tradesports, focusing on sports rather than politics and more financial based markets this time around.

Intrade was riding high in November 2012, as its user base correctly predicted the presidential vote in every state but Florida. Less than three weeks later, it fell off a cliff. The U.S. Commodity Futures Trading Commission sued the company, claiming that its exchange illegally enabled speculation on the value of gold and currencies and on the probability of acts of war. Intrade’s legal problems quickly drove away users fearful of being dragged into the proceedings; within months the company could no longer cover its remaining customers’ deposits. It shut down in March 2013.

A year later, co-founder Ron Bernstein is putting the Intrade shingle back up. The lettering is different, though; in lieu of the political betting that made it famous, Bernstein is restructuring his efforts around a company called Tradesports that allows speculators to buy and sell positions on sporting events. The company is starting a private test period for the site on Monday and plans to make it publicly accessible soon. “Sports are always going to be the best laboratory for prediction markets,” he says. “All those things now are just recorded and data-mined and discussed.”

The focus on sports has particular appeal for a company that was originally located in Ireland because of its market’s questionable U.S. legality. Bernstein wants to take advantage of a 2006 law that exempts fantasy sports from federal prohibitions on sports gambling. So long as fantasy contests can be considered games of skill that don’t rely on the outcome of a single sporting event, people are allowed to play them for money. Bernstein, who left Intrade in 2003 because he felt he had to choose between running the company and living in the U.S., will run the new site from New Jersey.

Contests on Tradesports will consist of a series of markets, where players buy stock based on events within a game. To not run afoul of the law, players cannot simply buy stock on whether the San Antonio Spurs will win; the position must also include a bet on, say, how many rebounds Tim Duncan will pull down. A pot is set in advance, and the stocks can be traded throughout the game. When the final buzzer sounds, players holding stock with the highest combined value win the pot. The company takes a cut of the entry fees.

There’s already a cottage industry of markets for so-called daily fantasy sports. These companies market their activities as coming with the satisfaction of gambling without the legal risks. The leading site, FanDuel, has raised $18 million in venture capital and says it became profitable late last year. It expects to pay out $400 million in prizes in 2014. Another competitor, DraftKings, has raised $35 million and says it will pay out more than $200 million in prizes this year.

Intrade’s most prominent investors once included Wall Street legend Paul Tudor Jones and Lachlan Murdoch, Rupert’s son. They’re long gone, but Bernstein raised some new funding last year and says he has enough capital to run the site for at least 18 months. While Tradesports is technically a different company from Intrade, its staff includes several Intrade veterans and it’s built on the original site’s technology. Bernstein will run both companies. He’s trying to stock the Tradesports pond partly through e-mail blasts to Intrade’s 200,000 or so past users. He’s also pitching the new site partly as a spectator sport, where the betting action results in accurate predictions about playoff wins and draft picks.

There may be something to that. While there’s little reason to look at other daily fantasy sites if you’re not betting, the quality of Intrade’s predictions was key to its prominence. When traffic was at its peak, the site’s audience was five times its trading pool. Intrade was a particular darling of economists, who use prediction markets to study the way markets operate or to measure the potential effects of a presidential debate or a surreptitiously recorded fundraiser. “We’re using prediction markets for two distinct purposes, neither of which is to gamble,” says Dave Rothschild, an economist at Microsoft. Whether sports will hold as much interest as politics is an open question.

Tradesports marks a return to Intrade’s roots. Bernstein became involved in Intrade when an investor from his previous startup shared the idea of a stock market for sports. No one involved quite knew how to create one, and Bernstein agreed to help out, moving with the other founders to Ireland before Intrade went live in 2001. Bernstein soon realized that a company based on sports wagers ruled out a return to the U.S. He left the company and came home to start a new career trading gold and sugar futures.

Lucky Fools And Rats

I had a purge of some obsolete / inactive blogs from the blog roll yesterday, and couldn’t help but wonder why so many betting / trading blogs fall by the wayside. The fact that any blog is ever born, means that the creator was at one time full of enthusiasm, excitement and positivity. Perhaps that moment was very brief, resulting in some blogs being created on a whim, while others might be started after a little more consideration. Either way, a lot of them seem to fall by the wayside fairly quickly. It’s been a while now, but at times in the past emails requesting to “please add my new blog to your blogroll” would come in fairly frequently.

The betting / trading blog world is in the midst of a dry spell, unless there are a whole bunch of new ones out there I just don’t know about. Why there are few new ones coming along is a different topic, but why do existing blogs of different vintages die, often after a very short life?

Apart from the possibility that the writer dies, or is somehow no longer physically or mentally able to post entries, (or Mrs Cassini says - in prison) the reason is simply that the writer has lost enthusiasm. I suspect the main trigger here is that the writer starts to lose money. What might be a fun hobby when you are winning is not such a fun hobby when you are losing and nothing dampens enthusiasm like losing money. Even if the blog isn’t about profit and loss necessarily, it is still understandable how enthusiasm for posting would be diminished. Being the sad person that I am, I maintain a ‘net-worth’ spreadsheet, and on an ‘up’ day I can’t wait to update the numbers. On a ‘down’ day, the enthusiasm diminishes significantly.


