Author Topic: WBA (H) Round Table  (Read 29823 times)

Offline GrkStav

  • Has a statuette of Lucas on the bonnet of his car which he polishes lovingly with Lucas Brasso. Glen Johnson's biggest fan. Doesn't have a "fucken clue" where L4 is
  • Legacy Fan
  • ******
  • Posts: 8,981
  • Not very good at 'banter'.
Re: WBA (H) Round Table
« Reply #160 on: November 1, 2013, 03:52:39 am »
By the way, I read the entire piece that was linked to, earlier.

I found the part of the paper entitled "The hot hand test statistics" to be very illuminating.

In effect, it deals with whether the data that would be required for any of the statistical tests that have been used to reject the null-hypothesis of there being no such thing as 'hot hand'/hot streak are so outlandishly unrealistic that the fact that there are ANY studies that show even modest statistical support and/or strong statistical support for extremely modest 'hot streak' effects is remarkable.

In short, if you can't even put together a simulation that reasonably models what 'common sense' classifies as a 'hot streak' and have it 'beat the statistical test' used, then there's something problematic about the statistical test.

As a result, there really is NO evidence disproving even the hot-hand/hot streak hypothesis, let alone the 'form' argument (which, as I argued earlier, is not, at present, a scientifically testable one).
Ludi Circenses!

Offline ElstonGunn

  • is Billy Beane
  • Legacy Fan
  • ******
  • Posts: 3,271
  • We all Live in a Red and White Kop
Re: WBA (H) Round Table
« Reply #161 on: November 1, 2013, 11:41:45 am »
So why pick teams then? Why buy expensive players? Why not just cobble together 25 lads for far less money, and hope that random chance turns them into champions? Because if a run of three good games does not lend itself to a REASONABLE EXPECTATION of a fourth good game, then the converse is also surely true - a run of 3 poor games doesn't mean the fourth will also be poor, so there's no need to drop a badly performing player, correct? Similarly, having a large body of quality work is also no predictor that Agger would do better than Skrtel if selected in his place, yes? So essentially, if form doesn't exist in any statistical instance, then managers can just throw names into a hat and select random combinations of 11 players for each game, with a 50% chance that they will be successful. But we both know that's nonsense. So why do some players get selected as ever-presents, and others don't? Why do some players score at a rate of 1 goal every 2 games, but a different striker in the same system scores only 2 in 20? Why do we select the higher scoring striker for each subsequent game, and not the low-scoring striker, if past performance is no indicator of immediate future performance? I know of the basketball studies that were performed in terms of the "hot hand" - but my challenge to that is - why would a player be selected with any regularity, if that is the case? If there is no statistical correlation between one performance and the next, then team-building is actually obsolete, it would seem.
Sorry, I had to leave last night, as I've been enjoying the discussion. I really don't understand why you think the existence or nonexistence of form has anything to do with this this parade of horribles about not being able to tell the difference between players' quality. As you've said yourself, form and quality are simply difference things. I've already explained the different as best as I know how--the question of team building based on past performance has literally nothing to do with whether or not form is real. 

Anyway, as I clarified (because I've been misstating it), it's not so much that form might not exist (as you and prof say, underlying psychological and physical condition of course matters), as whether or not it is predictable in any meaningful sense on the basis of recent performance, even for someone on the ground like the manager.  Regardless, perhaps it's best to leave this off now. I actually don't have anything riding on the answer either way! Maybe csgreen will come back eventually and say more what he believes.

Offline GrkStav

  • Has a statuette of Lucas on the bonnet of his car which he polishes lovingly with Lucas Brasso. Glen Johnson's biggest fan. Doesn't have a "fucken clue" where L4 is
  • Legacy Fan
  • ******
  • Posts: 8,981
  • Not very good at 'banter'.
Re: WBA (H) Round Table
« Reply #162 on: November 1, 2013, 02:39:36 pm »

Anyway, as I clarified (because I've been misstating it), it's not so much that form might not exist (as you and prof say, underlying psychological and physical condition of course matters), as whether or not it is predictable in any meaningful sense on the basis of recent performance, even for someone on the ground like the manager.  Regardless, perhaps it's best to leave this off now. I actually don't have anything riding on the answer either way! Maybe csgreen will come back eventually and say more what he believes.

