New Year's season ticker Resolution

(Spoken in the manic voice of Dr Heinz Doofenschmirtz):  

 "Behold! My newly improved FPL Incorporated SeasonTickerSpreadSheet-inator!"

Key: Dark blue is best and dark red the worst. The top tables relate to expected goals scored, and the bottom ones to expected goals conceded and clean sheets.

Towards the end of last season I made the decision to try to develop my own season ticker, as I liked the concept of identifying favourable and unfavourable fixture runs.  There were many tickers out there already, but I didn't like not knowing the methodology they used or how regularly they were updated.  Clearly, I have trust issues!  I also didn't like the thought of being at the mercy of somebody else's subjective opinions, as most tickers seem to rely on somewhat arbitrary weightings.  Okay, so I'm a control freak too!

For 2015-16, I've been relying solely on my own season ticker, and the results have been promising.  I enter the New Year doing well in my mini-leagues with 1,137 points and an overall ranking of 12,301.  This is despite some missteps along the way caused by various glitches, gremlins in the works, and several instances of misentered data.

One of those missteps saw me transfer in Lens in GW5 and I very nearly captained him in GW6!  I really ought to have cottoned onto something being amiss sooner.  It turned out that data in one table from Leicester down to Swansea was one line out, so Sunderland had been inadvertently credited with Stoke's results.  Doh!

As the season has wore on, I've smoothed out many wrinkles, and found more efficient methods of keeping my season ticker up to date with accurate data.  Now halfway through the present BPL season, I found myself having yet another eureka moment this past week whilst updating my season ticker.  Sparing you all the details, I realised there was a flaw in the logic of my final formula in a series of four calculations* my spreadsheet makes to arrive at an expected scoreline for every match in every gameweek.  I'm so confident that this realisation will boost the accuracy of my algorithms that I'm gearing up to start sharing the results with the FPL Community.

Certainly, if I'd had this insight earlier, I'd not have held onto Eriksen like I did this week even when I knew his price was about to drop.  Nor would I have brought van Dijk in for Wollscheid last week.  I'm happy to go without STO DEF cover for the foreseeable future, but SWA were clearly a much better option for a replacement than SOT (see below).

What I want my spreadsheet to be is a score predicting season ticker based on recent form only.  For the time being, I'm using the last 3 home games and the last 3 away games to guage teams' home and away form.  My expectation is that this results in my season ticker being much more responsive to changes in teams' form, which hopefully allows me to be ahead of the curve in my transfer dealings. 

I'm currently working on the most helpful way to present this information, but as a taster, this is how the teams with the best attacking potential over the next 6 games might be shown:

Thus, I might comment that those not selling Vardy will be pleased by this table.  As indeed will Ighalo owners.  And, looking at the projected season ticker I always do for the following week, I might reassure Arnautovic bandwagoners that Stoke are set to rise in this table in the coming weeks.  I might even warn WBA investors that these calculations do NOT factor in such things as Rondon's unavailability for the next match.
 
Similarly, the teams with the best Clean Sheet potential might be shown like this:

Again looking ahead, I could warn potential SUN DEF investors that the Black Cats are set to disappear from this list next GW, along with WAT.

Disclaimer:  The very elusive nature of form in football and the sensitivity of my algorithms does mean that fluctuations in my season ticker tables are inevitable.  Also, I should stress that my tables are to be used alongside your own judgement and knowledge, NOT instead of them!  Getting the most from them will depend greatly on synthesising them with your own well-informed instincts.

There are numerous options for which information to share from my spreadsheets, so I will deliberate about what works best, and is visually the most simple, over the next few days, and hope to launch something soon.

Watch this space.

In the meantime, please feel free to comment or ask any questions.

And best wishes for a Happy New Year!

Coley

Appendix

If I tell you that the top line of data, shown in the first screenshot at the top of the page, is actually line 221 of my spreadsheet, you'll understand that there's a lot going on before we get to the business end of my season ticker.  What is seen there is only the fourth part of the process.

*The four calculations consist of the following:

1. Adding up the goals scored and conceded in each team's past 3 home games and likewise for each team's past 3 away games.  The number of goals scored and conceded by each team are amended according to criteria** that take into account a wide range of contributory factors, such as whether they came from deflections, penalties, or unforced errors.  For example, teams are sometimes awarded penalties in situations they were unlikely to score from, so these might be counted as only a third of a goal, or 0.33 goals.
2. Ascribing teams different coefficients for the goals they score and concede both home and away.  These are weighted according to current form, as reflected in the last 3 matches played at home or away.  This tries to reflect the simple truth that some teams are easier to score against than others.  Currently, you could make the case that it's at least twice as difficult to score against Man Utd than it is Aston Villa.
3. Reassessing the total worth of the goals scored and conceded in each team's past 3 home games and likewise for each team's past 3 away games, in light of the weightings associated with the respective opponents faced.  By this method, 6 goals scored in a team's last 3 away games, say 3 against Villa, 2 against Everton and 1 against Newcastle, might be deemed to have the same value as just 3 goals scored by a different team in their last 3 away games, say 1 in each against Chelsea, Man City and Swansea.
4. Comparing the number of goals each home team is expected to score on average with the number each away team is expected to concede on average to calculate an expected scoreline.  