There are other reasons why enthusiasm might be lost. Writing posts, at least the more in-depth, wordy quality type of post that you find on a blog like this one for example, takes time. If you like quantity rather than quality, then there are probably blogs out there, but the formulaic posts or poorly written ones soon become tedious, at least to me, and the blogger ends up talking to himself. For my style of blog, an average of four of five posts a week is probably enough for most people. Blogs, like Tweets, can get boring fast. Again, it is quality rather than quantity that should be important. Time is precious, and priorities change as life and circumstances change, and even for me, blogging has to fit into my schedule rather than vice-versa.

Another reason why the non-formulaic, tipping or P/L blogs cease is that ideas simply dry up, but when you have a quality readership making intelligent comments and offering ideas and suggestions, this is not a problem, which brings me to a few comments and ideas suggested by a few of you.

Matthew commented on my Flash Back post comparing High-Speed Trading (HST) with Court-Sider Trading (CST):
I wonder if you would be interested in composing a devil's advocate article to this one. Maybe list some of the articles that support HFT and make a case for it and for court siders at tennis matches. I am aware that you may feel that there is not a credible counter argument but I would like to see you challenge yourself by presenting the other side of the debate.
I do love a challenge, but this might be a big one even for Cassini. Stand by though for a considered post on the subject in the next few days, and thanks for the idea Matthew.

I read Graeme’s post early yesterday, and was half expecting to see my name in there somewhere given that the theme was very similar to my own benchmark posts of recent days, but he did comment (fortunately for Blogger and their storage needs, with a little more brevity than his actual post) on my Benchmark post:
It’s interesting to see some of us scurrying around trying to look at underlying results and benchmarks after a poor season of results! As I’ve admitted a few times now on my blog, I’ve never really paid close attention to underlying results over the first 3 seasons as I had no reason to. With hindsight, even when you are making money betting, it’s probably not a bad idea to check that you just haven’t been operating in very favourable conditions but I guess it’s human nature to ignore these sort of things when things are going well. :)
I’ve actually been working on a blog post which is now on the blog which is on a similar topic as this blog post you’ve written.

I think the key learning for me and everyone else following my systems this season is that a season isn’t as long as any of us thought. Before this season, I would never have contemplated a losing season happening to my systems as it would have taken such a massive drop in profitability to go from the returns over the last 3 seasons to actually making a loss over a season. I struggled to imagine having a losing system at football, never mind a losing season for all system results combined but I’m now starting to face up to this reality this season unfortunately.
I think my blog post on the underlying results shows how fickle this game can be and ultimately, if you experience quite small % differences in the long-term averages of Homes/Aways/Draws in any league or any sample of bets (like backing all teams priced 2.50+), then it has a huge impact on the results you can achieve. You can escape in a month or a few months but if you are against the bias the whole season, it is difficult. You are finding this yourself as is Matt at FE I think.
Of course, there are ways around this and spreading the risk to ensure you aren’t too exposed to a particular group of bets is very obvious with hindsight but ultimately, I didn’t think about this during the first 3 seasons! Having 70%+ bets on longer priced aways seemed like a great idea based on the first 3 seasons and although it’s a rollercoaster due to the average odds you are backing at, over a course of a season, I thought things would always level out and a sample of 1,000+ bets was big enough to cover any variance. As my results show this season, that isn’t the case!
I don’t really believe my edge has vanished overnight and ultimately, I think there are a number of reasons my ratings have struggled this season but I’m sure I’ll learn more from this season than I learnt over the last 3 seasons. My only issue is that I need to decide whether I can start again basically and build something new based on what I’ve learnt over the last 4 seasons as what I have now isn’t fit for purpose in my opinion. That’s not to say there is anything wrong with all my systems but quite simply, the Barcelona of football systems should not be fighting relegation in your tipster league! End of. :)
In terms of your comment about tracking my results, I’ve said this before I think but my results are the easiest to monitor! Every system I have is proofed to 1pt level stakes and therefore, if anyone wants to see how the systems are doing, they only have to read my blog. What I can’t cater for is people cherry picking systems and creating their own portfolio of systems and creating profits/losses that look nothing like what others have experienced but that’s the beauty of TFA. Until this season, the returns varied for everyone but everyone had to make a profit as every system was profitable. This season, with a mix of profitable and (more predominantly) losing systems, the range of returns is as wide as it has always been for subscribers but unfortunately, some will be winning and some will be losing. That’s life I’m afraid.
I thought I’d be smart ass this season and create a unique system to go into the tipster league as I honestly thought it was impossible for me to build a losing system at football betting over the season but I no longer believe this is the case. My ego is back in check! With hindsight, I should have plucked one of my high turnover systems and just sent you the bets along with the rest of the subscribers each time but I wanted to possibly proof a new system. Instead, I’ve proofed and proven I can build a losing system and a really bad one at that!

