I am involved in a decent number of discussion regarding the relative performance of certain players/managers/front office personnel, and I often want to use stats for the basis of what I have to say. However, there isnt always a convenient source to look at just Revs performance, so I collated some of the information (courtesy of american soccer analysis) so that I can reference it here. 1st off, the revs performances to date relative to xG. https://docs.google.com/spreadsheet...single=true&widget=true&headers=false (The Red and Green highlights show wins (green), losses (red), and expected wins/losses (any xG value more than 0.4 from 0) A few observations: 1) The Revolution under Freidel significantly underperformed their xG. How much of that is due to variance in individual player performances vs. Freidel sabotaging his locker room is up for debate, and not something I want to delve too deeply into. Heck the numbers still think the Revs were a losing team under Freidel. However, tactically, his team was in a position to score considerably more goals than they actually did, and should have been much closer to a .500 outfit than they were at the end of his tenure. 2) The Revolution after Freidel have significantly overperformed their xG. Same caveats apply as before, however since Freidel was sacked (and without including the LAFC or SEA games for which I dont have data yet) the Revs still have the underlying numbers of a slightly sub .500 team. 3) Matt Turner is an excellent shot stopping GK. Looking at the Brad vs post-Brad xG numbers, the Revs defense is allowing roughly the same volume of scoring opportunities (22.5 in 12 games vs 18.3 in 11 games). The major difference is who has been tasked with keeping the ball out of the net. This is a stat that tends to fluctuate some from season to season, however I encourage you to look at his excellent 2018 numbers as well. I guess what the numbers ask is "are the Revs really that much better with Arena at the helm?" I'm not sure there is an objective answer. Its clear that Arena has brought a level of calm to the team that has translated to increased success. On the other, The team Freidel fielded week in and week out seems to have been a bit more hard done by bad luck (and Teal Bunbury's mercurial finishing skills) than perhaps he has been given credit for.
The question of "what correlates with consistently overperforming against xG" is an interesting question. But I think one reason we're overperforming xG is now we have Bou finishing chances, and he's going to finish at above xG rate. Another potential reason is that I've read xG models are "outperformed" by passive defenses (soak up pressure, allow shots under pressure instead of trying to win the ball), and Friedel was the opposite of that.
My takeaway is that we've gone from scoring 0.92 goals a game and conceding 2.5 goals per game to scoring 2.17 goals per game and conceding 1 goal a game. That is a marked difference on both sides of the ball. Pretty incredible, but expected given the results (or is it the other way around?) I added the Seattle result to the table provided to get my numbers.
But not the LAFC game? In both games the opponent scored fewer goals than their xG suggested, and against Seattle the Revs exceeded their xG by more than a goal. So in that regard nothing has changed regarding the trend under Bruce.
Fair enough and right things don't change much relative to the trend under Bruce being WAY better than under Friedel: updated numbers show us scoring 2 goals a game and giving up 1.08 goals a game under Bruce both way better - 'twice Ias good' figuratively speaking. The statistics would say in these categories that Friedel was 100% worse than Bruce - that's pretty damning.
Except that according to the underlying numbers (the ones measuring the chances the team had vs the chances the opponent has) the team has actually gotten worse under Bruce. While the team played like a .500 team under Freidel, they've played like a .200 team under Bruce. The only difference is that instead of massively underachieving their on field production for Freidel, they are now massively overachieving their on field production for Bruce. I worry, because over time these stats tend to regress toward the mean, which could result in the end of the season getting ugly.
There is a recent interview done by 538.com with political pollster Ann Seltzer, which is relevant to this discussion. In brief, her point is that past practice is predictive of future events, except if the factors that generated the past results have changed. Thus while one could argue that the Revs will inevitably regress to the mean (generally a good assumption), the counterargument is that in this case, the true mean is unknown.