On that level, for someone on the ground like the manager and his senior staff, I think it is predictable.

Still, there's form and then there's suitability for the tactics chosen and opponent faced. :wave
Ludi Circenses!

Offline Col

  • Shaves his tongue and shares makeup tips. May be a little camp.
  • RAWK Supporter
  • Legacy Fan
  • ******
  • Posts: 18,386
Re: WBA (H) Round Table
« Reply #163 on: November 1, 2013, 08:47:21 pm »
You have to do it from the outset though. You can't come out and say that players have to compete for places in one instance, and then switch it on a whim. Players lose respect for managers when they do that. So in some ways, Rodgers made a rod for his own back in having competition as his philosophy. But the other option - horses for courses regardless of form - has it's own drawbacks also. The most important thing is to have a philosophy and stick with it.
It's possible to do both, provided the communication is clear and the players involved buy in to the reasoning. It also shouldn't happen too frequently, as it gives off an aura of indecision and pandering to certain players.

Competition for places at the top level is great to have, on the proviso that when the stronger players in the group are out of form, or lose their place due to injury or suspension, then they have to earn their place back as we're currently seeing with Agger.

There's no doubt that Agger is a superior player to Skrtel, but whilst Skrtel's form is decent enough to not warrant being dropped from the team, things are rightly staying as they are.

(The other option is to change the system to accomodate both, which is not easy to do when you've got 4 players fighting for 2 or 3 spots, but we've done well with Suarez' return by tweaking things to get both him and Sturridge playing centrally at the expense of Aspas and our whole crop of 'wide' players - none of whom were playing well enough to demand a place in the starting team).

I think Rodgers has gotten both of these decisions spot on so far. Playing as the 3rd defender in a close group of 3 has allowed Skrtel to do what he feels comfortable doing, and despite the fact he's not as accomplished a player as Agger, right now circumstances dictate he has every right to be starting games ahead of him.
I don't have to sell my soul... he's already in me.

Offline csgreen

  • Going through RAWK like a dose of Verucca Salts
  • Anny Roader
  • ****
  • Posts: 473
  • Chairman Mao has never seen a greater show of red.
Re: WBA (H) Round Table
« Reply #164 on: November 1, 2013, 10:12:48 pm »
So why pick teams then? Why buy expensive players? Why not just cobble together 25 lads for far less money, and hope that random chance turns them into champions? Because if a run of three good games does not lend itself to a REASONABLE EXPECTATION of a fourth good game, then the converse is also surely true - a run of 3 poor games doesn't mean the fourth will also be poor, so there's no need to drop a badly performing player, correct? Similarly, having a large body of quality work is also no predictor that Agger would do better than Skrtel if selected in his place, yes? So essentially, if form doesn't exist in any statistical instance, then managers can just throw names into a hat and select random combinations of 11 players for each game, with a 50% chance that they will be successful. But we both know that's nonsense. So why do some players get selected as ever-presents, and others don't? Why do some players score at a rate of 1 goal every 2 games, but a different striker in the same system scores only 2 in 20? Why do we select the higher scoring striker for each subsequent game, and not the low-scoring striker, if past performance is no indicator of immediate future performance? I know of the basketball studies that were performed in terms of the "hot hand" - but my challenge to that is - why would a player be selected with any regularity, if that is the case? If there is no statistical correlation between one performance and the next, then team-building is actually obsolete, it would seem.


Sorry to jump back in late (after starting this debate ;-D). 

I think the issue here is with how you're using "good" and "poor" above.  In my world those terms are relative to the player's mean level of performance - they're not absolute values.  So, a "good" game for Messi is WAY better than a "good" game for Aspas (and frankly a "bad" game for Messi may also be).  So given that framework, and if things are independent, then indeed, a run of three "good" games (i.e. games where you play better than your average level of performance) does not lend itself to a REASONABLE EXPECTATION of a fourth "good" game.  The probability of Suarez scoring in the next game is no higher given that he scored in the last game than it would be if he hadn't scored in the last game.  The reason you pick Suarez though is that he has a much higher mean level of performance.  Simply put - his base probability of scoring is high.  So someone asked above whether I would bet on Sturridge scoring in the next game - and the answer is - I sure as hell wouldn't if the odds I was getting weren't worth it given his mean level of scoring for Liverpool.  The other kind of thinking is exactly why people go broke playing blackjack...   