If I give this another go next season, I will try to cater for the people who may only want to follow a traditional tipster service and just place bets on teams my ratings think offer value! A little bit different to how I’ve worked until now but it was easy when people were winning all the time, much trickier when some are losing I’m finding! It was impossible to lose for 3 seasons following my systems….too easy this season though!

As always, an interesting post.
Graeme has this to say about tracking just how good his results really are:
In terms of your comment about tracking my results, I’ve said this before I think but my results are the easiest to monitor! Every system I have is proofed to 1pt level stakes and therefore, if anyone wants to see how the systems are doing, they only have to read my blog.
The problem for many of us is that we are not interested so much in System 36B-1 or System Alpha-Centauri-X-Plus or System Since 1905 Hull City Have Never Beaten A Team Whose Name Starts With A Vowel On A Tuesday Night, because, as I understand it, the same selection can be in one or more systems. What people want to know is what is the return to a level (one point unit stake) of all selections (no duplicates), and that is what the FTL does. It removes fancy staking and multiples and duplicates, and just takes each selection a tipster offers one time, one unit bet at Pinnacle Prices. Skeeve comes out looking rather good, Peter Nordsted comes out looking rather bad, and everyone else is in between, although it has to be said, tending towards the bad.

Graeme also mentions that “I’ve never really paid close attention to underlying results over the first 3 seasons as I had no reason to. With hindsight, even when you are making money betting, it’s probably not a bad idea to check that you just haven’t been operating in very favourable conditions but I guess it’s human nature to ignore these sort of things when things are going well."

I couldn’t agree more, and it is reminiscent of the day-trading boom at the start of this century when everyone was a stock picking expert. Acquaintances with real qualifications were quitting their jobs to day-trade full-time. Special day-trading centres were springing up every week; it was like picking up money from the street. The problem was of course that my chemist friend had no edge. He, and thousands like him, tricked themselves into thinking they had uncovered some hidden talent when it was all down to favourable conditions. Stocks in general were in a boom, a great ride while it lasts, but booms don’t last. They bust, and when the inevitable bust came, day-traders all but disappeared, and acquaintances (I don't have many friends) went back to their chemist careers or whatever.
  
In hindsight, all very obvious, and as Graeme says, “it’s human nature”. In his book “Fooled by Randomness”, Nassim N. Taleb calls these people ‘lucky fools’ and psychologists call it “the self-attribution bias” – i.e. when things are going well, we take credit for it, when things go wrong, it’s outside forces to blame. As Mrs Cassini would be all to happy to confirm for you, it's one of my faults. Not so much now, but in the early days, a bad trade would be down to her making too much noise with the dishes, the dogs barking, Winston (now fully recovered by the way) chirping too loudly, or she (Mrs C - Winston is a boy) was slow to bring me my water (Winston doesn't bring me water - at least not in drinkable form) before it was ever down to me making a poor decision. The good trades were all down to my brilliance as an expert analyst. Mr. Taleb wrote:
Lucky fools do not bear the slightest suspicion that they may be lucky fools — by definition, they do not know that they belong to such category. They will act as if they deserved the money. Their strings of successes will inject them with so much serotonin (or some similar substance) that they will even fool themselves about their ability to outperform markets (our hormonal system does not know whether our successes depend on randomness).”
Outperform markets” is the key here. To use draws in football as an example, (an example close to my heart), if you choose a league which has a long-term average draw percentage of 30%, and pick a sub-set of matches to end as draws one season, at random you will almost certainly make money if the draws make up 50% of the results that season, and you will almost certainly lose money if the draws drop to 20%.

TFA - Two Furry Animals
Whether the selections are fantastic or useless cannot be measured by profit or loss, but as Graeme and I have both realised, it is how those selections measure up against a benchmark. For a rigid system, e.g. looking for a subset of draws, home teams in the 2.0 to 2.5 range, or teams who drew 1:1 on this day five years ago, it is a pretty simple process to establish such a benchmark and it can easily show the merits or otherwise, of a selection process.

If the selection process isn’t a rigid one, but flexible (i.e. the nature of the bet is not pre-determined, could be home, away, draw, under, over etc. based on value), a benchmark won’t work. For whatever reason, your model isn’t working. This could be a short-term hiccup or it could be that something has changed.

The other line of Graeme’s that jumped out at me was the reference to us “scurrying around” which rather brings to mind the image of black rats (in coal cellars?), which I’m not sure is quite the image we are hoping to project here! The truth is that as Graeme says, neither of us had any reason to look behind the numbers before. For me, it was only in the last few weeks that I realised this was an unusual season for draws. One reason for using five leagues is that it typically generates a reasonable number of selections each week – my idea of reasonable might be different to Graeme’s though, who’s idea of reasonable is 794 bets a week! – but a side benefit is that the overall average across all five leagues is more likely to reflect the long-term average. A slight increase in Italy may be balanced out by a slight drop in Spain, but as I keep mentioning, this season all five leagues are down on their 10-year averages.

