Awesome Rugby Computer System

Shakakan

Hi all

I thought I'd run something by your awesomely knowledgeable rugby brains

As part of a masters thesis, a colleague and I are analyzing data from rugby games. We have data from a couple of seasons of Rabo/Magners League, Amlin and HEC.

The data contains events from every game - so every pass, tackle, error, set-piece is captured along with who did it, where it was on the park and what the outcome was.

If you were a coach, or a player and you had a computer system that could tell you anything, what would you ask it?? Who is more likely to win was one question were looking at, but we're hoping for some more ideas that can give insight into player or team performance.

Any thoughts would be much appreciated!

Thanks
A

Work on defining each of the things you have said, break them into blocks if needed.

"Pass" can be a grubby pass along the ground to the wrong person, that should have resulted in a try but slowed the attack down, yet still counts as a pass, whilst "knock on" can be someone trying for an offload that would win a game when there's penalty advantage behind them, a veritable shot to nothing.

The definition of precise events in statistics is always a difficult place to work within. Too much precisions means your results can be meaningless, but oversimplification means that detail gets lost in the aggregate.

Things that interest me are decisions mostly. Sometimes players don't really have a decision to make, they take the ball and 0.5s later they are tackled. They were part of a passage of play but they didn't add more to that play than any other body would have done.

An interesting way to analyze this would be to look at the amount of times that "A Lock" takes the ball within the 10m channel of a breakdown, makes 3 steps, gets tackled and resets. After normalising the results, see what percentage of the time "Lock X" makes a decision that is different from the 'majority play' of hitting the deck and resetting, be it off loading, setting up a maul etc.

Rugby is a very structured game, its analyzing the events that take place outside the structures that will benefit you mostly in the above.

Another example would be something like. "On average in the RBP12, the out half kicks 40%, runs 15%, passes, 40% and errs 5%. Last year Ian Madigan was a massive outlier here with a percentage setup as follows 'runs 25% ,passes 50% and kicked and err'd 22.5% and 2.5% of the time respectively' ".

Find the outliers in the game and you find the big impact players and the 'bottlers' on the other side.

phog

Does it capture the lineouts - won/lost where the lineout was and where the thoirw went to, who the ref is, the weather conditions - dry, wet, wind direction, etc.

If you have a twitter account you could tweet @ruckingoodstats with your queries, he does stats for S15 and Internationals - site here. He used to share some info with Bernard Jackman.

[Jackass]

Bulk of this taken from BO rugby, but should give a good basis of areas to analyse.

Possession
Average points per minute of possession
Average points conceeded per minute without possession
Territory
Average points per territory advantage
Turnovers Won at rucks
Who won turnover
Turnovers Conceded at rucks
Who lost ball
Average Phases when in possession
Routes of that possession (pick and go / Drive / Expansive / kick and chase)
7+ Phases
Minutes in 22
Metres Gained by each individual
Line Breaks by each individual
Passer before line break
Intercepts for / against
Tackles (both individually and as a unit)
Missed Tackles (ditto)
Kicks out of hand (individual basis)
Metres Kicked
Kicks Out On The Full
Conversions (side of pitch and distance)
Missed Conversions (ditto)
Drop Goals (ditto)
Missed Drop Goals (ditto)
Penalties (ditto)
Missed Penalties (ditto)
Good Up and Unders (possesion retained kicker & Receiver)
Bad Up and Unders (possesion lost, kicker & receiver)
Up and Unders (individually)
Good Kicks (found touch or space / hard to define)
Bad Kicks (kicked possesion away with little return)
Knock-ons
Handling Errors
Forward Passes
Ball Carries
Yellow Cards
Red Cards
(lots of scope for analysis on effect of cards for and against)
Fights
Penalties Conceded
Penalties Won
Lineouts Won
Lineouts Lost
Lineouts Against Throw
Scrums Won
Scrums Lost
Scrums Against Put In
Rucks Won
Mauls Won
Injuries

CIARAN_BOYLE

I'd love to see an analysis of time spent from a tackle for the ball to come out of the ruck.

There are players and teams that take forever to get the ball out while others quickly recycle it.

blindsider

Interesting area.

How does your s/w tool differ from e.g. Sportscode Gamebreaker? I won't post the link, but you can google them - there are several companies who provide similar video analysis s/w.

This whole area is becoming more and more important in more and more sports.

It might be worth contacting the Video Analysis guys in your province or in the IRFU, and I know there's a guy in UCC who is heavily into analysis like this.

hardCopy

Is the system designed for Live use or for video playback?

Anything detailed will probably need a large team to do live analysis.

