Are we in a false position? Expected goals

[ColonelBlimp]

We're only 6 points off the play-offs, we can do it #COYR !!!

[Rab a dab doo]

The table does not lie. You are where you are on merit.

[Whoneedshuls]

I'd take top six now top two is great but I can see villa starting a run and really challenging us for second place let's just hope we maintain our form and get a couple of quality signings in and have a real go for it.

[jagerbob]

10 hours ago, Carl Sagan said:

Something I've jjust come across is the metric called "xG" which stands for "expected goals. Apparently it's an Opta thing and is explained in this video:

According to this statistic, we are massively over-performing based on the chances we're creating. If teams scored and conceded based on the quality of the chances, the top three would be Wolves, Villa and Cardiff with the Rams down in 13th place:

It is only one measure, but it's quite sobering to see, claiming we're not creating enough good chances and we're conceding too many good opportunities to be where we are. Or is it that we have such great quality in attack and defence that we're able to comfortably out-perform the rest of the division in these areas, and will be able to maintain our position?

kinda ********.

but i guess it is reflective of millwall creating loads of chances but never scoring (think morison has most shots at goal in the entire division, but only one goal )and ipswich having a ridiculously high conversion rate early season given they dont create clear chances anyway.

[jagerbob]

1 hour ago, Whoneedshuls said:

I'd take top six now top two is great but I can see villa starting a run and really challenging us for second place let's just hope we maintain our form and get a couple of quality signings in and have a real go for it.

top 6 is a shoe in.

i just cant see us finishing outside of the top 4, even with a fairly average 2nd half of the season. the points are on the table, and only the current top 3 have shown any sort of consistancy, the rest all all beat each other.

[Ambitious]

Interesting, this. I've always been intrigued by the numbers and what can be deciphered from them. In terms of professional sports & athletics, it's certainly the best tool in judging probability and making educated predictions. 

In this case, however, I don't believe we needed a complete rundown to tell us we've been fortunate on more than a few occasions this season. We don't necessarily have the lion's share of chances in the majority of our games, less possession than the opposition in more of them. Nevertheless, we've just got game management on our side, experience & quality at both ends of the pitch. 

I think we will be okay this season, but I'm certainly wary of the tide turning at some point.  

[Highgate]

3 hours ago, Rab a dab doo said:

The table does not lie. You are where you are on merit.

Of course tables generally reflect quite accurately the quality of football teams have played throughout a season but I've never understood why so many football fans have such an unwavering belief in their precision.  Some teams are bound to have ridden their luck or have had the benefit of more favourable refereeing decisions compared to others. 

And the very fact that there is 3 points for a win and 1 for a draw is a built in source of distortion. So no, league tables are not a perfect reflection of merit.

[Van der MoodHoover]

2 hours ago, Ambitious said:

Interesting, this. I've always been intrigued by the numbers and what can be deciphered from them. In terms of professional sports & athletics, it's certainly the best tool in judging probability and making educated predictions. 

In this case, however, I don't believe we needed a complete rundown to tell us we've been fortunate on more than a few occasions this season. We don't necessarily have the lion's share of chances in the majority of our games, less possession than the opposition in more of them. Nevertheless, we've just got game management on our side, experience & quality at both ends of the pitch. 

I think we will be okay this season, but I'm certainly wary of the tide turning at some point.  

I wonder whether - if this analysis had been done - we would have seen a similar  (if not more extreme) out performance in Billy Davies' promotion side?

[Kamikaze]

“Most people use statistics like a drunk man uses a lamppost; more for support than illumination”

[Tombo]

It's now hard to not change our target to be playoffs at a minimum. Not because we're all more excited, but because it would take a meltdown of McClaren proportions to mess this up now.

Most seasons, 75 points is enough to more or less guarantee 6th place. Sometimes it's as low as 68, and this is a weak year in the Championship so 75 should be more than enough. 

In our remaining 20 games, we could get 7 wins, 5 draws and 7 losses. P20 W7 D5 L7. That's a horrendous record by any stretch of the imagination, to only win 7 of your remaining 20 from that position in the league. Yet it still almost guarantees playoffs. 

If Rowett were to have got just 7 wins and 26 points from his first 20 games, he'd have been sacked by now. To not get in the playoffs now would be beyond all belief.

[ariotofmyown]

I wonder what our xg would be for the recent Millwall match? Seem to remember they had some great chances to score, but they were 3-0 down for most of them. We won easily, but might have got a better xg if we had scored 3 in the last 5 minutes instead, after we pressed them harder for 90 mins.

Don't think xg takes into account the match situation ie a team plays differently when winning.

Like against Ipswich, they had a few chances before scoring, then we became more defensive and they created little. If they had scored a bit earlier, we would have become defensive earlier too. This would have reduced their xg for the match without possibly changing the result.

[angieram]

What these (and all) statistics fail to take into account is human nature. Confidence, or lack of it, plays a big part in finishing. A player who has had 20 shots without scoring feels he's not going to score and this is amplified over time. 

We have a striker who thinks he's going to score every time he hits the ball. This makes him more likely to, in my opinion.

I'm not sure whether the same applies to defences, but Carson's goal has been living somewhat of a charmed life so far this season!

[Carl Sagan]

Comparing the expected goals (xG) table with the real table, we've scored 9 goals more than Opta say we should have done and conceded 7 goals fewer.

As others have said, that has to be down to how good Carson has been (as ever) and how lethal our strikers are for this division. And if we have that level of quality at both ends of the pitch, hopefully it augers well if we do end realizing that dream of promotion.

#CarsonForEngland!