Looking back, admittedly belatedly, the 10 year draw average (2001-11) in the top five leagues was 26.66%.

In season 2011-12, the draws hit at 26.56%. Down on the long term average, but only slightly and not by enough to erode the edge the XX Draws had that season. The league averages were down in England, France and Spain, up in Italy and Germany.

At the start of the 2012-13 season, the 10 year draw average was 26.78%, with the draws that season hitting at 25.96%. Down again, but again, not by enough to push the XX Draws into the red. League averages were up in England and Germany (again), down in France, Italy and Spain.

The 10 year average at the start of this season was 26.70%, and at the time of writing, draws are hitting at 23.79%, significantly lower, and down in all five leagues although Ligue 1 is very close.

With only 4 of the 15 league averages up in three seasons, it’s clear that it has not been a good time to start looking for draws, but if the benchmark is being beaten, i.e. the underlying selection criteria is performing better than the averages suggest, then the optimist in me looks forward to a stellar 2014-15 as the averages bounce back. More on this at the end of the season, if not sooner. That you can bet on.  

Thursday, 10 April 2014

Football Funnies

1. "If that had gone in, it would have been a goal." - David Coleman 
2. "I never comment on referees and I'm not going to break the habit of a lifetime for that prat." - Ron Atkinson 
3. "For those of you watching in black and white, Spurs are playing in yellow." - John Motson 
4. "My parents have been there for me, ever since I was about seven." - David Beckham 
5. "Germany are a very difficult team to play. They had 11 internationals out there today." - Steve Lomas 
6. "They're the second best team in the world, and there's no higher praise than that." - Kevin Keegan 
7. "Marseille needed to score first, and that never looked likely once Liverpool had taken the lead." - David Pleat 
8. "If we played like this every week, we wouldn't be so inconsistent." - Bryan Robson 
9. "If history is going to repeat itself I should think we can expect the same thing again." - Terry Venables
10. "I was saying the other day, how often the most vulnerable area for goalies is between their legs ..." - Andy Gray 
This season's winner was Matt Le Tissier with:
"Who'll win the league? It's a coin toss between three of them."

Flash Back

The issue of High-Frequency Trading (HFT) came to the fore with the Flash Crash of 6th May 2010 and has since drawn further attention and debate with proponents saying that the practice adds liquidity to the markets while opponents argue that the practice leads to front-running, market manipulation and insider trading. These activities are now under investigation by the Federal Bureau of Investigation, coincidentally (or perhaps not so coincidentally) announced the day after the publication of "Moneyball" author Michael Lewis’ latest book “Flash Boys: A Wall Street Revolt” which concludes that the market is rigged by HFT traders. Incidentally, Michael Lewis has the ability to take a complex topic and explain it very clearly in layman’s terms, and his book “The Big Short” is one of the best (in my opinion) at explaining clearly the events leading up to the financial crash of 2007-10.

As Wikipedia describes HFT, it is:

…a type of algorithmic trading, specifically the use of sophisticated technological tools and computer algorithms to rapidly trade securities. HFT uses proprietary trading strategies carried out by computers to move in and out of positions in seconds or fractions of a second. Firms such as British-based Algorates, one of the first to employ HFT, rely on advanced computer systems and the processing speed of their trades to record high earnings in the market.
High-frequency traders move in and out of short-term positions aiming to capture sometimes just a fraction of a cent in profit on every trade. HFT firms do not employ significant leverage, accumulate positions or hold their portfolios overnight. As a result, HFT has a potential Sharpe ratio (a measure of risk and reward) thousands of times higher than traditional buy-and-hold strategies. HFT firms make up for low margins with a high frequency of trades.
If you have the money to invest in setting up your operation, it appears to be rather lucrative. JP Morgan Chase & Co. did not lose money on any single day in 2013 (when intra-day trading is included, and why they conceal these numbers is another issue), but if you think that is impressive, another HFT firm, apparently a ‘titan’ called Virtu, had just one day in four years of trading when they lost money.

As the always excellent web site zerohedge.com concludes:
It also explains why fundamentals haven't mattered in years - the only thing that does matter is to quickly open one's own HFT stop, front-run as much order flow as possible, and scalp pennies ahead of the bid and ask... billions and billions of times, leading to the statistically improbable chart pictured above.
This, ladies and gentlemen, is why retail has given up - when companies want to go public and no longer even hide the "secret sauce" which confirms beyond a reasonable doubt that there is a two-tier market: one in which the HFTs just never lose, and one for everyone else, well: who would want to play in a casino so explicitly rigged?

Now some readers will know already where I am going with this, but it is hard not to see parallels with the Court-Sider Trading (CST) activities which came to widespread attention at the Australian Open earlier this year.

The type of trading is similar, relying of speed rather than ‘fundamentals’, and as the currently unregulated practice becomes more widespread and well-known, the public will similarly no longer want to play in a casino, where the ‘rules’ are so lax that Court-Sider Traders seldom lose.