Then to the question of why buy expensive players - if we think of a dice game where the goal is to roll the number 6.  Well - you're much better off buying an "expensive" die that has the number 6 on five faces than you are buying a cheaper die that has the number 6 on one face.  But neither die can be on a run of "form" - the probability of rolling a 6 with the expensive die is 83% irrespective of whether it was a 6 on the last roll or not and the probability of rolling a 6 with the cheap die is 16% irrespective of its previous roll result.

So the question of Skrtel depends on whether you think his recent results indicate an increase in his "mean level of performance" (which can absolutely be true - because that doesn't just depend on him - it also depends on things like the formation - his expected level of performance might very well be much higher given three at the back).  Picking Skrtel for that reason is totally rational.  Picking him because he's played above his average for three games in a row and you think he's thus more likely to play above average again is distinctly irrational... 


Let's go a little more philosophical then:

Do you think that the sun will rise tomorrow?

The sun example isn't really appropriate.  If the sun didn't rise today it does drastically affect the estimate of whether it will rise tomorrow (i.e. there is necessarily temporal dependence in that problem)... 

As for references - if you search for "the form fallacy" - you'll find a decent number of things for football.  However, the best data is - as pointed out by ElstonGunn - in basketball and baseball (primarily because those sports lend themselves to advanced metrics much better than football).   

Offline csgreen

  • Going through RAWK like a dose of Verucca Salts
  • Anny Roader
  • ****
  • Posts: 473
  • Chairman Mao has never seen a greater show of red.
Re: WBA (H) Round Table
« Reply #165 on: November 1, 2013, 10:21:27 pm »
In short, if you can't even put together a simulation that reasonably models what 'common sense' classifies as a 'hot streak' and have it 'beat the statistical test' used, then there's something problematic about the statistical test common sense.

Honestly though it's not that difficult to find examples in the world where science absolutely does reject the null hypothesis of temporal independence.  For instance, even though the results of a sequences of coin flips doesn't violate independence, people's guesses as to heads/tails during that sequence will almost certainly violate independence (in fact - it's almost impossible for an untrained human being to generate a sequence of independent choices).       

Offline Prof

  • fessor Yaffle. Full tosser.
  • RAWK Scribe
  • Legacy Fan
  • ******
  • Posts: 8,930
  • We all Live in a Red and White Kop
    • The Alternative Premier League Table
Re: WBA (H) Round Table
« Reply #166 on: November 1, 2013, 10:46:51 pm »
However, the best data is - as pointed out by ElstonGunn - in basketball and baseball (primarily because those sports lend themselves to advanced metrics much better than football).
Er, no.  It's because the statistics used have high frequencies, like pass completion in football.  High frequencies reduce the effect of experimental error, which is why it's better to have large sample sizes.

Where 'form' becomes relevant is where there are lower frequencies of specific events occurring which are, due to their low frequency, also affected more by psychology.

If you make a mistake with something you do 100 times in a game, you will probably put it right with only a statistically insignificant effect on your success rate.  It might even make you take more care/time to ensure successful completion, but effectively slow the game down.

If you make a mistake for a low frequency task, you have less opportunity to make up for this, and also, it will be more likely to affect your confidence.  Low frequncy events are less likely to be autonomous, making them more vulnerable to cognitive processes, affected by mistakes.

In any case, players who feel good about their game (confidence/efficacy), for whatever reason (including thinking they are in good form), are more likely to play well.  There is no way that can even be disputed with the amount of emperical evidence there is that demonstrates that.

What is frustrating me about this discussion is that it is even taking place.  Form has absolutely nothing to do with the hot-streak research.  It is being completely misused to argue something completely different.

Offline TSC

  • RAWK Supporter
  • Legacy Fan
  • ******
  • Posts: 25,342
  • We all Live in a Red and White Kop
Re: WBA (H) Round Table
« Reply #167 on: November 1, 2013, 10:52:18 pm »
Just read the last page of this, hard reading to be honest.  Here's a thought.  How about just enjoying a victory with probably the best collection of goals you're likely to see in one game this season.

