Mid season headline acts: part 1

Having looked at teams records and made some bold predictions, in this look at the season so and to add to the plethora of reviews, I will take a different look at specific players, popular ones and key players to different teams. Who is carving a path to glory? Who may be on the fade? Who is on the rise? Questions that I am unlikely to answer but will try to by jove! In two parts!


So how to choose who to look at? I have taken a slightly different route (the high road to others low road if you like) by going on how www.advancednflstats.com and www.profootballfocus.com have ranked players. In what will be an arbitary selection based on who I find interesting, I will look at different positions, including defensive positions just so we can discuss Kiko Alonso. I will be using two different types of average, where I use them, the mean (sum of numbers/total number counted) and the median (rank the numbers then pick the middle). I will be describing the consistency in terms of 95% confidence intervals i.e. I will be 95% confident future results will fall in this range. For the mean this will be reported as +/-no. and for the median, lowest<median<highest of range. Admittedly I started working on this in week 8 and we are nearly at week 10. So some of the data is for up to week 9 and some will be taking into account week 9. I am sorry about this but pretty graphs can take time. Also this will involve modelling but not that kind.

Let’s do this.


Peyton Manning

Are there any more superlatives or more byberbole that can be poured out across the ether about this man and the season he is having? Ever since that opening game against the Ravens (yeah remember the one where he threw 7 thouchdowns) and going for four games before throwing an interception (he has now thrown 6 picks) and leads the pack in passing touch downs with 29. Can he keep up with this pace? He leads in total passing yards and on current pace is on for 5800 yard season. But what is this? Matthew Stafford is not far behind? Indeed, he is third by the milestone standard on NFL.com and is set to reach 5200 yards. As of week 8, he has attempted 338 passes compared to Mannings 333 (admittedly Manning has the better completion rate, with 237 complete passes to Staffords 211). What about Drew Brees? What about other MVP contenders such as Andrew Luck, Aaron Rodgers (before his injury and as he is one of the greats, is included), Colin Kaepernick, Andy Dalton and new hero Jake Locker? I suspect a graph is going to be required.  But what graph? Well not all of the above are big on the passing front, as some do use their feet. So the first graph we will look at is touch downs per game to passing yard per game (just as a note, when coding this, Peyton Manning’s pass yards is pmpy and Aaron Rodgers is arpy. These just amuse me).



The line visualises the relationship between the pass yards thrown per game and the number of passing touchdowns thrown. The relationship is a significant one, the more yards a quarterback throws the more touchdowns he throws, with 1 touchdown scored every 140 yards. What is important to take from this is that there is no significant difference between the quarterbacks in relation to the number of passing touchdowns per game (ok for the quarterbacks looked at here but I as the likes of Philip Rivers, Tony Romo and Tom Brady all fall within the group of QBs looked at here I would say this applies to them). Have to say this comes as something of a surprise to me, but it is because of the odd bad game where a QB does not throw so many yards and therefore not so many touchdowns. Peyton may be leading the way but he isn’t head and shoulders above the rest. What about touchdowns in terms of the number of completed passes? To normalise the success rate I used the percentage of completed passes made. Ok so let us have a look at whether the number of touchdowns scored relate to the pass completion rate.


Again, look how clustered all our QBs are, so it should come as no surprise that there is no significant difference between our quarterbacks in the number of passing touchdowns thrown and their percentage of completed passes. Their is a significant positive relationship between the percentage of completed passes and the number of passing touchdowns. The line visualises the nature or the statistical model of the relationship. Ok so yes it does not predict a quarter back throwing 7  touchdowns but if I included every QB I suspect the line (a visualisation of the statistical model obtained from the data) would be a bit of a better fit. However the further above the line a player is, the better that player has performed and vice versa. So Peyton certainly has exceeded expectations this season so far. Incidentally, based on his percentage of complete passes, Nick Foles was predicted to throw 3 passing touchdowns, really emphasises how well he did  and the players he threw to did.

Anyway back to Peyton. Well he is on track for a career high of a season and may very well end up throwing significantly more touchdowns than anybody else (I stress the may). At the moment he has a significantly higher pass completion rate to Colin Kaepernick.

What about interceptions I hear you cry? I may come back to those in a different post as there is a different way I want to look at them rather than the rather rough and ready method of interceptions per game.

If quarterbacks and the top quarterbacks are all viable MVP options or basically have been as impressive as each other what about other star players? Onwards we go…

Kiko Alonso

The hero of many a Bills game and much loved by Buffalo fans and fans of other teams. He is a highly rated linebacker and having now looked into his performance, could be considered as a rookie of the year. Advancednflstats rate him as the number 1 linebacker so let us have a look at this endearing player. He has made the joint largest number of total tackles (full and assisted) of any linebacker, 81 so far (the Cowboys Sean Lee has also made 81 total tackles). He is also joint leader in number of interceptions, so important as this is a turnover. He only ranks 21st in terms of pass deflections, with 4 and has only managed 1 sack but for the Bills, Kiko has been a defensive rock along with Safety Da’Norris Searcey (53 total tackles and 2.5 sacks, is also ranked top Safety by AdvancedNFLstats). Oh ok how good is Kiko? He averages 4.5 complete tackles a game, with a 95%CI of 2<4.5<9 per game. Sean Lee manages 3<6.5<11 95%CI complete tackles per game, better but not significantly so. On combined tackles Kiko manages 4<9.5<12 95%Ci tackles per game, Sean Lee manages 4<10<16 95%CI tackles per game and leading tackler Karlos Dansby manages combined tackles of  5<8.5<11 95%CI. He leads both of them in terms of forced fumbles, he leads Lee in terms of sacks and leads Dansby in interceptions made. And this is why he is ranked high, he is a good all round OLB and promises to be one of the best this season (looking at a total of 160 combined tackles this season!)

All this modelling and plotting pretty graphs has taken a surprising amount of time and I have not finished writing about wide receivers and Jamaal Charles. Also we have probably had again enough stats for one article and I mean proper hardcore stats. If you are interested keep reading, if not then keep on checking our prolific twitter feed @GridironGents and our Facebook page, as well as our iTunes feed for all the latest updates. Of course keep on coming back to our website! Until part 2….

Toodle pip and onwards to statistics….




The way I analysed the touchdown to pass yards or pass completion was using a Generalised Linear Model with poisson error structure as touchdowns are whole integers and so are count data. Also you cannot get a negative number of touchdowns so poisson distribution works nicely. I used the optimal model generated to predict the number of touchdowns for a range of yards or completed passes. Oh for initially assessing whether there was a difference in pass completion between quarterbacks I used an square root arcsine transformation instead of a negative binomial model. And that is about it.