Perhaps I should write a book on the topic, and hope that the UK’s Gambling equivalent of the FBI will take a look at the practice the day after publication? But then I am not Michael Lewis. And the Gambling Commission is not the FBI.

Wednesday, 9 April 2014

Benchmark

It’s a curious (and mildly annoying) anomaly that the draw averages are down across all five leagues in the same season without an obvious common cause, and I find myself in a similar boat to Matt from Football Elite who, in his weekend summary email, which this week is actually more of a season-to-date summary than a weekly one, writes:

Obviously I totally understand that people don’t subscribe to a service to break-even but sometimes as a tipster things are out of your hands. This season when my kind of sides just didn’t start winning until February there’s not really much I can do.
He adds later:
I will defend my performance this season though as I think I’ve pretty much made the best of a bad situation and got the max possible in the circumstances.
I agree with much of this, although not all. For a start, I know that several subscribers to my XX Draws and More are using the draw selections to mitigate either current or potential Premium Charges on Betfair rather than to make their fortune. A low ROI on a large number of selections is what several people are looking for. It’s a paradox that if they are too successful, the Premium Charge issues are compounded rather than alleviated, but obviously making losses isn’t the form of relief we are looking for. Unlike Matt, who seems to have several Mr. Angrys among his clientele, I have yet to have one. 
My thoughts on this are that subscribers to a draw service understand that the nature of draws is such that long losing runs are inevitable, and that as the prospectus makes clear, they are not a ‘get rich quick’ scheme by any means. The limited feedback I have had, and my own experience, is that even in a disappointing season such as this one where the draws are mysteriously absent, relatively small losses at the official (but usually beatable) Pinnacle prices are even more modest if you use the exchanges or cherry pick the best prices with other sports-books, even if that is ultimately likely to be a short-lived opportunity.
As for Matt’s comment that “sometimes things are out of your hands”, this is true up to a point. Matt’s selections fall into a certain sub-set of matches, and as appears to be happening again this season, that sub-set is not proving to be profitable. A tipster can either ride out the storm, technically a drought I suppose, the keep calm and carry on approach, or he either change his selection strategy or abandon it altogether. One can become irrationally attached to stock selections, and I suspect that letting go of a previously profitable system is similarly harder than it should be, especially if you are selling them.

Whether selections are private or sold advice, a losing run doesn’t necessarily mean that the system no longer works. It’s why I have recently started talking about having a bench mark to compare your selections against, and why I think there is still some life in the XX Draws yet. The evidence (France excepted) suggests there is value still being found, and the below average number of draws across all five leagues in the same season is, I hope, a mere coincidence.

Matt could come up with a similar benchmark using the leagues and price ranges he typically picks from, and can tell whether he has been unlucky for almost two seasons, (i.e. his selections are beating the benchmark) or has the market adjusted to erode an edge that formerly existed? Are Matt’s selections beating the benchmark in certain leagues, but not others, or has the profitable target price ranges changed?

One problem when developing a model is that it can be discarded too soon. Two seasons of losses across all leagues and all prices is probably enough evidence that the edge has gone, but throwing everything away after a poor month would be a mistake.

Peter Nordsted’s Premier Account service is in a similar situation. As my post yesterday mentioned, assuming the official prices claimed are achievable, and with some of the books quoted, that would be quite a stretch for a winning account, a loss of 20.4% ROI over 265 matches is frankly terrible. Someone joked a few weeks ago that Premier Betting had two members, but that estimate will be two more than next season I imagine. The problem for Peter is that having taken in subscriptions, it’s hard to later admit that your methods have no merit.

Graeme, aka The Football Analyst, who after three winning seasons also finds himself struggling, is in a similar position to Matt, although his poor season is his first increasing the chances that it is just a blip. 

Without knowing how Graeme makes his selections, it’s hard to say if a benchmark could be created, and as has been much discussed already, simply finding out how his selections perform to a single unit stake isn’t easy.

One tipster who doesn’t have these problems, at least with selections this season, is Skeeve. Sure, his variable staking and multiples haven’t helped, but measured at Pinnacle prices as one unit singles, the results are impressive.

For myself, and the XX Draws, I shall ride the drought out (doesn’t sound quite right, but you get my drift), for this season and then take a look in even more detail. The pattern in Ligue 1 is bizarre.

