SMRTILYTICS is a bit of a hobby of mine. For a full explanation of this project, check out the main SMRTILYTICS page. -OG
Week 33
I’m back from the nether-reaches of the Sea of Cortez (or Baja Sur California as the locals call it) with a tan, some fairly cheap Liga MX replica kits and one helluva hangover!
This week’s SMRTILYTICS update is likely the penultimate 2016 entry as we enter the endgame of the Premier League Season. I have a few things to comment on as per the usual dross before going into a new space which has taken some work as well as grey matter, but more on that to come.
Where to start – perhaps the top is the most relevant. Leicester, if you recall previous SMRTILYTICS updates, have been outside of the Top Four in terms of Goals Conceded as well as Total Shot Ratio on Target (TSR*) to date. They had looked like a statistical outlier but as mentioned in my previous update, I speculated that we’d see them course correct over the next set of games and we’d see them shortly in the Top Four for those stats.
That has happened after a string of straightforward results has kept them firmly at the summit of the table. They now have the third best goals conceded stat in the League (31) behind United (30) and Tottenham (25). They are just outside the Top Four in terms of TSR* but even that should change in the next few games if they maintain current results. In terms of their Combined Metric, there again on the strength of Goals Conceded, they have entered the Top Four at third (0.40) behind City (0.37) and Tottenham (0.24).
They’re surely odds on favourites to win the League despite Tottenham leading all three metrics tracked.
In terms of Arsenal, the slide has been a long one. We continue to just make the Top Four for Goals Conceded but allowing 3 at West Ham at the weekend has undone some previous good work. That really has been the story of our season – good stretches of results and clean sheets punctuated by conceding multiple goals in a single match.
In contrast, Tottenham have never conceded more than 2 in a game this year, Leicester have done it once (we did it to them), and we’ve done it four times. Surprisingly two of those games ended in draws (Liverpool and West Ham) and while I’m dealing in “ifs and buts,” had all four of those games been wins – we’d be three back from Leicester with a game to play.
Watching us this season, there’s been a mental fragility present during critical periods – when we need a win to stay at or near the top, we’ve blown it spectacularly. None of the games where we’ve conceded high goal counts have come at the right time – contextually in the season, they’ve hit at exactly the wrong times. Southampton’s demolition and Liverpool’s late draw in Weeks 18 and 21 respectively ended what looked like brief revivals in our form after losing Santi and Coquelin for extended periods. The five games after Liverpool saw us win just two at the end of that run – only for us to be effectively eliminated from the title race at Old Trafford.
Click this graphic to see it in full resolution
Changing subjects ever so slightly, as you know SMRTILYTICS has endeavored to be able to predict outcomes – something I’m not completely comfortable with yet.
Despite that, I have been looking at the TSR* and recent results of teams as well as comparing to the Combined Metric (remember Combined is a result of TSR* and Goals Conceded). In each case, the metric has picked the winner 60% of the time over the course of the last 3 match days (18/30).
But that is an overly simplistic approach since there are draws in football and there were 8 draws in that same data set (about 27%). While the data is still very incomplete, I was able to look at the variance when a draw takes place and then create predictive rules based on that variance. Additionally, I then created a rule for when TSR* and Combined Metrics predicted different outcomes using the maximum draw tolerance observed over the same period.
All of this numbo jumbo (credit Ted Knutson) has produced the following predictions for Week 34 which I have little confidence in but hey, it’s fun:
Norwich vs Sunderland – Draw
Everton vs Southampton – Southampton win
Manchester United vs Aston Villa – Manchester United win
Newcastle vs Swansea – Swansea win
West Brom vs Watford – Watford win
Chelsea vs Manchester City – Manchester City win
Bournemouth vs Liverpool – Liverpool win
Leicester vs West Ham – Draw
Arsenal vs Crystal Palace – Arsenal win
Stoke vs Tottenham – Tottenham win
DO NOT BET ANY MONEY ON THESE PREDICTIONS
I really wanted to get these predictions in this week but my degree of confidence is rather low. I’m skeptical and the system still feels too unpolished to predict anything accurately. In particular I’m dubious about that City game and it does just look like the higher placed team has been picked in most cases.
The goal is to better 6 of the 10 predictions and then to do that regularly before I can feel like I’ve made any headway. It is likely that come the summer, I’ll do a much deeper dive to see if I can refine the system and the rules I’ve created in the background.
Right, that’s the lot of it for this week – I will keep track of the system predictions for the remaining fixtures so pay attention to @SMRTETA for those weekly game predictions to get you through to the end of season special.
On to the usual rabble…
Youknowit!
OG