I did a couple of games for RTE when I was in college and it took two of us using a pretty simple program to do live tracking of position, possession, kicks, rucks, mauls, tackles, wide balls etc. at team level, i.e. we didn't even try to track individual player stats, I don't think it'd be possible to do anything more in depth with just two people and a video feed.

The biggest difficulty was at times when a number of 'events' happen at once. For example when a player kicks downfield you have:
Kick by team A
Change of possession
Change of position from one area to another
The person doing the stats has to hit three different buttons quickly and then look up to watch the receiving players response so button placement and UI are very important.

This stuff is less of a problem if you're working off recordings and can pause and rewind at will.

If you have access to the players then there's massive data to be mined from GPS transponders on the players. I'm pretty sure I read about coaches monitoring Paul O'Connell's top speed and average speed during training matches to assess whether he had fully recovered from an injury.

He has the data, he's asking what to look for within it.

HellishHeat

hardCopy wrote: »

Is the system designed for Live use or for video playback?

Anything detailed will probably need a large team to do live analysis.

I did a couple of games for RTE when I was in college and it took two of us using a pretty simple program to do live tracking of position, possession, kicks, rucks, mauls, tackles, wide balls etc. at team level, i.e. we didn't even try to track individual player stats, I don't think it'd be possible to do anything more in depth with just two people and a video feed.

The biggest difficulty was at times when a number of 'events' happen at once. For example when a player kicks downfield you have:
Kick by team A
Change of possession
Change of position from one area to another
The person doing the stats has to hit three different buttons quickly and then look up to watch the receiving players response so button placement and UI are very important.

This stuff is less of a problem if you're working off recordings and can pause and rewind at will.

If you have access to the players then there's massive data to be mined from GPS transponders on the players. I'm pretty sure I read about coaches monitoring Paul O'Connell's top speed and average speed during training matches to assess whether he had fully recovered from an injury.

I am the aforementioned colleague. Just to be clear: We are not interested in collecting match data (That sounds horrendous). We have obtained data from a third party and want to do something with the data that will be useful for a coach/analyst/bookmaker etc..

We have some ideas and we have processed and (we feel) enriched the data, but we wanted to gauge what "rugby people" felt were important attributes or trends of rugby players and teams.

To answer an earlier question, we do not have data pertaining to weather conditions, unfortunately.

We do have event data for 3 years of Heinekin cup, Magners and Amlin; about 1500 events per match.

DeDoc

Have a look at the relationship between scoring and some of your data and see how it varies by team and style (maybe defined by kicking percentages). How much do teams score and concede for given levels of posession. What do winning teams have in common - both on a match by match basis, and also between the better teams

If I could have a look at your data to see what the format is and what information you have I'd give more suggestions and ideas. What analytical tools are you using?

durkadurka

I suggest you get in touch with Bernard jackman. He's on twitter and is quite approachable on there, and he's quite interested in the science and analysis side.


@bernardjackman is his handle there .

A mate of mine did video work for him when he was coaching clontarf.

blindsider

I was trying to think of the name of the book/film earlier.....Moneyball.

Have a look here...http://en.wikipedia.org/wiki/Sabermetric

There's a lot of room for research in this area IMO. Some rugby team (and GAA, soccer, hockey etc also) somewhere will get this right and do very well out of it.

tolosenc

Anywhere I can get a copy of that data?

HellishHeat

tolosenc wrote: »

Anywhere I can get a copy of that data?

I'm afraid not, we've signed an agreement with our source.

phog

Is this data/analysis only of use to the specific teams/players that it was collected from or to teams that will be playing against them?

HellishHeat

phog wrote: »

Is this data/analysis only of use to the specific teams/players that it was collected from or to teams that will be playing against them?

We hope not. We have data on all teams in all the aforementioned competitions. We want to use this data to say general things about Rugby Union.

profitius

HellishHeat wrote: »

I am the aforementioned colleague. Just to be clear: We are not interested in collecting match data (That sounds horrendous). We have obtained data from a third party and want to do something with the data that will be useful for a coach/analyst/bookmaker etc..

We have some ideas and we have processed and (we feel) enriched the data, but we wanted to gauge what "rugby people" felt were important attributes or trends of rugby players and teams.

To answer an earlier question, we do not have data pertaining to weather conditions, unfortunately.

We do have event data for 3 years of Heinekin cup, Magners and Amlin; about 1500 events per match.

Any chance the data could be available for the public, even at a small fee? I've started a thread recently looking for a website with good rugby stats but theres none really. Theres an area that could be exploited!