Danny, showing a wry sense of humour as well as an impressive knowledge of my reputed ancestor’s work in France, (although I prefer to claim the Italian heritage rather than the French - they adopted him on 14.July.1673) and had this to say:
For a man whose illustrious ancestor Giovanni Domenico Cassini famously had the measure of France, your troubles in Ligue 1 do indeed disappoint.
Could it be that much like when British Rail had the wrong kind of snow, you are currently suffering the wrong kind of draw in France??? 
Very droll, but clearly the draw selections are falling in the wrong places across the Channel. The XX Draws in France have won 28 of the 93 draws available (30%), and those 93 came from 320 matches (29%). So far so good, but here is where it gets ugly. The average price of the XX Draw selections is 3.23, and the 28 XX Draw winners were priced at an average of 3.28, but the average for draws in Ligue 1 is 3.475. Too many XX Draw winners at the low end. I've written in my email to subscribers that the draw on some selections in France is too low. Over exactly 500 XX Draw selections in France, 155 (31%) have been draws, implied odds of 3.23. Backing only XX Draws at this 3.23 or above would have meant a profit of 8.98 this season still well below the benchmark of 29.02 for all draws 3.23 to 3.99. At the risk of sounding like Pete the Weatherman, 3.23 would appear to have some significance this season.   

As my model evolves with ever more data, and my ‘true’ odds calculations improve, the Cassini Value Selections will likely play a bigger role. The more recent subscribers have been more interested in the Cassini Value Selections than the draws (not surprisingly), but as I said earlier, I think those draws still have some mileage, perhaps with some adjustments here and there.
One final comment is that in 2012-13 I offered a money-back guarantee (well, 50%), if the XX Draws failed to reach a certain percentage ROI. It seemed a good idea at the time, and luckily for me the target was exceeded, and although I didn’t offer the same in 2013-14, it is something I think I will do again next season if I continue the service. 
Whether the trigger is falling below the pre-defined benchmark (has the word ‘benchmark’ ever appeared so many times in one post?) or a set points total or ROI% is to be determined, but I think anyone selling selections should be accountable, at least to some extent. A full refund is probably too much, since there is a lot of effort that goes in, for most of us anyway, and some pros might not approve of this idea, but for someone like myself for whom this is a semi-professional activity, I’d rather take a hit than feel like I was offering a sub-par service. As I said at the time, I'm not doing this for the money. I also think that picking the top price from Oddschecker is unscrupulous. I'd rather take a small loss at a verifiable but beatable price than inflate my numbers by using books that will restrict anyone who appears to know what they are doing.

Tuesday, 8 April 2014

FTL Update 8.Apr.14

The weekend’s 49 top European matches are now in the books, and yet another below average number of draws. England saw none, Italy and Spain just one each, Germany two (from their nine fixtures) and only in France did we see an above average four draws. Overall, 8 draws from our (ok, my) pool of 49 matches and just 16.3%. It’s a challenging environment for the draw seekers among us, with the season’s benchmark of backing all draws priced at sub 4.0 currently down -38.42

The XX Draws found just two draws out of fifteen selections and dropped 7.94 points. Peter Nordsted’s Drawmaster found one from 6 dropping 2.62 while Jamie A changed tactics this weekend and also went hunting for draws. He picked a bad time and hit one from 12, and lost 8.62 points. Unofficial contender SOTDoc also hit one from 12 and lost another 8.79 points and the ROI now stands at -36%
Anyway, enough about draws. Here are the updates to the FTL (sponsored by TFA). The 'in-profit' part of the table now looks like this:
Skeeve had a 100% weekend finding four from four and closed the gap, but the Cassini Value Selections held on to their lead finding five winners from eight selections and making 3.21 points. Fedslam dropped a couple of places, as did the XX Bundeslayga selections, with Webbo moving up into the second money spot. Webbo has picked up 14.13 points in two weeks. The "in contention" portion of the table is actually falling out of contention with only the XX Draws under down by less than 10 points so the mid-section should be renamed to the "not-really-in-contention" group and includes those down by less than 20 points.
Fairfranco was the big mover here picking up 5.37 points, but of the other active players, it was all red. Hofs Hackers picked one winner from 13, the Draw hunters have already been whined about, and Football Elite had a 100% losing weekend, which brings us to the bottom of the table.
Again, all red with one notable exception. Punters' Friend Neil hit his second winning week in a row and moves off the bottom of the table for the first time since January. The recovery may be a little late to challenge for prize money, but Neil's joy is now Peter Nordsted's shame, as his Premier Betting Account Bets (Official selections) drop to last place with a negative ROI of over 20%. The modified selections for the FTL and using Pinnacle's prices are better in terms of points lost, -44.91, but the ROI on these is even worse at almost 25%. The good news is that somewhere, in an anti-universe, Peter's selections are probably up by the same amount. 

Graeme made a small loss of 0.27 points, and the bounty liability remains the same at £275.