Its a good point, and some things have changed (Bou and Arena). However the models that define xG have not. They dont care who the manager is, or what players they put on the field. All that matters for xG is how good the chances the players on the field produce. Right now, the Revs under Arena are outscoring their xG by +5 and are -10 compared with xGA. Those are not good indicators for future success! They are the kind of numbers that point to a team getting very lucky. If xG dictated results (I know it doesnt, thats why they play the games), they would be 3-3-7 (W-L-D)! These numbers are now past performance, and I worry that they predict the Revs future success.
IOW, we're finishing better. Part is explained by Bou, who clearly finishes a high percentage of his chances and Gil, who is suddenly not the primary threat, so he gets somewhat less attention. Well, that's just a disclaimer, not a statement of fact or truth. Just like "repeating the same thing over and over and expecting different results is the definition of insanity". It keeps getting repeated but it was never true.
That is correct, it's a disclaimer. But anytime you try to predict the future with statistical analysis and human beings are involved; you need a disclaimer!
At the risk of getting wonkish....but this is a Statistical discussion, right? I will grant that xG and xGA are impartial numbers and over the course of time across a big enough sample set, the numbers will come close to the expected. However, as with all statistics like this it depends on the sample size. For instance, would you expect Harry Kane and Gyassi Zardes (or Teal Bunbury for that matter) to have the same xG given the same shot opportunities? If the answer to that question is yes, then whether your team is above or below xG will, in part, depend on the quality of your scorers. I will grant you that Kane will create more chances, but still even presented the exact same chances I would expect Kane to be over the league average xG and Zardes to be below it. The same really goes for xGA as well -- if you have a better keeper/defense than average you should expect your xGA to be lower than a team with a lesser keeper/defense. That said, I do think our xGA is a concern, we do seem to give up a lot of good opportunities and need to do much better at limiting chances.
Thats an excellent point about quality of shot taker. There are a small handful of players who year to year will reliably exceed their xG. Their names include Ronaldo, Messi, etc... Bou, despite being good, is not in that category. He is good enough that over a 5 season stretch he should end up close to his xG though. Right now he has 3 goals on 1.55 xG. Its unlikely he continues to score more than double his xG. Gil might be the best Rev ever, and like Bou in a single season could outperform his xG. Right now he has 9 on 6.76 xG, which is (again) unsustainable. Teal has his merits, but consistent finishing isnt one of them. I'd expect him to consistently underperform his xG. And he has! Including his numbers from Freidel's reign he has 5 goals on 6.42 xG. Looking at 2018 he had 11 on 11.26, and from 2017: 7 vs 6.74, and 2016: 2 vs 5.79. Thats the picture of below average finishing! There is a not unpopular theory within the statistical community that finiahing ability is a myth. I'm not sure i buy it yet, but for players that arent in the superstar category shot volume tends to be an accurate predictor of goal production. The revs are definitely putting more shots on target under Arena (a good thing!) but the over correction in xG they are experiencing feels unsustainable
Interesting that if you add the xG and xGA from Friedel and Arena together the Revs appear to be right on track (from the table in the initial post, I need help understanding your point here. In looking at the numbers, during Friedel's time, we were 17.6 and 22.57 for xG and xGA, under Arena those numbers went to 18.8 and 18.34. While a point can be made that you could expect a .500 record during Arena's time given those numbers and that we have overachieved significantly, I am not sure how that says we have actually performed worse under Arena? We seem to be generating more xG and less xGA per game under Arena, how is that a worse performance?
That is VERY interesting -- so the mark of a scorer isn't the finishing, it's the chances they create...at least that is the theory. I am very skeptical...and I would be interested to see the numbers behind that theory. I would think that there is a range of finishing capability among forwards and that would show up with how well they stack against their xGs, per your comment about Teal, given his abilty you would expect him to underperform and he has.
