News The Anatomy Of A SC Average.

[Rowsus]

One of the things I've always found strange/amusing/bewildering in SC is how quickly and vastly people will change their opinions, and expectations of any given player.
Joe Blog came into the season, and very few people started him in their team. He looked like a 85-90/game plodder (Def), so it was understandable they avoided him. Joe starts the season with a bang, and is sitting on 8/102 after 8 Rounds. Suddenly people are talking like he's a must have, and discussing the best way to get him into their teams. They're treating him, and talking about him, like he is a Premium of at least 2 or 3 years history.
Ok, I understand the why part, as a big part of SC psychology is FOMO (Fear Of Missing Out). It's also one of the things that leads to the most trading mistakes made in SC. The part I really don't understand is the how?
How did people go from thinking he's a 85-90 plodder, to a 102/game must have?
I think the answer lies in the fact, that people don't fully understand the make up, or anatomy, of a SC average. So I thought I would start a thread that details the anatomy of various averages, and once you understand that anatomy, then use your new found knowledge to adjust your expectations accordingly, without wildly jumping 15 to 20 points/game on the back of what turned out to be a mini hot streak.

 

[Rowsus]

I decided to look at 4 averages: 90, 95, 100 and 110.
To increase the number of seasons I could look at, I also looked at averages 1 either side of those averages.
So for example, when I looked at players that averaged 90, I actually looked at every player that averaged between 89 and 91, from 2011 to 2017.

Something very interesting, and to me, unexpected happened. The numbers for the 90 seasons were nearly identical to the 95 seasons, and the numbers for the 100 seasons were very close to exact for the 110 seasons! For this reason we only need to make 2 analyses.

 

[Rowsus]

Players that averaged 90 or 95 from 2011 to 2017

There were 187 players that fitted the criteria I was looking at, and between them they played 3,471 games for me to analyse, or an average of 18.6 games/season.

The top part of the table below shows your expectations broken up into percentage ranges. The 3rd line, the average percentage, has actually been calculated off every score in those different ranges, it is just coincidence that they pretty much fell exactly where you'd expect them to. It then shows how many games the player should have in that range (assuming they play 22 games. You'll have to fiddle for yourselves, where expectation is lower than 22 games). The bottom part gives you examples of how that breaks up in scores, depending on what you expect a player to average.
So if you thought Joe Blog would average 92, then he will typically have:
3 scores lower than 64
2 scores between 64-74
3 scores between 74-83
3 scores between 83-92
3 scores between 92-101
3 scores between 101-110
2 scores between 110-120
3 scores greater than 120
I'm sure many of you will be surprised, that a 92/game player will typically have as many as 5 110+ scores, and that will include 3 120+ scores! Once you realise things like this, you start realise why it is important you don't drastically change your opinion on a player too quickly. Keep in mind, this is not a small sample, it is 187 players, across 3,471 games!

 

[Rowsus]

Players that averaged 100 or 110 from 2011 to 2017

There were 90 players that fitted the criteria I was looking at, and between them they played 1,743 games for me to analyse, or an average of 19.4 games/season. A smaller sample size than the previous one, as expected, but big enough to say it should hold up to scrutiny.

It may surprise some of you, that not only is it expected for a 110/game player to have a score below 66 (that on average is a 53!), but also another score below 77, and another 2 scores on top of that below 88. Those 4 games typically average 70, and yet the player goes onto average 110. In many cases there is a reason for one of two of these poor scores ie. injured during the game etc, but the bottom line is, they are typically there!

 

[Rowsus]

Ok, so now we have a better understanding of how these different season averages are put together, how do we use that information?

We use it to adjust our start of season expectations up and down as the season progresses.

Let's look at a few different examples.

 

[Rowsus]

Dylan Roberton
Roberton's lead up seasons to 2017 went: 20/75, 7/50, 20/91, 18/76 - he had played 102 games by the end of the 2016 season, so it was reasonable to suggest anything he did drastically above 90-92 could be a considered a spike season. I had him pegged at around 90, so let's start with that.

