EPS Form Watch

[Which Tyler]

Dammit - looks like a simply cocked up putting the formulas in for Freddie; the whole batch of K/P/R/MR had shifted one to the left; so what is kick up there is points scored; pass is actually kicks, etc. Only for Freddie as far as I can tell, who was the latest addition to the spreadsheet.
I've corrected the formulas for next time; but I'm not going to bother redoing the image and posting it up again right now.

Pretty hard. I can do it, I have all the raw data, it's "just" a case of collating it by the individual; but I don't think it would say much TBH; the amount of game time the kids are getting is too variable, and for most of them would be 10-20 minutes here or there once the match has been decided; and starts against weeker oponents (and no way I'm going to go getting the information from Low Value cup or A-league where it'd be comparable for the kids). About the only really useful stat to gather for them would be game-time; which is the biggest bugger to get hold of in the first place.

I'd suggest you need at least 350-400 minutes of rugby before you can start thinking the stats are actually representative of anything. For example, have a look at the scrum halves - Ben Spencer is way out ahead of the pack on stat.s. He scores more tries, runs further, better tackling, 100% off the tee... comes on for the last 20 minutes against beaten and knackered opponents with their heads down.

 

[LeinsterMan (NotTigsMan)]

I hope that the door isn't entirely closed on Feddie for England like it is Cips.

 

[ALJ]

I hope that the door isn't entirely closed on Feddie for England like it is Cips.

I can't see why it should be - He's a lot younger, has an excellent all round game and has none of the baggage.
Cant be a bit hot and cold with his form but I really rate him.
Only potential downside was how he downed tools at Glocs once he knew he was leaving - might suggest a small attitude problem but equally there might have been a lot more to it as Glocs has a bit of a history in the acrimonious departure department.

 

[Raggs]

Tough to keep performing for a team when the head coach doesn't include you in any of the team talks etc.

 

[j'nuh]

Every time you put these up I tell myself I'll have a good long look, and every time I realize I'm too lazy! I really under appreciate stats if I'm honest, although I stand by my assertion that they frequently obscure as much as they show (tackle % is a great example), if you have the patience to interpret them carefully they do have a lot to offer. Metres per run sounds like a useful one actually, I don't think I've seen that used before but it could be revealing.

Meters per run is difficult, because it is so open to outliers in the data. It is the one I loathe to read anything into.

For example, Vunipola makes 20 runs in a game. 19 of them are for one meter. One of them is for 61. Average = 4m per carry. But 19/20 of his runs were less than this! He ran into space just once in a game and it has inflated his stats to a point where they are meaningless. What I would love to see is a trimmed mean of all his runs: that is, an average excluding the outliers (the breaks). A median could work too.

But even then, there's a more pressing issue: where does he start running? Do you find it more impressive for a player to start 5m behind the gainline and get tackled 1m behind the gainline, which gives 4m per carry, or do you prefer a player that takes it flatter for fewer meters, breaking the gainline constantly? For me, gainline success is a far more useful metric than meters run for determining the effectiveness of carriers. Clean breaks and defenders beaten are also quite useful to know.

Another angle is kicking vs running. Some players run a lot, some players kick a lot. It is unfair to judge the players that kick a lot for not making the meters. May used to run from deep all the time. It looked great in his stats when he was routinely making more meters than any other fullback/winger in England. Now he's making fewer meters than anyone, but he's a much better player for knowing when to use the kicking game.

The problem with the data is that they are very poorly controlled, and sample sizes are incredibly small. Worst of all, you can infer many different, contradictory things using various metrics. Kicking percentages do not tell you the propensity to kick from range and from strange angles for example. I would love to run regression models on this kind of thing, if the data were available.

 

[ratsapprentice]

Meh... over the course of the season, if those "outliers" are making a significant impact to the average number then they aren't really outliers.

The metres made is generally reflective of the actual metres made you can expect of a player.

 

[j'nuh]

Meh... over the course of the season, if those "outliers" are making a significant impact to the average number then they aren't really outliers.

The metres made is generally reflective of the actual metres made you can expect of a player.

Let me give an example in another way. There's a product that two salesmen sell - this product can be bought as a one-off, or as a repeated service, month-on-month.

On a given day, person A sells £200 of this product and person B sells £300 of the product. Great, so you infer that person B is a better salesman. But lurking in the background is some data that says that person A sold more of the product on a repeat basis; half of person A's customers will be repeat buyers who have put orders in for future months, compared to only a quarter of person B. The problem with the £200-£300 data is that it doesn't provide much context beyond what was sold on the day.