I looked at xG on a game by game basis. Any game that totaled an xG difference greater than 0.4 i counted as an expected win or loss. Tallying up those results gives a few lopsided wins (3 wins by 2 ish goals each) and many losses (5 losses by 1ish goal each). Basically the good Arena (and Lapper) performances were really good, but in general the team has played worse than its opposition. If you include the Seattle game (1.8 xG, 2.5 xGA) and the LAFC game (0.5 xG, 2.8 xGA) the numbers get worse (they are included in the 3-3-7 but not in the spreadsheet. They werent published until after I wrote the original post)
I'm not sure whether the stats reflect it or not, but two very important factors for success in the game IMO are 'opportunism' and 'resilience'. In this context I'd define them as: opportunism - the ability to convert limited chances; the ability to create/convert chances when they are most needed resilience - the ability to withstand the efforts of the opponent even when being out-played or having suffered unfair or unfortunate events. So, if you can resist giving up goals while you are being outplayed, statistically and otherwise, and take advantage of the limited chances a quality defense may allow, I think the team is likely to find success in spite of the statistics (or at least until stats find a way to account for those qualities).h
I think taking stats like xG and xGA and applying them to an individual game is not really appropriate. I get that 'over a season' you would expect a return towards the mean, but for a particular couple of games not so much. I do not have the means, nor probably the wherewithal, to look at each of our opponents and see how they have performed over the season against their xG/A expectations. But it would be interesting to see how well it applies across the league. If xG/A stats are the same for ANY league in the world then I would continue to be skeptical of them - say what you will, but the xG for just about any team in USL would be expected to be lower than just about any team in the Premier League let's say. I can see that the goalies/defenders are less as well, but still you have to hit the ball into the right spot with the right pace. Just seems counter-intuitive to think the stats apply equally to all soccer players no matter the level.
I have 0 desire to look at each team's performance in each individual game and collate all that data, at least not without being paid to do it! However just looking at aggregate season stats the majority of teams seem to track pretty well with their xG numbers. Sorting by the xGD column, LAFC should have far and away the best record, but Chicago should be the 2nd best team in the league (way underperforming their chances and allowing easy goals). KC has also underperformed its stats. On the other end, the model predicts Vancouver and FCC to both be terrible. The outliers are DC (-13.2), NE (-8.8) and Portland (-5). The unifying theme is that all of them have gotten above average contributions from their gK. Bill Hamid has a track record of doing that, and Jeff Attinela did the same thing last season, so maybe there is something about Portland's defense. As to you PL comment: you rightly should expect attacking players to be better in better leagues. However, defenders and GK are also better. In the end the improved quality on both ends washes out and the numbers stay pretty normal.
Thanks, never implied you should do that work either, but it would an interesting piece of work. Thanks for the analysis you did do.
I went and read the explanation of xG. I interpret it a little differently. I believe that xG is simply an average shot completion. If a player takes a shot from the top of the 18, they should make it x percent of the time. Players like Bou and Gil make it x+; while players like Teal and Juan make it x-. So can the Revs continue to exceed their xG? Sure, as long as Bou and Gil continue to take the shots. According to the latest tables 6 of 19 players on the Revs have exceeded their xG. Conversely, 13 of 19 players have not exceeded their xGs. The "problem" with stats and sports is that stats tend to represent the 'average" player. Your better players should always exceed the average.
Another thing to look at is the correlation between key passes and assists. Castillo has 5 assists on 14 key passes. Meanwhile Bye has only 1 assist on 17! Other outliers include Agudelo (1/22) and Bunbury (1/23)! Key passes seem to correlate ~10:1 with xA, so both Castillo (and Penilla 4/22) seem to benefit from some exceptionally good finishing from their team mates.
Stats on the NYRB game are in. Revs xG: 0.5, xGA: 2.83. Your weekly reminder that I dont think the Revs recent success is sustainable (their record under arena is 7-6-1 with a +12 GD, despite their chances suggesting theyre playing like a 3-3-8 team with a -4.95 GD). On the season Matt Turner now leads the league in xG difference for keepers (-6.94). This despite facing shots that are on average closer than everyone else in the top 10 (shots taken from closer to the goal have a higher xG. This means he is stopping more harder shots as opposed to feasting on a diet of low % long distance shots). Kudos to Turner, I still love him and think people still under rate him. (After the LAFC game I had a guy in the parking lot try to convince me Knighton should be the GK. It still bothers me a little)
I know Knighton has been on bench, but did Cropper go back onto the milk carton again or was he loaned out?