Roberton started: 70, 105, 93, 126, 116, 149 (6/109.8 leaving him 16/82.6 to achieve 90 expectation).
If we lay those scores into a table showing the break up of a typical 90 season it looks like this:

That 149 in Round 6 filled the 2nd of his 3 available top score brackets, and gave him 5 out of 6 scores at the high end. The balance of probabilities say he is likely to overload in the high end from here, and he's closest to doing that at the top score bracket. While many were thinking "Roberton's already scoring at 110, surely he can go at 100 from here", I think that's way too big of a jump on 6 games information. Here's how I suggest you tackle it:
The two scores in the top score bracket total 275, which is 15 points higher than the average top bracket score (130) says they should. The closest bracket to the top score bracket that hasn't hit yet is the 81-90 bracket, that has an average of 86. We are going to let one of those top bracket scores replace one of the expected 81-90 bracket scores. That gives us an adjustment of 130 - 86 = 44, and then we'll add in the 15 points that top score bracket is overperforming by, so total adjustment is 44 + 15 = 59 points.
Original expectation: 22 x 90 = 1,980 points + 59 point adjustment = 2,039 points, or a season average of 2,039 / 22 = 92.7

Roberton has scored 659 points so far, so that leaves the balance of his season at 16/86.3.
So while many were thinking "gotta get Roberton, he's going to go at least 95, probably 100 from here". My suggestion says, stay away, he'll still only average 86 from here.

Now keep in mind, we have basically removed a score from Roberton's top score bracket, and placed it in the 81-90 bracket. We have no need to worry if Roberton scores another one there quickly, as that is the reason we "made room" there. So we adjust Roberton's table to reflect the new 92.7 expectation, and the shifted score, and it looks like this:

 

[Rowsus]

The next 7 Rounds Roberton scored: 96, 84, 119, 89, bye, 83, 119 (season now 12/104.1 leaving him 10/79.0 to get 22/92.7)
His table now looks like this:

Keeping in mind, we are not too concerned if the brackets in the middle get overloaded, as they are right in the wheel house we expected, we do need to look at possibly adjusting again though. The 2nd top bracket has overloaded by 1, and even though there is room in the brackets either side to accomodate this overload, the top 4 brackets have 7 scores to the bottom 4's 5 scores, and the very bottom bracket has no scores at all. Shifting one of the 2nd top score bracket scores to the bottom bracket evens things out nicely. It leaves 4 scores in the top 3 brackets, and enough room for him to score 4 more there in the last 10 Rounds, and gets the bottom bracket moving. The 2nd top bracket has 3 scores averaging 118, and the bottom bracket averages 53, so we need to adjust our total expectation up by:
118 - 53 = 65 + 2,039 = 2,104 as our new season expectation.
2,104 - 1,249 already scored = 855, leaving Roberton looking at 10/85.5 in the last 10 games.
His new table, with the adjusted new expected average of 2,104 / 22 = 95.6 looks like this:

 

[Rowsus]

There was potentially an adjustment that could have been made after Round 18, where Roberton's next scores leading into that were:
80, 60, 79, 101, 91, 82 (season total now 18/96.8 leaving him 4/90.5 to meet our adjusted expectation of 95.6).
It would only be a small adjustment back down again, so we will leave it.
Roberton's final adjusted table looked like this, after we add in those scores, and his last 4 of: 85, 100, 78, 31:

We ended up slightly slanted to the bottom end, and it probably looks better if we hadn't adjusted it all. In fact, here's how it looks, if we just stuck to our guns, and not made any adjustments:

As it turns out, our very first adjustment, where we said Roberton will score 16/86.3 for the rest of the season was just about spot on! He actually went 16/86.1 from that point.
it really rams home the point, that making big adjustments to your expectations, on players with fairly established scoring patterns, is always a dangerous thing to do!