HellishHeat

profitius wrote: »

Any chance the data could be available for the public, even at a small fee? I've started a thread recently looking for a website with good rugby stats but theres none really. Theres an area that could be exploited!

Sorry, as stated earlier, we have signed an agreement with the third party who provided the data.

I'm sure we can share our conclusions with the boards at the end of August.

profitius

profitius wrote: »

Any chance the data could be available for the public, even at a small fee? I've started a thread recently looking for a website with good rugby stats but theres none really. Theres an area that could be exploited!

If the data is comprehensive, it is worth thousands of euros. There's no hope of getting it free unless it's for academia I'd imagine.

prospect

I would attempt to categorise player types, based on common stats, for example, number of passes, short or long passes, box kicks, runs, tackles, types of tackle, ball carrying, grubber kicks (sp?).

You could have a comprehensive list of categories that the system places each type of your players in (on a given day) and compare them to their opposite number.

This way, when you look at the types of players on the opposition you can see what type of play has the most success in beating them man to man.

irishbucsfan

From a player/coach perspective,

The best use I can see for this is a team-by-team breakdown of patterns in play. For gameplanning purposes. So for example, it would be absolutely fantastic if I could use a system like this to analyse precisely what a team is likely to do in a certain situation.

So my input could be: Edinburgh Lineout, 30 meters out, 3rd quarter of the game, 5 point lead.

And the output would be what they are likely to do with the lineout itself and also what they are likely to do in the next few phases if they are succesful with the lineout.


Similarly a breakdown of player habits and patterns would be very very useful for a player. So a guy playing 13 for the Scarlets against Connacht could say "What is Eoin Griffin likely to do when he receives the ball on first phase from a set piece in the opposition half?" Or how is the opposition out half likely to use possession in his half when the team is down by 10 points or more in the second half, kick it away or keep possession?

I'm not sure how possible any of this is for you, but that would be fantastic to see.

.ak

irishbucsfan wrote: »

From a player/coach perspective,

The best use I can see for this is a team-by-team breakdown of patterns in play. For gameplanning purposes. So for example, it would be absolutely fantastic if I could use a system like this to analyse precisely what a team is likely to do in a certain situation.

So my input could be: Edinburgh Lineout, 30 meters out, 3rd quarter of the game, 5 point lead.

And the output would be what they are likely to do with the lineout itself and also what they are likely to do in the next few phases if they are succesful with the lineout.


Similarly a breakdown of player habits and patterns would be very very useful for a player. So a guy playing 13 for the Scarlets against Connacht could say "What is Eoin Griffin likely to do when he receives the ball on first phase from a set piece in the opposition half?" Or how is the opposition out half likely to use possession in his half when the team is down by 10 points or more in the second half, kick it away or keep possession?

I'm not sure how possible any of this is for you, but that would be fantastic to see.

+1, but it should look like this and have a distinct British accent when answering your queries....

attic

Shakakan wrote: »

Hi all

I thought I'd run something by your awesomely knowledgeable rugby brains

As part of a masters thesis, a colleague and I are analyzing data from rugby games. We have data from a couple of seasons of Rabo/Magners League, Amlin and HEC.

The data contains events from every game - so every pass, tackle, error, set-piece is captured along with who did it, where it was on the park and what the outcome was.

If you were a coach, or a player and you had a computer system that could tell you anything, what would you ask it?? Who is more likely to win was one question were looking at, but we're hoping for some more ideas that can give insight into player or team performance.

Any thoughts would be much appreciated!

Thanks
A

If you correlate something with who wins matches - I think you could have something world class.

padser

Something that's often not caught is the quality of a pass and it's outcomes.

So (taking Leinster as an example) it's easy to see that Madigan makes more running yards than Sexton when he gets the ball.

It's easy to see that Sexton make more passes than Madigan when he gets the ball.

What I would love to know is "Where is the ball likely to be 10 seconds after Madigan gets it in position x rather than Sexton". e.g. it could be 60% likely to be in Leinster possession and an average of 10 yards further up the pitch with Sexton and 55% likely to be in Leinster possession and an average of 12 yards up the pitch with Madigan (telling me that Madigan is higher risk, higher reward and a great option to bring on for the last 10 minutes when we are 8 points down).

A very crude way to start getting to this would be to look at measure the yards gained by Player B when Player A makes a pass to him and attribute those yards back to Player A's pass.

prospect

This kinda reminds me of the expression:

"Paralysis by analysis"

profitius

HellishHeat wrote: »

Sorry, as stated earlier, we have signed an agreement with the third party who provided the data.

I'm sure we can share our conclusions with the boards at the end of August.

Any updates?