Offline csgreen

  • Going through RAWK like a dose of Verucca Salts
  • Anny Roader
  • ****
  • Posts: 473
  • Chairman Mao has never seen a greater show of red.
Re: WBA (H) Round Table
« Reply #168 on: November 1, 2013, 11:01:54 pm »
Er, no.  It's because the statistics used have high frequencies, like pass completion in football.  High frequencies reduce the effect of experimental error, which is why it's better to have large sample sizes.

I shouldn't have said "lend themselves to advanced metrics" - I should have said lend themselves to base statistical analysis (and then through that to advanced metrics).  Those sports have a large number of clearly defined "states" that can considered roughly equivalent and thus compiled across.  Football doesn't.   

What is frustrating me about this discussion is that it is even taking place.  Form has absolutely nothing to do with the hot-streak research.  It is being completely misused to argue something completely different.

It has everything to do with hot-streak research - at least in how "form" is used on this forum (maybe you have a different definition which is fine).  But on this forum look at how many times in the past week someone has suggested that you "don't want to drop a player while they're on form" or "you want to milk their current good form for all it's worth".  That suggests a clear belief that good performance at time one is predictive of good performance at time two - above and beyond the base probability of good performance.  That is 100% reflective of a belief that players can be on a "hot streak"... 


Offline Sangria

  • In trying to be right ends up wrong without fail
  • Legacy Fan
  • ******
  • Posts: 19,075
  • We all Live in a Red and White Kop
Re: WBA (H) Round Table
« Reply #169 on: November 1, 2013, 11:55:57 pm »
I shouldn't have said "lend themselves to advanced metrics" - I should have said lend themselves to base statistical analysis (and then through that to advanced metrics).  Those sports have a large number of clearly defined "states" that can considered roughly equivalent and thus compiled across.  Football doesn't.   

It has everything to do with hot-streak research - at least in how "form" is used on this forum (maybe you have a different definition which is fine).  But on this forum look at how many times in the past week someone has suggested that you "don't want to drop a player while they're on form" or "you want to milk their current good form for all it's worth".  That suggests a clear belief that good performance at time one is predictive of good performance at time two - above and beyond the base probability of good performance.  That is 100% reflective of a belief that players can be on a "hot streak"... 

It's indicative of not much more than the fact that fans have less data to base judgements on than is ideal. It doesn't mean form does not exist, as in a single word expression of a conglomerate of physical, mental and psychological condition for a task. It might have been Lobanovskiy who said he picked teams based on their performances in video games designed to test their reactions. Whatever the suitability of the data (and the players he cited as best performers were virtually a who's who of 80s Soviet greats), that's an example of data not available to fans, but which is used to measure "form". With less information at hand, fans use other metrics to measure the same thing. The rawest data is past results, and whatever their crudity, they do at least express with some accuracy an important part of form, confidence. Players who score goals are happier. You don't need scientific measurements for that. You just need to have played the game and experienced games where you've scored and games where you haven't.
"i just dont think (Lucas is) that type of player that Kenny wants"
Vidocq, 20 January 2011

http://www.redandwhitekop.com/forum/index.php?topic=267148.msg8032258#msg8032258

Offline i6uuaq

  • RAWK Supporter
  • Kopite
  • ******
  • Posts: 645
  • Hmm... what's this personal text thing, then?
Re: WBA (H) Round Table
« Reply #170 on: November 2, 2013, 12:07:21 am »
Er, no.  It's because the statistics used have high frequencies, like pass completion in football.  High frequencies reduce the effect of experimental error, which is why it's better to have large sample sizes.

Where 'form' becomes relevant is where there are lower frequencies of specific events occurring which are, due to their low frequency, also affected more by psychology.
...
What is frustrating me about this discussion is that it is even taking place.  Form has absolutely nothing to do with the hot-streak research.  It is being completely misused to argue something completely different.

I don't think it has nothing to do with hot streak research. I think that your argument is that we cannot apply the results of hot streak research in baseball to football, due to the different nature of the events. But I think that the basic research methodology and statistical methods can still be applied to try and find the influence of form on different metrics in football.