 

[Rowsus]

Rory Sloane

Sloane's a very interesting one to look at for this exercise. He had lots of high and low scores, with no scores at all between 92 - 119 for the entire season! In the seasons leading into 2017 he went: 21/107, 22/115, 16/106, 21/109 and I had him pegged for another 109 season again. In all honesty, if you had pegged much higher than that, then he was, or should have been, in your starting team!

He started the season: 77, 146, 168, 124, 139, 140 making his table look like this:

That 140 make his top 3 brackets filled already after only 6 Rounds. The balance of probabilities says there will be more to come, so we shift that 140 back to the nearest bracket yet hit, which is the 109-120 bracket, and adjust our expectation up.
We started at 22 x 109 = 2,398. The 131-142 bracket has scored a total of 7 points higher than expected, and we are shifting a score back to a bracket with an expectation that is 22 points lower, so we need to adjust Sloane's expectation up by 22 + 7 = 29. Expectation now 2,427 which equals 110.3 game. After starting 6/132.3, that leaves his season expectation at 16/102.1, and his table now looks like this:

 

[Rowsus]

Sloane's next 4 scores went: 80, 70, 119, 177 making his table look like this:

That 177 has overfilled the top score bracket, and needs to be shifted back to the next bracket that "needs it the most", in this case the 99-110 bracket. This gives an adjustment of: The top score bracket has overscored expectations by 11 (2 x expected ave 167 - 168 - 177 = 11), and we are shifting a score back to a bracket that is 167 - 105 = 62 lower, so we need to adjust his expectation up by 62 + 11 = 73. => 2,427 + 73 = 2,500. New expected season average 113.6. Season so far 10/124.0, expectation from here 12/105.0. His table now looks like:

This next bit might make some you think the whole thing is too complicated, or "too special", but you'll surprised how often this type of adjustment system works better than just looking at Rory's early 6/132.3, and just jumping straight up to "Gee, I better get Sloane into my team, he's going to go 120 this season!". Just a side note, if you did think that way, it would still leave him at 16/115.4, not the 120 you have in your head. After Round 6 he was priced at pretty much exactly 115, so he wasn't "value" anyway. I digress, the part you might find too special is, that the shift in the low and high score bracket values, has caused the 2nd lowest bracket to now be overfilled. We need to shift that 77 to the next bracket that "needs it the most". In this case, that is the 91-102 bracket. It's not the lowest bracket for 2 reasons. The low score end of the table already has 3 out of the 4 spots filled. Secondly, the lowest score bracket only needs one to be filled, where the 91-102 still needs three scores to fill it's quota. So we have another adjustment to make, before he's even played another game.
The 2nd lowest bracket has underscored by 174 x 2 - 70 - 77 = 1, and we are shifting a score up to a bracket with an expectation 23 higher, so we need to adjust our expectation down by 23 + 1 = 24. New expectation 2,500 - 24 = 2,476 which is 112.5. The table now looks like this:

 

[Rowsus]

In most circumstances, it will be unusual if you need to make more than 3 or 4 adjustments on a player during season. It might be why Sloane's 2017 season is a good one to look at, as it has a lot more adjustments than we'd normally expect.
Sloane's next 3 scores were: 89, 66, bye, 80 making his table look like this:

The 3rd lowest score bracket has overfilled, so we need to shift that 80 to the bracket that "needs it the most". This open to interpretation, but my thoughts are the 125-136 bracket, though I could accept an argument for the 102-113 bracket. I actually like that there is some possibility for a little variance, otherwise we all follow a similar path, plus the difference in the adjustment is minimal anyway.
The 79-91 bracket has underscored by 3 x 86 - 80 - 89 - 80 = 9. We are shifting a score to a bracket that has a 45 higher expectation, so we need to adjust our expectation down by 45 + 9 = 54. New expectation 2,476 - 49 = 2,427. New season expectation 110.3, he's sitting at 13/113.5, so we expect 9/105.8 from here. The table now looks like this:

 

[Rowsus]

Sloane's next 3 scores were: 90, 91, 50 giving him a table looking like this:

That 50 needs to shift to the 100-111 bracket, and we need to adjust our expectation down again.
The low score bracket has overscored by 53 x 2 - 66 - 50 =10. We are shifting the 50 to a bracket that is 53 higher in expectation, so we need to adjust his expectation down by 53 - 10 = 43. New expectation 2,427 - 43 = 2,384 or 108.4/game. His table now looks like:

The shift in the bracket values has caused the 66 to move from lowest score bracket, to the 2nd lowest, now overloading that bracket again. We need to shift it back down to the lowest again, and make another adjustment.
The 2nd lowest bracket has now underscored by 2 x 71 - 70 - 66 = 6, and we are shifting a score to a bracket with 71 - 52 = 19 lower expectation. So we need to adjust our expectation up by 19 - 6 = 13. New expectation 2,384 + 13 = 2,397 or 109.0/game. So now the table looks like:

He's sitting on 16/106.6, so our expectation for the remainder of the season is 6/115.2, after he scored that 50 Sloan's price fell to $460,500 or close enough to 92/game. If you had a spot left to fill, or an injury to replace, he was underpriced by 23/game on our new expectations.

 

[Rowsus]

I fear the Sloane example might have scared you off the whole concept. You have to trust me in that, Sloanes season with no scores anywhere near his final season average is one out of the box. It is unprecedented for a player to play 18 + games, and have no scores within 9 or 10 of his average.
I reiterate, most players in this system will have 3 or 4 adjustments, if that, and it's rare to have to make 2 adjustments in the one Round, to the same player. That might happen once or twice a season, and it happened to Sloane twice last year!

Sloane finished the season: 169, 36, 151, 92, 140, 130 (6/119.7) just to highlight how up and down his season was! Each of those scores, excluding the 92, and of course the final 130, would have caused an adjustment. Getting that late into the season, some of you might have finished your teams, and not need to keep track of things like this. I don't want to further complicate this example by doing those last 4 adjustments, 3 of which were up, and 1 down.
His final table, if you had followed through and done it all, looked like this:

If you had just filled the original starting table with Sloane's season scores, and made no adjustments at all during the season, his table looks like this:

I think we can easily see how unusual Sloane's season was, so please don't let it sway you from this concept. You won't see another season like it in a long time! Sloane only scored one score in the middle brackets, that theoretically should have contained 8 scores for his season. The chances of this happening under a normal distribution of scores is 1 in 1,656 or around 0.06% of the time.

 

[Rowsus]

You've probably got numbers swirling in your head right now, and are thinking this is just way too complicated. If need be, put this aside for a few days, and come back and read it without even looking at the Sloane example. It will look so much simpler if you do that. I umm'ed and aah'ed about including the Sloane example, and maybe I should have left it out.

I need to do one more example, which will become obvious why in the summary. it's a simple one, I promise.

Remember: come back in 2 or 3 days, and re-read this, without looking at the Sloane posts between posts #9 & #13

I think most of you'll be glad you did, and it will change (hopefully) the way you think about player expectations in the coming season!

 

[Rowsus]

Matt Crouch

Let's say you were interested in Matt Crouch coming into Round 1. You thought he could average 105 in 2017, but you had 3 or 4 midpricers, and you decided Matt Crouch was just one too many, so he missed your starting line up. Coming into 2017 he had scored: 8/59, 15/68, 20/93.
Crouch started the season: 113, 76, 88, 78, 124, 111, 97, 79 - 8/96.0, leaving him 14/110.3 to reach our 105 season expectation. His table looks like this:

The 3rd bottom score bracket has overloaded, and the bottom 3 score brackets have 3 out of 4 spots filled, while the top 3 score brackets have none. The 79 needs to be shifted to a higher scoring bracket. Once again, good arguments can be made for picking two different locations. The 95-105 bracket, and 126-136 bracket. Taking it to the 95-105 bracket would get those middle 2 brackets 50% full, while taking it to the 126-136 bracket only 25% fills that top 3 score area. I choose the 126-136 bracket. That 3rd lowest score bracket has underscored by 3 x 80 - 76 - 78 - 79 = 7. Moving the 79 up to 126-136 bracket is an adjustment down of 131 - 80 = 51, so the total adjustment down on our expectation is 51 + 7 = 58. The new expectation is 2,310 - 58 = 2,252 which equates to 102.4, and leaves him 14/106.1 to meet our new expectation. His price is currently $506,600 which equates to a value of 101/game. His new table now looks like:

 

[Rowsus]

Crouch's next score was an 88. There is no adjustment, but his price has now fallen to $491,700 he's sitting on 9/94.1, and he needs to score 13/107.5 to meet our new season expectation of 102.4. He's now priced at 98.3/game, so he is potentially 9.2/game underpriced.

 

[Rowsus]

Crouch followed this with a 123, and this does lead to an adjustment. His chart now looks like this:

We're going to shift that 123 back to the 102-112 bracket, and make the necessary adjustment. Now keep in mind, when we shifted the 79 to the 123-133 bracket, we mad the necessary adjustment, that for all intents and purposes, changed that 79 to a 128, the average of that bracket. So that bracket has underscored by 128 x 3 - 124 - 128(79) - 123 = 9. We are shifting the score from a bracket that expects 128 to a bracket that expects 107, so the adjustment in total becomes 128 - 107 - 9 (the underscore) = 12. The new expectation is 2,252 + 12 = 2,264 which equates to 102.9. He has scored 10/97.7 so that leaves him 12/107.3 to reach his new expectation. He's now priced at $492,200 which is 98.4/game, and is underpriced by 8.9/game on our expectations. His table now looks like this:

 

[Rowsus]

Crouch's next 3 scores were:
105 - now 11/98.4 and needing 11/107.5 to reach our expectation. Now priced at $503,400 or 101/game.
100 - now 12/98.5 and needing 10/108.2 to reach our expectation. Now priced at $518,100 or 104/game.
bye
131 - now 13/101.0 and needing 9/105.7 to reach the expectation coming into this round. We need to make another adjustment.
The table looks like this:

The 131 needs to shift to either the 93-103 bracket, or the 62-72 bracket. As it stands, the percentages are identical on either shift, so given we don't care if the middle brackets overload, we will shift it to the 93-103 bracket. The 123-134 bracket has underscored by 3, and the brackets have an expectation difference of 31, so we need to adjust our expectation up by 31 - 3 = 28. It now becomes 2,264 + 28 = 2,292, or 104.2/game. He now needs to score 9/108.8 to meet our new expectation, and is priced at $532,200 or 106/game. His table looks like this:

 

[Rowsus]

Crouch's next 4 scores were: 132, 116, 104, 134.
The 134 overfills that 3rd top score bracket again. We need to shift it to 63-73 bracket, and make the necessary adjustment. The 125-135 bracket has overscored by 6, and we are shifting the score to a bracket 62 lower, so we need to adjust the expectation up 62 + 6 = 68. The expectation is now 2,292 + 68 = 2,360 which is 107.3/game. He has scored 17/105.8 and needs 5/112.2 to meet our new expectation. He's now priced at $582,000, or 116/game, and his adjusted table looks like this:

The shift in the bracket values has caused the middle brackets to overfill, but as stated earlier, we aren't worried about them overfilling, as they represent what we expect anyway.

 

[Rowsus]

Crouch finished the season: 120, 139, 139, 91, 146 for a 5/127 finish. Getting to round 18 and beyond is too late to worry about showing any more adjustments.
Crouch's original table would have looked like this, if we made no adjustments along the way.