**The criteria considered for amending goals:- Deflected cross/shot on/off target

Gifted chance to score

Penalty when unlikely to score

Soft penalty

Deflected shot on target

Injury time on the break vs team chasing equaliser

Intended cross Lucky rebound

Penalty when 50/50 to score

Goalkeeper/defender should have done better

Penalty when more likely than not to score

Goal vs 9 men

Penalty when very likely to score

Goal vs 10 men

Goal denied by foul play

Goal incorrectly disallowed*

So Much Mumbo Jumbo!

When it comes to FPL team management, are you in the words of Stevie Wonder "very superstitious"? 

Very superstitious, writing's on the wall,
Very superstitious, ladders 'bout to fall..

It seems to me that a sizeable majority of active FPL managers are prone to irrational thought processes tantamount to superstition.  Now, if superstition is just your way of adding to your enjoyment of the game, then that's fine and dandy.  Good for you.  I'm not here to rain on your parade.  If you take your overall rank or mini-league standings seriously, however, then the more aware you can become about illogical reasoning and weak argumentation the better.

Don't get me wrong, we are all susceptible to making flawed assumptions and jumping to over-generalised conclusions, but some of the misuse of statistics I see from high profile FPL managers to support their decision-making is laughably ludicrous.  For instance, imagine if in summarising why I was nominating Wijnaldum as my captain pick for GW13 I concluded by stating the following:

Finally, another novel reason which validates Wijnaldum as captain is based on an interesting fact that I discovered through Astrology.  He's a water sign which appears to be the most optimum sign for dream players of the week, as follows:

Fire             (Aries, Leo, Sagittarius)          25.00%

Water          (Cancer, Scorpio, Pisces)       41.67%

Air               (Gemini, Libra, Aquarius)        08.33%

Earth           (Taurus, Virgo, Capricorn)      25.00%

How seriously would YOU take such an explanation?  And yet, this is no different to the fuzzy logic and woolly thinking put forward by a prominent pundit last gameweek.  He's not alone though.  Far from it.  I see countless examples by other leading lights every gameweek and it frustrates me enormously the significance that is routinely attached to ridiculously small sample sizes.  Remember, the law of large numbers tells us that flicking a coin a thousand times is much more likely to result in heads nearer half the time than if we repeat the experiment with just ten coin tosses.

Obviously, if you measure the frequency of four arbitrary factors, you shouldn't be too surprised if there are differences in how they score.  To then make inferences based on the results though is just the stuff of whimsey.  Granted, these perpetrators of statistical crimes can be forgiven for trying to make the often repetitive and tedious nature of FPL articles more interesting to read.

At the outset of this season, I was assured by the FPL community on more than one occasion that I was making a big mistake starting without any Chelsea players in my team, because historically they ALWAYS started strongly.  That's the thing with statistics though, they're only true until they're not; they tell us what has happened in the past, not what will happen in the future.

It isn't just the FPL prophets who are guilty of this either.  Their disciples are just as often willing accomplices to this type of groupthink.  Take captaincy picks based on the record of a particular player against a particular team.  Many of you will easily remember the times such thinking came off, such as Rooney's lucky assist vs Spurs on the opening day of the current season, or likewise, Oscar's fortunate goal vs Swansea perhaps?  Less will readily recall the times these coincidences didn't pay off.  In poker, this is called results oriented thinking.  How about Rooney vs Aston Villa this season, the team he has scored more goals against than any other?  He blanked.

Think about it like this - if every team consisted of eleven cloned Oscars, would you expect the score of every match to be the same?  Of course not!  Why?  Because of all the countless pivotal moments and numerous turning points that comprise the beautiful game.  And most of the Oscars would have a team that they score more against than any other.  I guess what I'm trying to say is that such coincidences are inevitable.

That all said, it is easy to understand why we make these mistakes over and over again, as researchers have repeatedly shown how vulnerable to confirmation bias we all are.  As a consequence, we readily accept the stats that support our beliefs, but dismiss those that contradict them.  Furthermore, our brains are lousy at correctly calculating probability.  Try this problem for example if you haven't come across it before:

How many randomly chosen FPL managers do you think there would need to be in a mini-league for the chances of two of them sharing a birthday to be more likely than not? *

These tendencies and many other unhelpful ones besides ('gamblers fallacy', 'primacy effect', 'recency effect', etc.) are often hardwired into our brains, and thus, difficult to eradicate, but I want to argue that we should try to be more honest with ourselves when applying stats to our FPL justifications.  Otherwise, we are doomed to the fate Stevie Wonder warned us about:

When you believe in things that you don't understand,
Then you suffer,
Superstition ain't the way

*The chances are that, even when I tell you the answer to the birthday problem above, your brain will refuse to accept it!  Believe it or not, but the answer is 23. Most people start instinctively by halving 365 days and revise downwards from around 180. From personal experience, I can report that betting someone you'll find at least 2 people from 30 random strangers in a pub who share a birthday is both a fun and profitable game to play next time you're out with your (preferably gullible) friends! ;o)