It's the same problem everyone brings up about statistics in football - it's inherently a low-scoring game, so the number of specific events are usually too low to be statistically meaningful. But that hasn't stopped people from trying to find meaningful statistics, such as final-third pass completion, shot-on-target percentage etc. Would it be possible to find one of these metrics to apply the hot-streak methods to?
"I've not seen it and I'm not being Arsene Wenger," Dalglish said. "If there's something untoward then I am sure the governing body will act appropriately."

Offline Prof

  • fessor Yaffle. Full tosser.
  • RAWK Scribe
  • Legacy Fan
  • ******
  • Posts: 8,930
  • We all Live in a Red and White Kop
    • The Alternative Premier League Table
Re: WBA (H) Round Table
« Reply #171 on: November 2, 2013, 12:25:46 am »
Hot-streak research is about the claim that a player is exceding their normal rate of scoring for a specific condition.

We are talking about Skrtel playing well, not Gerrard's shooting.  A centre back's performance is assessed by all parts of his game, not a solitary, discrete skill.  Skrtel is playing well, he's in form, not on a hot streak where random error is at play.  They are quite clearly completely different concepts, so extrapolating hot-streak research to dismiss Skrtel's good form is not valid.

Not wanting to drop a player in form is more about wanting players not to lose their confidence and to reward the hard-work that has seen them play well.  Not because he's on a lucky streak.

Offline GrkStav

  • Has a statuette of Lucas on the bonnet of his car which he polishes lovingly with Lucas Brasso. Glen Johnson's biggest fan. Doesn't have a "fucken clue" where L4 is
  • Legacy Fan
  • ******
  • Posts: 8,981
  • Not very good at 'banter'.
Re: WBA (H) Round Table
« Reply #172 on: November 2, 2013, 04:40:49 am »
Honestly though it's not that difficult to find examples in the world where science absolutely does reject the null hypothesis of temporal independence.  For instance, even though the results of a sequences of coin flips doesn't violate independence, people's guesses as to heads/tails during that sequence will almost certainly violate independence (in fact - it's almost impossible for an untrained human being to generate a sequence of independent choices).       

Don't cross out what I wrote and substitute some bullshit.

I meant precisely what I wrote. I read the long piece and paid attention to that specific part I referenced.

I've been around the block a couple of hundred times with regard to using statistics outside of sports. Not in the natural sciences, fair enough, but in a 'scientific' field nonetheless.

In this area, you have to find a way mathematically to model, as validly as possible, the nebulous common-sense notion of a hot-streak (to go back to the research that HAS been done), let alone the even more nebulous and much more connotative concept of 'form'. Then, you have to devise a statistical test that is sensitive enough to be a true test of the 'hot streak' theory. As things stand, no such tests have been developed (or rather deployed).

I already stated (politely) that the theory regarding 'form' is not, at present, empirically testable. Its proponents CAN indeed effectively demonstrate that 'form' is NOT the functional equivalent of 'hot streak'. What they cannot do, yet, is shoot down any quantitative or simulation research that attempts to state the 'form theory' formally, deduce some in principle testable hypotheses, attempt validly to operationalize the main variables, and attempt some empirical testing. Until and unless the proponents of the 'form theory' can tell us, in quite a bit of detail, what kind of empirical data would disprove it, it's not a scientific theory.
Ludi Circenses!

Offline csgreen

  • Going through RAWK like a dose of Verucca Salts
  • Anny Roader
  • ****
  • Posts: 473
  • Chairman Mao has never seen a greater show of red.
Re: WBA (H) Round Table
« Reply #173 on: November 2, 2013, 05:00:39 am »
Don't cross out what I wrote and substitute some bullshit.

I meant precisely what I wrote. I read the long piece and paid attention to that specific part I referenced.

I've been around the block a couple of hundred times with regard to using statistics outside of sports. Not in the natural sciences, fair enough, but in a 'scientific' field nonetheless.