[Owen87ITK]

4 hours ago, ariotofmyown said:

I wonder what our xg would be for the recent Millwall match? Seem to remember they had some great chances to score, but they were 3-0 down for most of them. We won easily, but might have got a better xg if we had scored 3 in the last 5 minutes instead, after we pressed them harder for 90 mins.

Don't think xg takes into account the match situation ie a team plays differently when winning.

Like against Ipswich, they had a few chances before scoring, then we became more defensive and they created little. If they had scored a bit earlier, we would have become defensive earlier too. This would have reduced their xg for the match without possibly changing the result.

[RamNut]

Don't know whether its relevant, but its quite noticeable this season that we shoot a lot more from outside the box.

at one time we rarely scored from outside box and were accused of wanting to walk it in to the net.

 

[RamNut]

3 minutes ago, OwenB87 said:

2-1 would have been fair.

[Comrade 86]

The team above us has more points and those below us less so I'd say we're where we deserve to be on balance. Could be argued that our nearest rivals Cardiff are punching above their weight given they've lost 4 from their last 5. Think Opta need a HTB stat - 'hard to beat'. That'd offer a more telling metric than where we have scored our goals from. 

[Van der MoodHoover]

1 hour ago, 86 points said:

The team above us has more points and those below us less so I'd say we're where we deserve to be on balance. Could be argued that our nearest rivals Cardiff are punching above their weight given they've lost 4 from their last 5. Think Opta need a HTB stat - 'hard to beat'. That'd offer a more telling metric than where we have scored our goals from. 

This is possibly Cardiff reverting to THEIR mean.

So logically, the same could happen to us. In fact, the statistics are suggesting that it is more likely we lose the next 4 than win the next 4, if we continue to play the same way. Note....."likely", not "certain".

[Van der MoodHoover]

5 hours ago, ariotofmyown said:

I wonder what our xg would be for the recent Millwall match? Seem to remember they had some great chances to score, but they were 3-0 down for most of them. We won easily, but might have got a better xg if we had scored 3 in the last 5 minutes instead, after we pressed them harder for 90 mins.

Don't think xg takes into account the match situation ie a team plays differently when winning.

Like against Ipswich, they had a few chances before scoring, then we became more defensive and they created little. If they had scored a bit earlier, we would have become defensive earlier too. This would have reduced their xg for the match without possibly changing the result.

Interesting point and it prompted a thought as to how often Opta rework their averages to take account of more recent data? I suppose it might be after each and every match they could just add in the latest shots, goals etc and update the xg

There's any number of spurious adjustments they could make to the data - eg giving more weight or "credibility" to recent results than older results but doesn't alter the basic fact that you measure what happened in the past and use that in some way to "guess" the future.

[brady1993]

23 hours ago, brady1993 said:

(This ended up being longer winded than planned will write about interpretation in a little while)

Oook so there's two elements to discuss here the validity of the statistic and the interpretation of the statistic. 

Statistic

Alright so the statistic is attempting to measure the number of goals you'd expect to score or concede based on the quality of the chances a team has. The quality of chance is defined through position on the pitch, the assist type, whether it's a head or a foot and whether it's a big chance. It's then assigned a percentage based on the likelihood of the "average" player scoring from there based on a training dataset. Then it's simply a matter of adding the percentage value for each chance to give you an overall expected goals scored or xG.

There are a few things about this that immediatley spring to mind:

First is that it doesn't appear to account for proximity to defenders, i.e. how likely a defender is to getting a block in.

It doesn't appear to directly account for what position the player taking the shot plays in (although is likely mitigated with the training data set).

They don't detail how the training data set was sourced (different leagues may have different probabilities from scoring from different positions). 

There is an implicit assumption by using such a cumulative statistic that if a player takes 10 shots, in the same situation, with say an expected 10% chance of scoring he will score a goal. However it's important to bear in mind that each shot is independent of what has happened before it and therefore the expected number goals would be lower. The probability of scoring a goal in that situation is not 1 and the model appears to overlook that.

Also I'd be very interested to see how strong the correlation is between xG and actual goals and how strong the correlation between x+/- is with results. For the obvious reason in both cases that it doesn't account for good finishing or good last ditch defending. 

It's probably a pretty good statistic for helping to judge a sides performance analytically but it isn't the be all end all and you should consider carefully how you apply it. 

Interpretation

You should always be very careful about how you interpret metrics such as these  because there outcome is dependant on quite a few variables going in and they will not be all encompassing.

@Carl Sagan it's easy to see why you proposed your original interpretation and it could prove true. However it's obviously directly contradicted by actual number of goals scored and actual number of goals conceded. We have scored the 2nd most goals in the league and we have conceded the 2nd most goals in the league, given that this is over a 26 game period I'd argue it's unlikely that our position is radically false. 

What's interesting though is if you take xG along with actual goals and the brilliant graphs that @Ellafella posted the other day (that I meant comment on but forgot). What they indicates is that defensively we concede roughly an average number of shots/chances but the proportion of those shots that go in is very low. This likely suggests that we are either forcing the opposition to shoot from far out/bad angles, good at getting blocks in, we have a good keeper or some combination of the above. From an offensive stand point we are having an average number of shots but a very good ratio of shots to goals. Suggesting that the chances we create are of higher quality than average or we have better finishing than average or both. 

All this is perfectly inline with our growing identity as a good counter attacking team and how games have gone in general this season. How many games have we had where the opposition has a good amount of the ball with quite a few shots but we never look like conceding ? How many games have we had were a few good chances and put them all away ? 

Interestingly it also suggests how we can improve; prevent more shots, create more shooting opportunities. The first could be aided with more legs in midfield and the second with more creativity in forward areas. Which are literally the two areas nearly everybody has suggested we could improve.