Various data sites track all kinds of weird and wonderful stats, one of those stats that I decided to track was Final Third Passes. Every game features an image like this:
I assume that the totals featured in the image above are attempted passes rather than completed passes, but who knows. I counted the numbers in the images above for all 190 games of the season so far.
The results look like this:
| Plus/Minus | Ratio | |
|---|---|---|
| Man City | 1095 | 60.29 |
| Tottenham | 916 | 59.77 |
| Arsenal | 1058 | 59.14 |
| Chelsea | 884 | 58.53 |
| Everton | 646 | 56.23 |
| Swansea | 552 | 55.6 |
| Man United | 488 | 54.78 |
| Liverpool | 493 | 54.66 |
| Southampton | 368 | 53.94 |
| West Brom | -183 | 48.14 |
| Newcastle | -199 | 47.84 |
| West Ham | -224 | 47.73 |
| Norwich | -461 | 45.61 |
| Stoke | -627 | 43.53 |
| Sunderland | -652 | 43.46 |
| Hull | -698 | 42.98 |
| Cardiff | -698 | 42.59 |
| Aston Villa | -760 | 42.06 |
| Fulham | -826 | 41.56 |
| Palace | -1172 | 38.59 |
Taking the image above and the other 189 from the rest of this seasons game I was able to create a final Third Pass +/- and Final third Possession%.
There's some pretty cool information in that sortable table. Information that tells us Man City and Arsenal both post +/- numbers better than Liverpool and Man United's +/- numbers combined.
Obviously Final Third Passes are susceptible to massive score effects and time decay and the like. But once again, as I keep stating, good teams will post good numbers given a large enough sample.
Reliablility Of Final Third Pass Ratio
To test the reliability in the 13/14 season I put the final third pass ratio from games 1-10 in one column and the final third pass ratio from games 11-19 in another column. The results of that process are here:
The relationship between the two buckets is really strong at 0.88. Sunderland are a huge outlier, they posted a ~50% number in weeks 1-10 and a ~42% numbers in weeks 11-19. Score effects, some drop off in their play are likely reasons.
Although final third pass ratio is highly repeatable that doesn't automatically make this stat usual for all that much. I ran the numbers to see what level of predictability this stat has in predicting future goal difference. To do this I ran the first 10 weeks of final third pass ratio with weeks 11-19 of goal difference.
We know that final third pass ratio is repeatable but we needed to know if could predict anything of worth. The answer is yeeaaa...maybe:
An r2 of 0.389 isn't terrible, in fact it's good. The problem is that we have far better predictors of future performance than final third pass ratio.
In case anybody is interested this is how Final Third Pass Ratio correlates to Goal Difference on a week to week basis:
********
A lot of maths here but the things to take away from this article are simple:
- Final third pass ratio is a pretty stable stat from one chunk of games to the next.
- Final third pass ratio has a pretty good week-to-week correlation with goal difference.
- Final third pass ratio is not that great a predictor of future goal difference.
- Final third pass ratio probably tells us a lot about how good teams are at pinning their opponents into their own defensive third and passing the ball in spite of the oppositions defensive pressure.