In this area, you have to find a way mathematically to model, as validly as possible, the nebulous common-sense notion of a hot-streak (to go back to the research that HAS been done), let alone the even more nebulous and much more connotative concept of 'form'. Then, you have to devise a statistical test that is sensitive enough to be a true test of the 'hot streak' theory. As things stand, no such tests have been developed (or rather deployed).

I already stated (politely) that the theory regarding 'form' is not, at present, empirically testable. Its proponents CAN indeed effectively demonstrate that 'form' is NOT the functional equivalent of 'hot streak'. What they cannot do, yet, is shoot down any quantitative or simulation research that attempts to state the 'form theory' formally, deduce some in principle testable hypotheses, attempt validly to operationalize the main variables, and attempt some empirical testing. Until and unless the proponents of the 'form theory' can tell us, in quite a bit of detail, what kind of empirical data would disprove it, it's not a scientific theory.

I agree with this 100% with regard to form (and note: I wasn't being snarky to you in my crossing out - more toward the many who have more or less told me to fark off because 'common sense' says that form is a thing).  My tendency is to think that things that cannot be tested for have no value...

Offline csgreen

  • Going through RAWK like a dose of Verucca Salts
  • Anny Roader
  • ****
  • Posts: 473
  • Chairman Mao has never seen a greater show of red.
Re: WBA (H) Round Table
« Reply #174 on: November 2, 2013, 05:10:29 am »

We are talking about Skrtel playing well, not Gerrard's shooting.  A centre back's performance is assessed by all parts of his game, not a solitary, discrete skill.  Skrtel is playing well, he's in form, not on a hot streak where random error is at play.  They are quite clearly completely different concepts, so extrapolating hot-streak research to dismiss Skrtel's good form is not valid.


Let's imagine that Anelka put away the chance where he clearly beat Skrtel last weekend.  Would Skrtel still be considered to be in a vein of good form?  Or if WBA had scored when Skrtel headed the ball directly to them at the top of our 18?  If either of those had occurred my guess is that the majority of the forum would be calling for Skrtel to be dropped due to his recent poor form despite the fact that he would have played exactly the same game as the one that got him lauded when WBA didn't put either chance away.  So of course random error is at play in determining "form" (or not random error, but a system that is complex enough that we can't make any reasonable predictions).   

Offline deadsetred

  • Legacy Fan
  • ******
  • Posts: 1,056
  • We all Live in a Red and White Kop
Re: WBA (H) Round Table
« Reply #175 on: November 2, 2013, 06:02:38 am »
Let's imagine that Anelka put away the chance where he clearly beat Skrtel last weekend.  Would Skrtel still be considered to be in a vein of good form?  Or if WBA had scored when Skrtel headed the ball directly to them at the top of our 18?  If either of those had occurred my guess is that the majority of the forum would be calling for Skrtel to be dropped due to his recent poor form despite the fact that he would have played exactly the same game as the one that got him lauded when WBA didn't put either chance away.  So of course random error is at play in determining "form" (or not random error, but a system that is complex enough that we can't make any reasonable predictions).   

Not so fast sonny. You can't do that. All defenders will make minor mistakes in a game - the more you make, the more chance overall the opposition has of scoring. Skrtel's made relatively few in recent weeks, and hasn't hidden, which is another thing you must consider. Also, I don't think its constructive to question a player's place in the team just because the opposition capitalizes on a mistake they make. You've got to look at their overall reliability and the context in which the mistake is made. Not too fond of your analysis truth be told.

Offline GrkStav

  • Has a statuette of Lucas on the bonnet of his car which he polishes lovingly with Lucas Brasso. Glen Johnson's biggest fan. Doesn't have a "fucken clue" where L4 is
  • Legacy Fan
  • ******
  • Posts: 8,981
  • Not very good at 'banter'.
Re: WBA (H) Round Table
« Reply #176 on: November 2, 2013, 06:29:54 am »
I agree with this 100% with regard to form (and note: I wasn't being snarky to you in my crossing out - more toward the many who have more or less told me to fark off because 'common sense' says that form is a thing).  My tendency is to think that things that cannot be tested for have no value...

Fair enough on the first part.

As for them having "no value" . . . I am not so sure. There's plenty of value in undisprovable, untestable propositions. Just don't rely on them to 'bet' (literally or figuratively).  :wave
Ludi Circenses!