Sunday, 6 April 2014

Worth A Munchen

In my review of where the draws were last week, or rather where they were not, I wrote this about Bayern Munich:

While an understanding of human nature might have suggested a sub-par performance was in prospect for a game sandwiched between the highs of winning the Bundesliga in Berlin on Tuesday and the importance of a Champions League Quarter Final first leg at Arsenal a week later, few would have would have expected a previously 100% at home Bayern Munich and having with a goals for / against margin of 39 to 7, to come up with a 3:3 draw. The Pinnacle Sports price was 8.61 for the draw (the 13th longest in this league so far this season).
In other words, Bayern Munich's domestic fixtures are now meaningless. Motivation is a subjective factor. Maybe Bayern's players are motivated to continue their 53 game unbeaten run. Maybe they don't care. Maybe management don't care and want to blood new players. The unpredictable nature of motivation means that any bets on such teams, and at this time of the season the list increases week by week, are even riskier than usual. It was probably a surprise to subscribers to my Cassini Draws And More service that the value selections included this one:
I'll make the lame excuse with the benefit of hindsight that my model's price of 1.16 assumed the match would be played with both teams doing their best. A price of 1.4 at the time the list was compiled looked good enough, with a 20.4% edge, to justify inclusion, although I did caution that:
"One comment is that Bayern Munich have clinched the league and their motivation may not be what it was."
Augsburg won 1:0 in case you missed it. Bayern "fielded an experimental side". Models don't work too well with 'experimental sides'.

Life can be funny. Some time ago, a commenter by the name of Danny and myself didn't exactly see eye to eye on a few things. Roll the clocks forward a few months, and we have our own mutual appreciation society, although if the XX Draws don't start hitting soon, the club may be disbanded in record time.

I am reminded of the old joke about a father and his teenage son who didn't get on, and had little to do with each other for several years. After re-establishing their relationship, the son was amazed at how much his Dad had learned in the intervening years. This is not to imply that Danny has changed, in fact it may come as a surprise to readers that my sense of humour and faux arrogance can easily, and often is, taken the wrong way, but the bottom line is that Danny is, and probably always was, a decent individual (even if his choice in football club leaves something to be desired).

Danny had this to say this morning:
So yesterday was my first day of betting on Cassini Value and Bayern the first match. When I came to place the bet they had drifted to 1.60 and as I was doing a full Kelly (such is my total belief in you!) and because the 'true' price was 1.16 it recommended a 63% bet - FULL BANK JOB!!!
I was at the point of sending a bullion van round to Pinnacle when the memory banks kicked in and I had that deja vu feeling. A few years ago I had a big bet in very similar circumstances on what turned out to be the Carlos Tevez game, the infamous Man U West Ham match. I backed United at a certain price like 1.50 something for a lot of money only to see them drift like the proverbial barge followed by that familiar sinking feeling as United were listless and inept. It was a horrible experience at the time but a good lesson learned.
So Bayern brought up the familiar question - you identify huge value on something, back it, only to see the price spiral away into the stratosphere. What to do, who to listen to? Well these days I always listen to the market, hopefully you did too. However this is blatant after-timing of course!!!
My approach in these situations is typically the "if it looks too good to be true, it probably is". I like my value to be over 10%, but if it is too high, I suspect that I am missing something. Danny mentions the price drifting out to 1.6 which makes the edge 37.9%. Frankly an edge that size in the generally efficient top league football markets is improbable, and the other canard of "don't throw good money after bad" seems appropriate.

With Elche (the Crystal Palace of La Liga and not conceding in their last seven home games) rounding off a good weekend for the Cassini Value selections (Bayern Munich debacle not withstanding) by scoring a 90th minute winner, the ROI on these is at about 26% (official prices not yet available) after 98 selections, and an edge of this size is likely unsustainable (although I am doing my best!). 98 is of course a low sample size, as Danny mentions later in his email:
Dare I ask the $64K dollar question - If XX Draws disappoints this season will you still do it next year? I know you love it but is it really the best use of your talents? This Cassini value looks really interesting and is starting to look like more than statistical noise, however it is still early days.
I suspect there will be more on this topic later in the season. The inclusion of Elche in the Cassini Value Selections is another topic of interest. When I did my preliminary run through the matches in midweek, Elche were 2.14 with Pinnacle, the "official sportsbook of the FTL". This mean the edge was 9.7%, and thus just missed out on inclusion. When the final email was prepared, the price had drifted out to 2.19 (a 12.3% edge) and was now thus a selection. The market moved against Elche, for what reasons I know not, but it highlights the limitations of a model that uses macro-data (form and ratings) rather than micro-data (team line-ups). Unfortunately most of us do not have the time to drill down to this level, although Skeeve with his focus on two minor leagues does a decent job - at least to level stakes he does - although as with most tipsters, he doesn't suggest 'true' prices.

Saturday, 5 April 2014

Crystal Balls And The Perfect Model

Hejik left a comment on my Zero-Inflation adjustment to the Poisson distribution pointing out that the implied assumption that Poisson is an effective tool in modelling football matches is debatable, and indeed it is certainly a valid topic for debate.   
There seems to have been much debate about Poisson on various forums of late and the danger would appear to be the perfectly natural beginners obsession with building the perfect model.
Such a thing does not exist as you'll well be aware and 'best guess' is a reasonably fair assessment of what professional bettors are doing.
As long as your 'guess' is extremely well educated you have half a chance at the books who are also, apparently surprisingly to some, employing 'best guess' pricing models, not psychic witches with crystal balls.
For what it's worth, I think it's important that we question how efficiently Poisson fits to football and observation would suggest it's far from perfect.