So with meters run: when you add both meters made in a clean break, and "hard meters made", you miss out on important information. We cannot tell from meters per run how effective someone is at taking the ball into contact. Making the hard meters and setting up next phase ball is the bread and butter responsibility of a number 8. Judging them on meters per run, which effectively only tells you how good someone is at making breaks, is unfair.

 

[ratsapprentice]

I disagree.

 

[ALJ]

Tough to keep performing for a team when the head coach doesn't include you in any of the team talks etc.

I believe it was something of a 'cause and effect' type scenario - no doubt he'll say he was left out and that affected his game. Davies would probably say he was left out because of his attitude.
He had a lot of injuries that year too but how much of which was kinda difficult to say.

Mind you, I'm not Davies' biggest fan anyway. Entirely too much ego (and god knows what he got up to while at the Scarlets - Can't remember if what I've heard are rumours or established facts).

 

[Every Time Ref]

So with meters run: when you add both meters made in a clean break, and "hard meters made", you miss out on important information. We cannot tell from meters per run how effective someone is at taking the ball into contact. Making the hard meters and setting up next phase ball is the bread and butter responsibility of a number 8. Judging them on meters per run, which effectively only tells you how good someone is at making breaks, is unfair.

See your point here, there are two types of skillset conflated in that data. I also agree on the issue of where you take the ball - fullbacks always "make" a lot of metres by running 20m unopposed into contact

 

[ratsapprentice]

That stuff only really applies if you don't actually watch the players you are analysing at all - or at least enough to understand what style of player they are.

The FB thing doesn't really matter - you're only really comparing their metres made stats with other FB's.

 

[RedruthRFC]

Meh... over the course of the season, if those "outliers" are making a significant impact to the average number then they aren't really outliers.

On that basis, isn't the week that you win the lottery an outlier over a lifetime of buying lottery tickets?

My very old maths A-Level is serving me poorly here (I suspect it's have enjoyed it a lot more and done a lot better if statistics was related a lot more closely to real life applications like this), but isn't it possible to calculate the probability that each average is an accurate indicator in the same way that it is to say whether a gambler is genuinely +EV or just running hot?

http://goalkickers.co.za/ provides an excellent example of how much more deeply statistics need to be considered before being considered a useful indicator when considered in isolation.

For anyone who's interested, Googling "sabermetrics rugby" throws up a few interesting articles, although they're all pretty anecdotal. Maybe fodder for a new thread here.

- - - Updated - - -

That stuff only really applies if you don't actually watch the players you are analysing at all - or at least enough to understand what style of player they are.

The FB thing doesn't really matter - you're only really comparing their metres made stats with other FB's.

Agreed 100%. I look at this the same way as dealing on money markets, you need a combination of statistical and fundamental analysis to draw useful conclusions and need to understand how to interpret statistics when you do look at them.

In sport, statistics make simple column inches for lazy journalists and an easy way to "prove" a point for lazy minded supporters. Everyone here appears to be thinking on a deeper level which is making for an interesting discussion.

 

[ALJ]

Agreed 100%. I look at this the same way as dealing on money markets, you need a combination of statistical and fundamental analysis to draw useful conclusions and need to understand how to interpret statistics when you do look at them.

In sport, statistics make simple column inches for lazy journalists and an easy way to "prove" a point for lazy minded supporters. Everyone here appears to be thinking on a deeper level which is making for an interesting discussion.

This.

Stats are clearly important for many reasons but they are meaningless in isolation.

 

[ratsapprentice]

On that basis, isn't the week that you win the lottery an outlier over a lifetime of buying lottery tickets?

Maybe I'll change my mind when Jordan Crane runs 100 million metres in a game.

 

[RedruthRFC]

Maybe I'll change my mind when Jordan Crane runs 100 million metres in a game.

To an extent, I was engaging in reductio ad absurdum in order to make my point. To clarify it another way, how large of a sample size do you need before you would be (say) 95% confident that what may appear to be outliers are actually a valid part of your data set?

 

[Which Tyler]

This.

Stats are clearly important for many reasons but they are meaningless in isolation.

Exactly, we're also comparing like with like. Billy may make 1-2 longer runs per match from long kicks down... But so does Morgan, and so does Hughes outliers thst are regularly repeated aren't outliers... by definition.

