Monday, November 19, 2012

A week of missed chances

We focus a lot on shots and chances created around these electronic pages, based on the idea that they are more predictable than goals scored. Over the course of a season this is undoubtedly true, however, this week gave us a good reminder of what can happen in a sport where the odd goal can decide games and one moment of magic can give more rewards than 89 minutes of sweat and toil:

No real lessons here, it just struck me as particularly noteworthy that five of the week's winners were all out shot by their opponents, by some distance in a couple of instances. Indeed, we've seen four instances all season where teams had 20+ shots but lost, with two of them coming this week. The others were QPR's crazy 5-0 loss to Swansea in gameweek one where they out shot their opponents 21-13, and Spurs' 4-2 loss to Chelsea in gameweek eight in which they had as many SoT (10) as their opponents had total shots yet came away with nothing.

The 'pundits' would like us to think that all this is totally meaningless without goals ("goals win games" remember), and in a way they're absolutely right, all these highlighted teams earned exactly zero points, but if we replayed the fixtures 100 times, I think we can be confident that, for example, Newcastle at least manage to score more than once on 32 shots (thirty two!) fairly often, winning the game in a decent proportion of those outcomes.

Nothing more to take away here, just a reminder that however confident we are in forecasts, in a single gameweek, things can always go unexpectedly.


James Richard Klien Jr said...

It is precisely for this logic (i.e. the latter part of the third para) that your blog is the only FPL blog I can read. You should really develop upon this trail of thought by experimenting with a Monte Carlo simulator rather than the current statistical methods featured in your forecasts.

FPL is one of those areas where the use of statistics is unreliable, and as you have mentioned many times you are not a statistician (it’s a good thing!), the simulator should play to your strengths. It may also help to focus more toward who to avoid (more reliable) rather than who to pick (which is more variable/random and better left on those other blogs).

JT said...

I'm a huge fan of your blogs Chris and although stats will never allow us to forecast real-life events 100% accurately, they sure do help a lot in terms of decision-making. There will always be more work to be done and there are a lot of sceptics, but keep up the good work!

I've been looking into Baines' chances created (according to the opta stats) vs. assists and over the last 2 seasons his chances have been converted at a much lower rate than many other players. It would seem that most of the chances he creates are from crosses as opposed to through balls which would go some way to explaining the rate but this has never stopped other wide players from doing well, especially when you consider the sheer volume of crosses and his crossing accuracy.

I don't know whether you've touched upon this in the past but does the model take into account crosses, cross success and through balls when assessing assist potential? I'd love to hear your thoughts on the topic.