However, even without that information it's easy enough to arrive at the same conclusion when we consider that some of the biggest bookmakers in the world are actively promoting it as a method of beating them. I'm not sure they'd be so inclined to do that were it even near as effective as they suggest.
Going through the comment in point order, the impossible perfect model that Hejik speaks of does not and cannot exist, and as I have commented on before, the perfect is the enemy of the good. A football match has far too many random variables for any model to get anywhere near to perfection, but that shouldn't be a reason to give up on creating a model that predicts outcomes more favourably than the available odds imply.

The beauty of football betting is that there is no shortage of matches to apply your model to, although there does seem to be a tendency for people to start shouting about their ROIs before they have even a few hundred matches under their belts. Compare this sport to the NFL, with its 256 regular season games a year.

As for Hejik's point about the bookmakers promoting Poisson, there is an article from last August that Pinnacle Sports are promoting, but interestingly there is no mention of the zero-inflation issue which, given the frequency of team scores being binary is a rather key omission. Fortunately you have me to help you out here! I suspect that there is a good reason why the topic was omitted. The cynic in me suggests one reason, but it may be as simple as that the mention of the added and significant complexity to calculations would be beyond many of their readers who prefer the 'weatherman' approach to putting in any real effort.

One well-known and understood issue with using Posisson for football matches is that Poisson assumes the totals to be independent, but everyone knows that goals beget goals. Studies from 2006 have shown that this is truer in lower levels of football than the top levels, but it is true nonetheless.

If you are trading a match in-play, then understanding how goal expectancy is increased after a goal is critical, but if you are pricing up a match for a punt, you take your best guesstimate, and if there's a first minute red card, penalty and a goal, maybe it works for you and maybe it doesn't. Random fluctuations of probability will happen, but that doesn't mean your model is flawed. The thickness of a coat of goal-post paint can make the difference between a win and a draw, an over or an under, but such happenings need to be 'brushed' off.

A second (sort of related) comment was from my new best friend Danny:
Cardiff 0 Crystal Palace 3 was a Poisson buster alright!
It certainly was. It was in October 1997 when Palace last scored three goals in a Premier League away game (3:1 v Sheffield Wednesday) and with a total of 6 in 15 trips this season, the likelihood of 3 in one game was remote. I had 0:3 priced at 70.2, and as XX Draws and More subscribers will know, a +0.3 rating to Cardiff City meant this was an XX Draw selection. Some draw losses hurt more than others though! Palace finished the 1997-98 season in last place, and while I am not celebrating yet, to say that Palace have exceeded my wildest expectations this season after a promotion that was in many ways almost accidental last season, would be an understatement.

Zero-Inflation Revisited

I had a couple of questions about Elo based ratings, one I can no longer find but fortunately I recall the gist of it, and the second was on my clarification about how I handled the zero-inflation problem in this recent post.

To address the second one first, the comment from Alkalmazasok was this:

Hi Cassini! I'm a bit confused here... You say you track the goal expectancy against the actual number of goals scored. The zero-inflation comes in when you compute the probabilities for different number of goals (with the help of the goal exp.). So how does the comparison of goal expectancies and the actual results say anything about the zero-inflation?
Zero-inflation is a problem with the Poisson Distribution when applied to football - essentially it under estimates the likelihood of a team scoring zero goals and to a lesser extent, of it scoring one goal. The input into the formula is the number of goals you expect a team to score, a number that will not usually be an integer, but something like 1.67341. When you plug this number into Poisson, it tells you what percentage of the time the time the team will score 0 goals, one goal, two goals etc. The numbers for zero and one are too low, and need to be 'inflated' with a manual adjustment.

Cassini Top 10
My solution to this, and there are most likely better ways out there, was to experiment a little and come up with an inflation formula that appeared to make the correct, or close to correct adjustments. In simple numbers, if the goal expectancy is 1.5, Poisson tells me that 22.3% of the time, the team will not score, and 33.5% of the time they will score one goal. If over a period of time my goal expectancy of 1.5 is on average correct, but the team scores zero goals 30% of the time, and one goal 36% of the time, then I can adjust my inflation accordingly. If my average is 1.5 and over a large number of matches teams are scoring 3 goals, that's a problem with the expected number of goals.

Club Elo Top 10
The first question was how do I determine the starting ratings for each team, and again, this is a best guess, although one good resource is to use the numbers already calculated by a site such as clubelo.com. Once you start applying your own bespoke adjustments to them, they will settle down soon enough. I started my own ratings about five or six years ago now, using league points along with the EA coefficients for each league as a starting point, and the rankings are not hugely dissimilar to those at ClubElo although my scale appears to be a little more stretched out. Paris St Germain suffer because Ligue 1 is rated the lowest league, but as points also transfer between clubs as a result of Champions League and Europa League matches, a run in the Champions League will help them. Since the ratings are used for domestic league games, how Bayern Munich compare to say Manchester City is of academic interest only anyway.