No one is criticising a FH who kicks a lot for not running far, though if anything, their average run should be further, as they only go when there's a gaping hole in the defence; but the K/P/R stat will make it bloody obvious that they kick more than they run, so in that case,context is provided in the raw stats!


Raw stats are very useful, but only if you know that they actually tell you, rather than projecting; and provided you have seen enough of the players concerned to draw sensible conclusions.
Eg. BMI is a brilliant tool for assessing health risks, and is often rubbished because idiots are stupid, use it as the ONLY piece of information, and assume because it's bloody obviously not 100% perfect, that it must therefore be100% useless.

 

[RedruthRFC]

Exactly, we're also comparing like with like. Billy may make 1-2 longer runs per match from long kicks down... But so does Morgan, and so does Hughes outliers thst are regularly repeated aren't outliers... by definition.

Good point, I hadn't thought of it that way.

I like your line on BMI BTW, a little knowledge and all that.

 

[j'nuh]

That stuff only really applies if you don't actually watch the players you are analysing at all - or at least enough to understand what style of player they are.

The FB thing doesn't really matter - you're only really comparing their metres made stats with other FB's.

You've highlighted fullbacks. Here's a good example:
Fullback A: Carries 10 times from deep for an average of 10m.
Fullback B: Carries 10 times from deep for an average of 10m. Carries 10 times in the attacking line for an average of 5m.

Fullback A would have an average of 10m per carry. Fullback B would have an average of 7.5m per carry. The fact that Fullback B is getting through more attacking work is hurting his stats.

This isn't rare. The more that fielding players only field, and the less they get involved in attack, the better their stats will look. This is because fielding results in by far the easiest meters a player will ever make: you get 10-15m of unchallenged running, something an amateur could do. Any other type of run you do, unless you are making non-stop clean breaks, will hurt that average.

Jonny May used to be like this. He ran absolutely everything and had outrageous stats for it. However, he often put Gloucester under heaps of pressure by running it when he had no support. It looked great in his stats, but hurt his team. Now his running stats look terrible contrasted to other English wingers, but it's only because he's become very wise on when to run the ball. This isn't reflected in the stats.

Exactly, we're also comparing like with like. Billy may make 1-2 longer runs per match from long kicks down... But so does Morgan, and so does Hughes outliers thst are regularly repeated aren't outliers... by definition.

We know they all do it, we don't know how often they do it though. It seems to me that Billy does it more often than the others, and it may result in higher meters made.

Aside from that, consider who they are all playing for. Vunipola and Hughes play for the top two teams in the league, teams which can stretch defences and create the gaps like no other. Morgan has a significantly harder job of making clean breaks. If Gloucester and Sarries swap their 8s, you can bet that Morgan's stats will improve and Billys will go in the opposite direction.

 

[ratsapprentice]

You've highlighted fullbacks. Here's a good example:
Fullback A: Carries 10 times from deep for an average of 10m.
Fullback B: Carries 10 times from deep for an average of 10m. Carries 10 times in the attacking line for an average of 5m.

Fullback A would have an average of 10m per carry. Fullback B would have an average of 7.5m per carry. The fact that Fullback B is getting through more attacking work is hurting his stats.

No it's not.

It's showing that his average meters per carry is lower (because it is).

His average metres per game will be higher (because it is).

Jonny May used to be like this. He ran absolutely everything and had outrageous stats for it. However, he often put Gloucester under heaps of pressure by running it when he had no support. It looked great in his stats, but hurt his team. Now his running stats look terrible contrasted to other English wingers, but it's only because he's become very wise on when to run the ball. This isn't reflected in the stats.

You're right, it's not: which is why you have to watch the sport in order to properly utilise stats. And unless someone decides to add a stat that specifically records every possible quantifiable event in a game, you always will.

What those stats do still do (which is all that stats have ever, and will ever do) is quantify exactly what a player does. It's then up to you to use your underpinning knowledge of the game to understand and interpret them.

 

[Which Tyler]

J'nuh, that's where interpreting stats by actually watching rugby, akin to seeing a BMI stat and seeing a BMI of 40, then opening your eyes and seeing if your office contains an Adonis or a failed jelly.

Quite apart from what RA says, your also forgetting that I'm not taking 1 data point, I'm providing 2 whole years, and then the season so far, any outliers have been negated by sample size.

Your entire negativity is about idiots being idiots. It is perfectly possible to interpret these stats in a useful way, idiots won't, so dismiss the idiots, not the stats.